00:00:00,560 --> 00:00:03,669

All right. I am Will Serber. Um, and it

00:00:03,669 --> 00:00:03,679
All right. I am Will Serber. Um, and it
 

00:00:03,679 --> 00:00:06,470
All right. I am Will Serber. Um, and it
is my uh my great pleasure to be

00:00:06,470 --> 00:00:06,480
is my uh my great pleasure to be
 

00:00:06,480 --> 00:00:09,190
is my uh my great pleasure to be
introducing uh this section and giving

00:00:09,190 --> 00:00:09,200
introducing uh this section and giving
 

00:00:09,200 --> 00:00:11,830
introducing uh this section and giving
you a little bit of background uh before

00:00:11,830 --> 00:00:11,840
you a little bit of background uh before
 

00:00:11,840 --> 00:00:14,070
you a little bit of background uh before
turning it over to friends from OpenAI

00:00:14,070 --> 00:00:14,080
turning it over to friends from OpenAI
 

00:00:14,080 --> 00:00:16,710
turning it over to friends from OpenAI
and uh Pacific Northwest National Labs.

00:00:16,710 --> 00:00:16,720
and uh Pacific Northwest National Labs.
 

00:00:16,720 --> 00:00:19,590
and uh Pacific Northwest National Labs.
Um I uh you know the title of the talk

00:00:19,590 --> 00:00:19,600
Um I uh you know the title of the talk
 

00:00:19,600 --> 00:00:21,109
Um I uh you know the title of the talk
here designing, deploying and scaling

00:00:21,109 --> 00:00:21,119
here designing, deploying and scaling
 

00:00:21,119 --> 00:00:23,750
here designing, deploying and scaling
autonomous labs with OpenAI and PNL. Um

00:00:23,750 --> 00:00:23,760
autonomous labs with OpenAI and PNL. Um
 

00:00:23,760 --> 00:00:26,790
autonomous labs with OpenAI and PNL. Um
I'm the head of automation at GKO uh and

00:00:26,790 --> 00:00:26,800
I'm the head of automation at GKO uh and
 

00:00:26,800 --> 00:00:28,550
I'm the head of automation at GKO uh and
uh have been there for a handful of

00:00:28,550 --> 00:00:28,560
uh have been there for a handful of
 

00:00:28,560 --> 00:00:30,790
uh have been there for a handful of
years. Um there's a disclaimer of course

00:00:30,790 --> 00:00:30,800
years. Um there's a disclaimer of course
 

00:00:30,800 --> 00:00:32,870
years. Um there's a disclaimer of course
this is required. Um I've been uh I've

00:00:32,870 --> 00:00:32,880
this is required. Um I've been uh I've
 

00:00:32,880 --> 00:00:34,549
this is required. Um I've been uh I've
been in GKO for about three years going

00:00:34,549 --> 00:00:34,559
been in GKO for about three years going
 

00:00:34,559 --> 00:00:36,150
been in GKO for about three years going
backwards on this. Before that I was

00:00:36,150 --> 00:00:36,160
backwards on this. Before that I was
 

00:00:36,160 --> 00:00:38,950
backwards on this. Before that I was
head of automation at Zyrgen. Before

00:00:38,950 --> 00:00:38,960
head of automation at Zyrgen. Before
 

00:00:38,960 --> 00:00:40,470
head of automation at Zyrgen. Before
that I was a commercial photographer for

00:00:40,470 --> 00:00:40,480
that I was a commercial photographer for
 

00:00:40,480 --> 00:00:42,389
that I was a commercial photographer for
5 years. Uh not very helpful for my

00:00:42,389 --> 00:00:42,399
5 years. Uh not very helpful for my
 

00:00:42,399 --> 00:00:43,750
5 years. Uh not very helpful for my
current career. Before that I was an

00:00:43,750 --> 00:00:43,760
current career. Before that I was an
 

00:00:43,760 --> 00:00:45,510
current career. Before that I was an
astrophysicist for a while. Again not

00:00:45,510 --> 00:00:45,520
astrophysicist for a while. Again not
 

00:00:45,520 --> 00:00:48,150
astrophysicist for a while. Again not
the most helpful. Um but uh it's been

00:00:48,150 --> 00:00:48,160
the most helpful. Um but uh it's been
 

00:00:48,160 --> 00:00:49,750
the most helpful. Um but uh it's been
fun combining all these things as much

00:00:49,750 --> 00:00:49,760
fun combining all these things as much
 

00:00:49,760 --> 00:00:51,670
fun combining all these things as much
as I can into into one interesting

00:00:51,670 --> 00:00:51,680
as I can into into one interesting
 

00:00:51,680 --> 00:00:53,670
as I can into into one interesting
career. I've got a dog in the bottom

00:00:53,670 --> 00:00:53,680
career. I've got a dog in the bottom
 

00:00:53,680 --> 00:00:55,750
career. I've got a dog in the bottom
right hand corner named Soba. And uh

00:00:55,750 --> 00:00:55,760
right hand corner named Soba. And uh
 

00:00:55,760 --> 00:00:57,430
right hand corner named Soba. And uh
that's her looking very bored in our

00:00:57,430 --> 00:00:57,440
that's her looking very bored in our
 

00:00:57,440 --> 00:00:59,990
that's her looking very bored in our
workshop where we make racks.

00:00:59,990 --> 00:01:00,000
workshop where we make racks.
 

00:01:00,000 --> 00:01:02,790
workshop where we make racks.
So the mission of the company is to make

00:01:02,790 --> 00:01:02,800
So the mission of the company is to make
 

00:01:02,800 --> 00:01:05,750
So the mission of the company is to make
biology easier to engineer. And we've

00:01:05,750 --> 00:01:05,760
biology easier to engineer. And we've
 

00:01:05,760 --> 00:01:07,990
biology easier to engineer. And we've
got about 400 or so employees, a little

00:01:07,990 --> 00:01:08,000
got about 400 or so employees, a little
 

00:01:08,000 --> 00:01:10,469
got about 400 or so employees, a little
north of that. And we're expecting about

00:01:10,469 --> 00:01:10,479
north of that. And we're expecting about
 

00:01:10,479 --> 00:01:12,550
north of that. And we're expecting about
200 million in revenue this year just to

00:01:12,550 --> 00:01:12,560
200 million in revenue this year just to
 

00:01:12,560 --> 00:01:14,550
200 million in revenue this year just to
tell you a tiny bit about GKO. And we're

00:01:14,550 --> 00:01:14,560
tell you a tiny bit about GKO. And we're
 

00:01:14,560 --> 00:01:16,469
tell you a tiny bit about GKO. And we're
really focusing more and more on

00:01:16,469 --> 00:01:16,479
really focusing more and more on
 

00:01:16,479 --> 00:01:18,550
really focusing more and more on
automation.

00:01:18,550 --> 00:01:18,560
automation.
 

00:01:18,560 --> 00:01:21,350
automation.
So the talk is organized really uh there

00:01:21,350 --> 00:01:21,360
So the talk is organized really uh there
 

00:01:21,360 --> 00:01:23,510
So the talk is organized really uh there
are three fundamental segments to it.

00:01:23,510 --> 00:01:23,520
are three fundamental segments to it.
 

00:01:23,520 --> 00:01:26,870
are three fundamental segments to it.
Um, one just a uh description of our

00:01:26,870 --> 00:01:26,880
Um, one just a uh description of our
 

00:01:26,880 --> 00:01:28,469
Um, one just a uh description of our
belief that autonomous labs will

00:01:28,469 --> 00:01:28,479
belief that autonomous labs will
 

00:01:28,479 --> 00:01:30,630
belief that autonomous labs will
transform biotechnology.

00:01:30,630 --> 00:01:30,640
transform biotechnology.
 

00:01:30,640 --> 00:01:31,749
transform biotechnology.
Second, I'm going to tell you a little

00:01:31,749 --> 00:01:31,759
Second, I'm going to tell you a little
 

00:01:31,759 --> 00:01:33,190
Second, I'm going to tell you a little
bit about what we think those labs might

00:01:33,190 --> 00:01:33,200
bit about what we think those labs might
 

00:01:33,200 --> 00:01:35,990
bit about what we think those labs might
look like. And then third, just a super

00:01:35,990 --> 00:01:36,000
look like. And then third, just a super
 

00:01:36,000 --> 00:01:38,149
look like. And then third, just a super
quick word on kind of how we go to

00:01:38,149 --> 00:01:38,159
quick word on kind of how we go to
 

00:01:38,159 --> 00:01:41,270
quick word on kind of how we go to
market with that in practice.

00:01:41,270 --> 00:01:41,280
market with that in practice.
 

00:01:41,280 --> 00:01:43,830
market with that in practice.
Starting with the first. So, this is an

00:01:43,830 --> 00:01:43,840
Starting with the first. So, this is an
 

00:01:43,840 --> 00:01:46,230
Starting with the first. So, this is an
image uh you might have seen Jason uh

00:01:46,230 --> 00:01:46,240
image uh you might have seen Jason uh
 

00:01:46,240 --> 00:01:48,469
image uh you might have seen Jason uh
our CEO show off, but I do quite like

00:01:48,469 --> 00:01:48,479
our CEO show off, but I do quite like
 

00:01:48,479 --> 00:01:51,590
our CEO show off, but I do quite like
it. So this is an old ad for IBM in

00:01:51,590 --> 00:01:51,600
it. So this is an old ad for IBM in
 

00:01:51,600 --> 00:01:53,830
it. So this is an old ad for IBM in
which they were saying, you know, one of

00:01:53,830 --> 00:01:53,840
which they were saying, you know, one of
 

00:01:53,840 --> 00:01:57,030
which they were saying, you know, one of
their computers will place 150 engineers

00:01:57,030 --> 00:01:57,040
their computers will place 150 engineers
 

00:01:57,040 --> 00:01:58,870
their computers will place 150 engineers
with slide rules. So there's a couple

00:01:58,870 --> 00:01:58,880
with slide rules. So there's a couple
 

00:01:58,880 --> 00:02:02,630
with slide rules. So there's a couple
reasons I really like this. Uh one is um

00:02:02,630 --> 00:02:02,640
reasons I really like this. Uh one is um
 

00:02:02,640 --> 00:02:04,630
reasons I really like this. Uh one is um
you know, very clearly this is uh you

00:02:04,630 --> 00:02:04,640
you know, very clearly this is uh you
 

00:02:04,640 --> 00:02:05,830
you know, very clearly this is uh you
know, someone might look at this and

00:02:05,830 --> 00:02:05,840
know, someone might look at this and
 

00:02:05,840 --> 00:02:07,670
know, someone might look at this and
think like, oh my god, our jobs as

00:02:07,670 --> 00:02:07,680
think like, oh my god, our jobs as
 

00:02:07,680 --> 00:02:09,749
think like, oh my god, our jobs as
engineers are at risk as a result of

00:02:09,749 --> 00:02:09,759
engineers are at risk as a result of
 

00:02:09,759 --> 00:02:11,830
engineers are at risk as a result of
this technology. But we all know that

00:02:11,830 --> 00:02:11,840
this technology. But we all know that
 

00:02:11,840 --> 00:02:14,550
this technology. But we all know that
there's way way way more software

00:02:14,550 --> 00:02:14,560
there's way way way more software
 

00:02:14,560 --> 00:02:16,390
there's way way way more software
engineers today in the world uh than

00:02:16,390 --> 00:02:16,400
engineers today in the world uh than
 

00:02:16,400 --> 00:02:18,630
engineers today in the world uh than
there were back then. So that's the way

00:02:18,630 --> 00:02:18,640
there were back then. So that's the way
 

00:02:18,640 --> 00:02:20,790
there were back then. So that's the way
I feel about um automation in the lab

00:02:20,790 --> 00:02:20,800
I feel about um automation in the lab
 

00:02:20,800 --> 00:02:22,710
I feel about um automation in the lab
space too. And so it's nice to see this

00:02:22,710 --> 00:02:22,720
space too. And so it's nice to see this
 

00:02:22,720 --> 00:02:25,030
space too. And so it's nice to see this
uh fantastic example. The other thing

00:02:25,030 --> 00:02:25,040
uh fantastic example. The other thing
 

00:02:25,040 --> 00:02:27,110
uh fantastic example. The other thing
that's kind of interesting uh sort of

00:02:27,110 --> 00:02:27,120
that's kind of interesting uh sort of
 

00:02:27,120 --> 00:02:29,510
that's kind of interesting uh sort of
represented on this slide is that it's

00:02:29,510 --> 00:02:29,520
represented on this slide is that it's
 

00:02:29,520 --> 00:02:30,710
represented on this slide is that it's
these two things are not really

00:02:30,710 --> 00:02:30,720
these two things are not really
 

00:02:30,720 --> 00:02:33,190
these two things are not really
equivalent. The slide rule is extremely

00:02:33,190 --> 00:02:33,200
equivalent. The slide rule is extremely
 

00:02:33,200 --> 00:02:35,350
equivalent. The slide rule is extremely
flexible but not automated. You need a

00:02:35,350 --> 00:02:35,360
flexible but not automated. You need a
 

00:02:35,360 --> 00:02:37,589
flexible but not automated. You need a
human to run it. Uh it feels similar to

00:02:37,589 --> 00:02:37,599
human to run it. Uh it feels similar to
 

00:02:37,599 --> 00:02:39,670
human to run it. Uh it feels similar to
a pipet in that sense. And then in the

00:02:39,670 --> 00:02:39,680
a pipet in that sense. And then in the
 

00:02:39,680 --> 00:02:41,589
a pipet in that sense. And then in the
top lefthand corner you have the things

00:02:41,589 --> 00:02:41,599
top lefthand corner you have the things
 

00:02:41,599 --> 00:02:43,750
top lefthand corner you have the things
that IBM was selling which at the time

00:02:43,750 --> 00:02:43,760
that IBM was selling which at the time
 

00:02:43,760 --> 00:02:45,830
that IBM was selling which at the time
were not particularly general purpose

00:02:45,830 --> 00:02:45,840
were not particularly general purpose
 

00:02:45,840 --> 00:02:47,990
were not particularly general purpose
computers uh but they were heavily

00:02:47,990 --> 00:02:48,000
computers uh but they were heavily
 

00:02:48,000 --> 00:02:50,470
computers uh but they were heavily
automated and it was really the advent

00:02:50,470 --> 00:02:50,480
automated and it was really the advent
 

00:02:50,480 --> 00:02:51,910
automated and it was really the advent
of something in the top right hand

00:02:51,910 --> 00:02:51,920
of something in the top right hand
 

00:02:51,920 --> 00:02:53,270
of something in the top right hand
corner the you know the personal

00:02:53,270 --> 00:02:53,280
corner the you know the personal
 

00:02:53,280 --> 00:02:56,229
corner the you know the personal
computer that transformed the world and

00:02:56,229 --> 00:02:56,239
computer that transformed the world and
 

00:02:56,239 --> 00:02:58,949
computer that transformed the world and
is why we have billions of computers uh

00:02:58,949 --> 00:02:58,959
is why we have billions of computers uh
 

00:02:58,959 --> 00:03:02,229
is why we have billions of computers uh
in in the world today. Um so that high

00:03:02,229 --> 00:03:02,239
in in the world today. Um so that high
 

00:03:02,239 --> 00:03:04,949
in in the world today. Um so that high
automation, high flexibility is kind of

00:03:04,949 --> 00:03:04,959
automation, high flexibility is kind of
 

00:03:04,959 --> 00:03:07,350
automation, high flexibility is kind of
the the golden spot to be um where you

00:03:07,350 --> 00:03:07,360
the the golden spot to be um where you
 

00:03:07,360 --> 00:03:09,030
the the golden spot to be um where you
can really have a transformative effect

00:03:09,030 --> 00:03:09,040
can really have a transformative effect
 

00:03:09,040 --> 00:03:12,309
can really have a transformative effect
on the industry.

00:03:12,309 --> 00:03:12,319
on the industry.
 

00:03:12,319 --> 00:03:14,070
on the industry.
Continuing this analogy a little bit to

00:03:14,070 --> 00:03:14,080
Continuing this analogy a little bit to
 

00:03:14,080 --> 00:03:16,949
Continuing this analogy a little bit to
transportation. Um you can imagine uh

00:03:16,949 --> 00:03:16,959
transportation. Um you can imagine uh
 

00:03:16,959 --> 00:03:19,030
transportation. Um you can imagine uh
trains or subways being in the top

00:03:19,030 --> 00:03:19,040
trains or subways being in the top
 

00:03:19,040 --> 00:03:21,670
trains or subways being in the top
lefthand corner again uh where totally

00:03:21,670 --> 00:03:21,680
lefthand corner again uh where totally
 

00:03:21,680 --> 00:03:23,509
lefthand corner again uh where totally
automated but you have to go where the

00:03:23,509 --> 00:03:23,519
automated but you have to go where the
 

00:03:23,519 --> 00:03:25,589
automated but you have to go where the
rails are laid or a car in the bottom

00:03:25,589 --> 00:03:25,599
rails are laid or a car in the bottom
 

00:03:25,599 --> 00:03:27,110
rails are laid or a car in the bottom
right hand corner. And this is the way

00:03:27,110 --> 00:03:27,120
right hand corner. And this is the way
 

00:03:27,120 --> 00:03:29,110
right hand corner. And this is the way
the world has been for a long time. Like

00:03:29,110 --> 00:03:29,120
the world has been for a long time. Like
 

00:03:29,120 --> 00:03:30,949
the world has been for a long time. Like
you basically had these two choices. And

00:03:30,949 --> 00:03:30,959
you basically had these two choices. And
 

00:03:30,959 --> 00:03:32,710
you basically had these two choices. And
we're starting to see the emergence of

00:03:32,710 --> 00:03:32,720
we're starting to see the emergence of
 

00:03:32,720 --> 00:03:34,309
we're starting to see the emergence of
something in the top right hand corner.

00:03:34,309 --> 00:03:34,319
something in the top right hand corner.
 

00:03:34,319 --> 00:03:36,229
something in the top right hand corner.
Whimo being the best example where

00:03:36,229 --> 00:03:36,239
Whimo being the best example where
 

00:03:36,239 --> 00:03:38,149
Whimo being the best example where
within their constrained environment it

00:03:38,149 --> 00:03:38,159
within their constrained environment it
 

00:03:38,159 --> 00:03:40,390
within their constrained environment it
is both autonomous and flexible. And it

00:03:40,390 --> 00:03:40,400
is both autonomous and flexible. And it
 

00:03:40,400 --> 00:03:41,830
is both autonomous and flexible. And it
will be interesting to see how this

00:03:41,830 --> 00:03:41,840
will be interesting to see how this
 

00:03:41,840 --> 00:03:43,270
will be interesting to see how this
changes the world and the way that

00:03:43,270 --> 00:03:43,280
changes the world and the way that
 

00:03:43,280 --> 00:03:44,789
changes the world and the way that
cities are laid out and people lead

00:03:44,789 --> 00:03:44,799
cities are laid out and people lead
 

00:03:44,799 --> 00:03:47,030
cities are laid out and people lead
their lives. I suspect this will be a

00:03:47,030 --> 00:03:47,040
their lives. I suspect this will be a
 

00:03:47,040 --> 00:03:49,350
their lives. I suspect this will be a
you know quite a transformation as well

00:03:49,350 --> 00:03:49,360
you know quite a transformation as well
 

00:03:49,360 --> 00:03:50,949
you know quite a transformation as well
which is also why so much money pours

00:03:50,949 --> 00:03:50,959
which is also why so much money pours
 

00:03:50,959 --> 00:03:53,350
which is also why so much money pours
into that space.

00:03:53,350 --> 00:03:53,360
into that space.
 

00:03:53,360 --> 00:03:55,750
into that space.
In the lab space, we again have things

00:03:55,750 --> 00:03:55,760
In the lab space, we again have things
 

00:03:55,760 --> 00:03:57,910
In the lab space, we again have things
in these two corners. We have the lab

00:03:57,910 --> 00:03:57,920
in these two corners. We have the lab
 

00:03:57,920 --> 00:04:00,869
in these two corners. We have the lab
bench, extremely flexible, but not a lot

00:04:00,869 --> 00:04:00,879
bench, extremely flexible, but not a lot
 

00:04:00,879 --> 00:04:02,789
bench, extremely flexible, but not a lot
of automation, no automation. And then

00:04:02,789 --> 00:04:02,799
of automation, no automation. And then
 

00:04:02,799 --> 00:04:04,390
of automation, no automation. And then
you have the traditional work cells in

00:04:04,390 --> 00:04:04,400
you have the traditional work cells in
 

00:04:04,400 --> 00:04:06,630
you have the traditional work cells in
the top lefthand corner uh filling this

00:04:06,630 --> 00:04:06,640
the top lefthand corner uh filling this
 

00:04:06,640 --> 00:04:09,429
the top lefthand corner uh filling this
out. And to date, there's basically been

00:04:09,429 --> 00:04:09,439
out. And to date, there's basically been
 

00:04:09,439 --> 00:04:11,429
out. And to date, there's basically been
nothing in the top right hand corner.

00:04:11,429 --> 00:04:11,439
nothing in the top right hand corner.
 

00:04:11,439 --> 00:04:13,830
nothing in the top right hand corner.
And so really what GKO is going after is

00:04:13,830 --> 00:04:13,840
And so really what GKO is going after is
 

00:04:13,840 --> 00:04:15,830
And so really what GKO is going after is
that top right hand corner. We don't

00:04:15,830 --> 00:04:15,840
that top right hand corner. We don't
 

00:04:15,840 --> 00:04:17,830
that top right hand corner. We don't
view ourselves, even though we have lab

00:04:17,830 --> 00:04:17,840
view ourselves, even though we have lab
 

00:04:17,840 --> 00:04:19,430
view ourselves, even though we have lab
equipment. That's what we're selling. We

00:04:19,430 --> 00:04:19,440
equipment. That's what we're selling. We
 

00:04:19,440 --> 00:04:20,949
equipment. That's what we're selling. We
don't view ourselves as competing

00:04:20,949 --> 00:04:20,959
don't view ourselves as competing
 

00:04:20,959 --> 00:04:24,070
don't view ourselves as competing
fundamentally with the work cell per se.

00:04:24,070 --> 00:04:24,080
fundamentally with the work cell per se.
 

00:04:24,080 --> 00:04:25,990
fundamentally with the work cell per se.
It's competing as much with the bench as

00:04:25,990 --> 00:04:26,000
It's competing as much with the bench as
 

00:04:26,000 --> 00:04:28,710
It's competing as much with the bench as
the work cell. You know, we are uh very

00:04:28,710 --> 00:04:28,720
the work cell. You know, we are uh very
 

00:04:28,720 --> 00:04:31,189
the work cell. You know, we are uh very
much looking to transform the way that

00:04:31,189 --> 00:04:31,199
much looking to transform the way that
 

00:04:31,199 --> 00:04:33,110
much looking to transform the way that
people do science in the way that the

00:04:33,110 --> 00:04:33,120
people do science in the way that the
 

00:04:33,120 --> 00:04:35,270
people do science in the way that the
computer did or the way that perhaps

00:04:35,270 --> 00:04:35,280
computer did or the way that perhaps
 

00:04:35,280 --> 00:04:38,230
computer did or the way that perhaps
autonomous driving will.

00:04:38,230 --> 00:04:38,240
autonomous driving will.
 

00:04:38,240 --> 00:04:40,150
autonomous driving will.
So, this raises this question of okay,

00:04:40,150 --> 00:04:40,160
So, this raises this question of okay,
 

00:04:40,160 --> 00:04:41,430
So, this raises this question of okay,
how do you get there? How do you

00:04:41,430 --> 00:04:41,440
how do you get there? How do you
 

00:04:41,440 --> 00:04:43,830
how do you get there? How do you
actually build that?

00:04:43,830 --> 00:04:43,840
actually build that?
 

00:04:43,840 --> 00:04:45,270
actually build that?
Part two.

00:04:45,270 --> 00:04:45,280
Part two.
 

00:04:45,280 --> 00:04:48,390
Part two.
So, one of our key beliefs is that an

00:04:48,390 --> 00:04:48,400
So, one of our key beliefs is that an
 

00:04:48,400 --> 00:04:50,629
So, one of our key beliefs is that an
autonomous lab must allow for the same

00:04:50,629 --> 00:04:50,639
autonomous lab must allow for the same
 

00:04:50,639 --> 00:04:53,350
autonomous lab must allow for the same
work you do in a traditional lab. So,

00:04:53,350 --> 00:04:53,360
work you do in a traditional lab. So,
 

00:04:53,360 --> 00:04:54,790
work you do in a traditional lab. So,
you need to be able to do all those same

00:04:54,790 --> 00:04:54,800
you need to be able to do all those same
 

00:04:54,800 --> 00:04:56,629
you need to be able to do all those same
things. You can't be leaving out key

00:04:56,629 --> 00:04:56,639
things. You can't be leaving out key
 

00:04:56,639 --> 00:04:58,870
things. You can't be leaving out key
parts of the workflow.

00:04:58,870 --> 00:04:58,880
parts of the workflow.
 

00:04:58,880 --> 00:05:00,550
parts of the workflow.
There are, this is the wordiest slide. I

00:05:00,550 --> 00:05:00,560
There are, this is the wordiest slide. I
 

00:05:00,560 --> 00:05:02,150
There are, this is the wordiest slide. I
apologize for this one, but there are

00:05:02,150 --> 00:05:02,160
apologize for this one, but there are
 

00:05:02,160 --> 00:05:04,310
apologize for this one, but there are
basically six things that we think are

00:05:04,310 --> 00:05:04,320
basically six things that we think are
 

00:05:04,320 --> 00:05:06,550
basically six things that we think are
critical to replicate uh to get to that

00:05:06,550 --> 00:05:06,560
critical to replicate uh to get to that
 

00:05:06,560 --> 00:05:08,629
critical to replicate uh to get to that
top right hand corner. So, we think you

00:05:08,629 --> 00:05:08,639
top right hand corner. So, we think you
 

00:05:08,639 --> 00:05:10,790
top right hand corner. So, we think you
need reliable liquid handling that is

00:05:10,790 --> 00:05:10,800
need reliable liquid handling that is
 

00:05:10,800 --> 00:05:12,629
need reliable liquid handling that is
easily programmable. We called this one

00:05:12,629 --> 00:05:12,639
easily programmable. We called this one
 

00:05:12,639 --> 00:05:14,390
easily programmable. We called this one
out just because liquid handling is so

00:05:14,390 --> 00:05:14,400
out just because liquid handling is so
 

00:05:14,400 --> 00:05:16,070
out just because liquid handling is so
much at the core of so many different

00:05:16,070 --> 00:05:16,080
much at the core of so many different
 

00:05:16,080 --> 00:05:18,550
much at the core of so many different
workflows and so it feels like it worth

00:05:18,550 --> 00:05:18,560
workflows and so it feels like it worth
 

00:05:18,560 --> 00:05:21,430
workflows and so it feels like it worth
special mention. Number two, you need

00:05:21,430 --> 00:05:21,440
special mention. Number two, you need
 

00:05:21,440 --> 00:05:23,590
special mention. Number two, you need
reliable material transport to any

00:05:23,590 --> 00:05:23,600
reliable material transport to any
 

00:05:23,600 --> 00:05:25,430
reliable material transport to any
device. The reliable part is the hard

00:05:25,430 --> 00:05:25,440
device. The reliable part is the hard
 

00:05:25,440 --> 00:05:27,029
device. The reliable part is the hard
part there. Like lots of people have

00:05:27,029 --> 00:05:27,039
part there. Like lots of people have
 

00:05:27,039 --> 00:05:29,029
part there. Like lots of people have
robot arms, but they're not really

00:05:29,029 --> 00:05:29,039
robot arms, but they're not really
 

00:05:29,039 --> 00:05:32,070
robot arms, but they're not really
reliable enough to get to where we want.

00:05:32,070 --> 00:05:32,080
reliable enough to get to where we want.
 

00:05:32,080 --> 00:05:34,469
reliable enough to get to where we want.
Uh number three is parameterized control

00:05:34,469 --> 00:05:34,479
Uh number three is parameterized control
 

00:05:34,479 --> 00:05:36,390
Uh number three is parameterized control
of devices. You need, you know, a good

00:05:36,390 --> 00:05:36,400
of devices. You need, you know, a good
 

00:05:36,400 --> 00:05:38,870
of devices. You need, you know, a good
API. you need to be able to do many

00:05:38,870 --> 00:05:38,880
API. you need to be able to do many
 

00:05:38,880 --> 00:05:40,469
API. you need to be able to do many
things with the device and get kind of

00:05:40,469 --> 00:05:40,479
things with the device and get kind of
 

00:05:40,479 --> 00:05:43,990
things with the device and get kind of
full control over it. Uh fourth is you

00:05:43,990 --> 00:05:44,000
full control over it. Uh fourth is you
 

00:05:44,000 --> 00:05:46,390
full control over it. Uh fourth is you
need support for as many devices in a

00:05:46,390 --> 00:05:46,400
need support for as many devices in a
 

00:05:46,400 --> 00:05:48,230
need support for as many devices in a
single setup as you would have in a

00:05:48,230 --> 00:05:48,240
single setup as you would have in a
 

00:05:48,240 --> 00:05:50,469
single setup as you would have in a
traditional lab. And so we think, you

00:05:50,469 --> 00:05:50,479
traditional lab. And so we think, you
 

00:05:50,479 --> 00:05:52,070
traditional lab. And so we think, you
know, you don't want to have lots of

00:05:52,070 --> 00:05:52,080
know, you don't want to have lots of
 

00:05:52,080 --> 00:05:53,990
know, you don't want to have lots of
different work cells with humans walking

00:05:53,990 --> 00:05:54,000
different work cells with humans walking
 

00:05:54,000 --> 00:05:55,749
different work cells with humans walking
between them. Ideally, you don't want to

00:05:55,749 --> 00:05:55,759
between them. Ideally, you don't want to
 

00:05:55,759 --> 00:05:57,189
between them. Ideally, you don't want to
be missing pieces of equipment that are

00:05:57,189 --> 00:05:57,199
be missing pieces of equipment that are
 

00:05:57,199 --> 00:05:59,189
be missing pieces of equipment that are
on the bench in the idealized form of

00:05:59,189 --> 00:05:59,199
on the bench in the idealized form of
 

00:05:59,199 --> 00:06:00,710
on the bench in the idealized form of
this. Fortunately, there's, you know,

00:06:00,710 --> 00:06:00,720
this. Fortunately, there's, you know,
 

00:06:00,720 --> 00:06:03,029
this. Fortunately, there's, you know,
there's value to be, uh, to be gotten

00:06:03,029 --> 00:06:03,039
there's value to be, uh, to be gotten
 

00:06:03,039 --> 00:06:05,430
there's value to be, uh, to be gotten
along the path to this vision, but this

00:06:05,430 --> 00:06:05,440
along the path to this vision, but this
 

00:06:05,440 --> 00:06:07,749
along the path to this vision, but this
is the this is the ultimate vision. Um,

00:06:07,749 --> 00:06:07,759
is the this is the ultimate vision. Um,
 

00:06:07,759 --> 00:06:10,629
is the this is the ultimate vision. Um,
fifth, uh, support for parallel work.

00:06:10,629 --> 00:06:10,639
fifth, uh, support for parallel work.
 

00:06:10,639 --> 00:06:12,390
fifth, uh, support for parallel work.
So, if you imagine a traditional lab

00:06:12,390 --> 00:06:12,400
So, if you imagine a traditional lab
 

00:06:12,400 --> 00:06:14,390
So, if you imagine a traditional lab
with a bunch of grad students or RAS

00:06:14,390 --> 00:06:14,400
with a bunch of grad students or RAS
 

00:06:14,400 --> 00:06:17,029
with a bunch of grad students or RAS
running around in it, like they are

00:06:17,029 --> 00:06:17,039
running around in it, like they are
 

00:06:17,039 --> 00:06:18,710
running around in it, like they are
coordinating constantly to share

00:06:18,710 --> 00:06:18,720
coordinating constantly to share
 

00:06:18,720 --> 00:06:20,870
coordinating constantly to share
equipment and to make the use the best

00:06:20,870 --> 00:06:20,880
equipment and to make the use the best
 

00:06:20,880 --> 00:06:23,110
equipment and to make the use the best
use out of the the capital expenditure.

00:06:23,110 --> 00:06:23,120
use out of the the capital expenditure.
 

00:06:23,120 --> 00:06:25,749
use out of the the capital expenditure.
And so we believe that really to uh

00:06:25,749 --> 00:06:25,759
And so we believe that really to uh
 

00:06:25,759 --> 00:06:27,110
And so we believe that really to uh
achieve what we want to achieve in the

00:06:27,110 --> 00:06:27,120
achieve what we want to achieve in the
 

00:06:27,120 --> 00:06:29,590
achieve what we want to achieve in the
lab, you need to support tens to

00:06:29,590 --> 00:06:29,600
lab, you need to support tens to
 

00:06:29,600 --> 00:06:31,909
lab, you need to support tens to
hundreds of parallel protocol requests

00:06:31,909 --> 00:06:31,919
hundreds of parallel protocol requests
 

00:06:31,919 --> 00:06:33,990
hundreds of parallel protocol requests
from scientists and critically ones that

00:06:33,990 --> 00:06:34,000
from scientists and critically ones that
 

00:06:34,000 --> 00:06:35,909
from scientists and critically ones that
change daily. It's not the same, you

00:06:35,909 --> 00:06:35,919
change daily. It's not the same, you
 

00:06:35,919 --> 00:06:37,350
change daily. It's not the same, you
know, 10 protocols they're going to be

00:06:37,350 --> 00:06:37,360
know, 10 protocols they're going to be
 

00:06:37,360 --> 00:06:38,870
know, 10 protocols they're going to be
running over and over again. They're

00:06:38,870 --> 00:06:38,880
running over and over again. They're
 

00:06:38,880 --> 00:06:40,390
running over and over again. They're
going to want to push the frontiers of

00:06:40,390 --> 00:06:40,400
going to want to push the frontiers of
 

00:06:40,400 --> 00:06:42,070
going to want to push the frontiers of
what is possible and to be running

00:06:42,070 --> 00:06:42,080
what is possible and to be running
 

00:06:42,080 --> 00:06:45,189
what is possible and to be running
different uh protocols. And then lastly,

00:06:45,189 --> 00:06:45,199
different uh protocols. And then lastly,
 

00:06:45,199 --> 00:06:47,510
different uh protocols. And then lastly,
uh it should say an experience as easy

00:06:47,510 --> 00:06:47,520
uh it should say an experience as easy
 

00:06:47,520 --> 00:06:49,670
uh it should say an experience as easy
for scientists to use as the bench. So,

00:06:49,670 --> 00:06:49,680
for scientists to use as the bench. So,
 

00:06:49,680 --> 00:06:51,430
for scientists to use as the bench. So,
we think we really, really need to lower

00:06:51,430 --> 00:06:51,440
we think we really, really need to lower
 

00:06:51,440 --> 00:06:52,950
we think we really, really need to lower
the barrier to entry to using these

00:06:52,950 --> 00:06:52,960
the barrier to entry to using these
 

00:06:52,960 --> 00:06:55,029
the barrier to entry to using these
tools.

00:06:55,029 --> 00:06:55,039
tools.
 

00:06:55,039 --> 00:06:56,309
tools.
I'm going to tell you just a little bit

00:06:56,309 --> 00:06:56,319
I'm going to tell you just a little bit
 

00:06:56,319 --> 00:06:58,070
I'm going to tell you just a little bit
about our solution. If you're excited to

00:06:58,070 --> 00:06:58,080
about our solution. If you're excited to
 

00:06:58,080 --> 00:06:59,830
about our solution. If you're excited to
see more, you should come to our booth.

00:06:59,830 --> 00:06:59,840
see more, you should come to our booth.
 

00:06:59,840 --> 00:07:01,990
see more, you should come to our booth.
It's the big pink one uh in kind of the

00:07:01,990 --> 00:07:02,000
It's the big pink one uh in kind of the
 

00:07:02,000 --> 00:07:03,830
It's the big pink one uh in kind of the
middle of the the floor. Uh very

00:07:03,830 --> 00:07:03,840
middle of the the floor. Uh very
 

00:07:03,840 --> 00:07:06,230
middle of the the floor. Uh very
Valentine's Day themed. So, we've made

00:07:06,230 --> 00:07:06,240
Valentine's Day themed. So, we've made
 

00:07:06,240 --> 00:07:08,230
Valentine's Day themed. So, we've made
modular hardware and we've kind of

00:07:08,230 --> 00:07:08,240
modular hardware and we've kind of
 

00:07:08,240 --> 00:07:10,710
modular hardware and we've kind of
wrapped each instrument in hardware to

00:07:10,710 --> 00:07:10,720
wrapped each instrument in hardware to
 

00:07:10,720 --> 00:07:12,629
wrapped each instrument in hardware to
make it more reliable and more robust so

00:07:12,629 --> 00:07:12,639
make it more reliable and more robust so
 

00:07:12,639 --> 00:07:14,230
make it more reliable and more robust so
that it really behaves as well as

00:07:14,230 --> 00:07:14,240
that it really behaves as well as
 

00:07:14,240 --> 00:07:15,749
that it really behaves as well as
possible. and we put all the accessories

00:07:15,749 --> 00:07:15,759
possible. and we put all the accessories
 

00:07:15,759 --> 00:07:17,990
possible. and we put all the accessories
needed into this modular sort of Lego

00:07:17,990 --> 00:07:18,000
needed into this modular sort of Lego
 

00:07:18,000 --> 00:07:20,790
needed into this modular sort of Lego
building block. Um,

00:07:20,790 --> 00:07:20,800
building block. Um,
 

00:07:20,800 --> 00:07:23,749
building block. Um,
you can put multiple uh racks, we call

00:07:23,749 --> 00:07:23,759
you can put multiple uh racks, we call
 

00:07:23,759 --> 00:07:25,510
you can put multiple uh racks, we call
them reconfigurable automation carts

00:07:25,510 --> 00:07:25,520
them reconfigurable automation carts
 

00:07:25,520 --> 00:07:27,589
them reconfigurable automation carts
together and they plug together kind of

00:07:27,589 --> 00:07:27,599
together and they plug together kind of
 

00:07:27,599 --> 00:07:29,670
together and they plug together kind of
like Lego and they just work. And so

00:07:29,670 --> 00:07:29,680
like Lego and they just work. And so
 

00:07:29,680 --> 00:07:31,110
like Lego and they just work. And so
again, it's like this focus on

00:07:31,110 --> 00:07:31,120
again, it's like this focus on
 

00:07:31,120 --> 00:07:33,670
again, it's like this focus on
reliability and robustness.

00:07:33,670 --> 00:07:33,680
reliability and robustness.
 

00:07:33,680 --> 00:07:35,670
reliability and robustness.
You can start very small as in the top

00:07:35,670 --> 00:07:35,680
You can start very small as in the top
 

00:07:35,680 --> 00:07:37,749
You can start very small as in the top
lefthand corner with just a few racks.

00:07:37,749 --> 00:07:37,759
lefthand corner with just a few racks.
 

00:07:37,759 --> 00:07:39,430
lefthand corner with just a few racks.
So it's a small investment, begin

00:07:39,430 --> 00:07:39,440
So it's a small investment, begin
 

00:07:39,440 --> 00:07:41,589
So it's a small investment, begin
automating at that scale and then expand

00:07:41,589 --> 00:07:41,599
automating at that scale and then expand
 

00:07:41,599 --> 00:07:43,990
automating at that scale and then expand
over time in a kind of nice linear way

00:07:43,990 --> 00:07:44,000
over time in a kind of nice linear way
 

00:07:44,000 --> 00:07:47,350
over time in a kind of nice linear way
to add capacity or capability. Uh and

00:07:47,350 --> 00:07:47,360
to add capacity or capability. Uh and
 

00:07:47,360 --> 00:07:49,670
to add capacity or capability. Uh and
you can get to a very large system. I

00:07:49,670 --> 00:07:49,680
you can get to a very large system. I
 

00:07:49,680 --> 00:07:51,029
you can get to a very large system. I
don't want to steal too much of Todd's

00:07:51,029 --> 00:07:51,039
don't want to steal too much of Todd's
 

00:07:51,039 --> 00:07:52,870
don't want to steal too much of Todd's
thunder, but you know that system on the

00:07:52,870 --> 00:07:52,880
thunder, but you know that system on the
 

00:07:52,880 --> 00:07:55,990
thunder, but you know that system on the
the right is a 97 rack system purchased

00:07:55,990 --> 00:07:56,000
the right is a 97 rack system purchased
 

00:07:56,000 --> 00:07:59,029
the right is a 97 rack system purchased
by uh the Department of Energy by PNL.

00:07:59,029 --> 00:07:59,039
by uh the Department of Energy by PNL.
 

00:07:59,039 --> 00:08:01,510
by uh the Department of Energy by PNL.
Um and so this this technology really

00:08:01,510 --> 00:08:01,520
Um and so this this technology really
 

00:08:01,520 --> 00:08:04,550
Um and so this this technology really
scales from small to large.

00:08:04,550 --> 00:08:04,560
scales from small to large.
 

00:08:04,560 --> 00:08:05,990
scales from small to large.
And then the other part that we've

00:08:05,990 --> 00:08:06,000
And then the other part that we've
 

00:08:06,000 --> 00:08:07,589
And then the other part that we've
really invested a lot in is our

00:08:07,589 --> 00:08:07,599
really invested a lot in is our
 

00:08:07,599 --> 00:08:10,150
really invested a lot in is our
software. And there's a ton to be said

00:08:10,150 --> 00:08:10,160
software. And there's a ton to be said
 

00:08:10,160 --> 00:08:11,749
software. And there's a ton to be said
about this. Um, you know, we've tried to

00:08:11,749 --> 00:08:11,759
about this. Um, you know, we've tried to
 

00:08:11,759 --> 00:08:13,830
about this. Um, you know, we've tried to
make it super easy to use. Uh, by which

00:08:13,830 --> 00:08:13,840
make it super easy to use. Uh, by which
 

00:08:13,840 --> 00:08:15,990
make it super easy to use. Uh, by which
I mean writing protocols, launching

00:08:15,990 --> 00:08:16,000
I mean writing protocols, launching
 

00:08:16,000 --> 00:08:17,830
I mean writing protocols, launching
protocols, monitoring them, getting the

00:08:17,830 --> 00:08:17,840
protocols, monitoring them, getting the
 

00:08:17,840 --> 00:08:19,510
protocols, monitoring them, getting the
data off. But the one thing I wanted to

00:08:19,510 --> 00:08:19,520
data off. But the one thing I wanted to
 

00:08:19,520 --> 00:08:21,270
data off. But the one thing I wanted to
emphasize in this talk in this small

00:08:21,270 --> 00:08:21,280
emphasize in this talk in this small
 

00:08:21,280 --> 00:08:24,710
emphasize in this talk in this small
intro is just the interle of protocols.

00:08:24,710 --> 00:08:24,720
intro is just the interle of protocols.
 

00:08:24,720 --> 00:08:26,230
intro is just the interle of protocols.
And so we've uh, you know, we've got

00:08:26,230 --> 00:08:26,240
And so we've uh, you know, we've got
 

00:08:26,240 --> 00:08:28,070
And so we've uh, you know, we've got
distributed plate transport and then we

00:08:28,070 --> 00:08:28,080
distributed plate transport and then we
 

00:08:28,080 --> 00:08:29,749
distributed plate transport and then we
put a lot of time into our constraint

00:08:29,749 --> 00:08:29,759
put a lot of time into our constraint
 

00:08:29,759 --> 00:08:31,670
put a lot of time into our constraint
solver that is actually scheduling the

00:08:31,670 --> 00:08:31,680
solver that is actually scheduling the
 

00:08:31,680 --> 00:08:34,149
solver that is actually scheduling the
work. And so the cool thing here is the

00:08:34,149 --> 00:08:34,159
work. And so the cool thing here is the
 

00:08:34,159 --> 00:08:36,469
work. And so the cool thing here is the
different colors of the bars up there.

00:08:36,469 --> 00:08:36,479
different colors of the bars up there.
 

00:08:36,479 --> 00:08:37,750
different colors of the bars up there.
It's basically a little gant. The

00:08:37,750 --> 00:08:37,760
It's basically a little gant. The
 

00:08:37,760 --> 00:08:39,670
It's basically a little gant. The
different colors are different protocols

00:08:39,670 --> 00:08:39,680
different colors are different protocols
 

00:08:39,680 --> 00:08:41,829
different colors are different protocols
running on the same system at the same

00:08:41,829 --> 00:08:41,839
running on the same system at the same
 

00:08:41,839 --> 00:08:44,470
running on the same system at the same
time. So run by different people,

00:08:44,470 --> 00:08:44,480
time. So run by different people,
 

00:08:44,480 --> 00:08:46,389
time. So run by different people,
totally different workflows requiring

00:08:46,389 --> 00:08:46,399
totally different workflows requiring
 

00:08:46,399 --> 00:08:47,910
totally different workflows requiring
different instrumentation. And it

00:08:47,910 --> 00:08:47,920
different instrumentation. And it
 

00:08:47,920 --> 00:08:49,670
different instrumentation. And it
figures out how to schedule it and to

00:08:49,670 --> 00:08:49,680
figures out how to schedule it and to
 

00:08:49,680 --> 00:08:51,430
figures out how to schedule it and to
get the maximum utilization out of your

00:08:51,430 --> 00:08:51,440
get the maximum utilization out of your
 

00:08:51,440 --> 00:08:53,030
get the maximum utilization out of your
instruments.

00:08:53,030 --> 00:08:53,040
instruments.
 

00:08:53,040 --> 00:08:55,110
instruments.
So here is just a clear example where

00:08:55,110 --> 00:08:55,120
So here is just a clear example where
 

00:08:55,120 --> 00:08:56,790
So here is just a clear example where
that gant you can see there's a whole

00:08:56,790 --> 00:08:56,800
that gant you can see there's a whole
 

00:08:56,800 --> 00:08:58,870
that gant you can see there's a whole
bunch of colored blocks all in the same

00:08:58,870 --> 00:08:58,880
bunch of colored blocks all in the same
 

00:08:58,880 --> 00:09:00,550
bunch of colored blocks all in the same
vertical line which means multiple

00:09:00,550 --> 00:09:00,560
vertical line which means multiple
 

00:09:00,560 --> 00:09:02,630
vertical line which means multiple
things are moving simultaneously on the

00:09:02,630 --> 00:09:02,640
things are moving simultaneously on the
 

00:09:02,640 --> 00:09:04,630
things are moving simultaneously on the
system. So you've got multiple robots

00:09:04,630 --> 00:09:04,640
system. So you've got multiple robots
 

00:09:04,640 --> 00:09:07,430
system. So you've got multiple robots
transferring plates at the same time.

00:09:07,430 --> 00:09:07,440
transferring plates at the same time.
 

00:09:07,440 --> 00:09:10,870
transferring plates at the same time.
You can also uh have one rack in error

00:09:10,870 --> 00:09:10,880
You can also uh have one rack in error
 

00:09:10,880 --> 00:09:12,389
You can also uh have one rack in error
and the rest of the system keeps

00:09:12,389 --> 00:09:12,399
and the rest of the system keeps
 

00:09:12,399 --> 00:09:13,670
and the rest of the system keeps
running. So you get very high

00:09:13,670 --> 00:09:13,680
running. So you get very high
 

00:09:13,680 --> 00:09:15,590
running. So you get very high
availability. If there's a second

00:09:15,590 --> 00:09:15,600
availability. If there's a second
 

00:09:15,600 --> 00:09:16,949
availability. If there's a second
identical rack, it'll fail over

00:09:16,949 --> 00:09:16,959
identical rack, it'll fail over
 

00:09:16,959 --> 00:09:18,470
identical rack, it'll fail over
automatically.

00:09:18,470 --> 00:09:18,480
automatically.
 

00:09:18,480 --> 00:09:21,190
automatically.
And then this is the just the dark mode

00:09:21,190 --> 00:09:21,200
And then this is the just the dark mode
 

00:09:21,200 --> 00:09:23,910
And then this is the just the dark mode
version of that same uh sort of gant.

00:09:23,910 --> 00:09:23,920
version of that same uh sort of gant.
 

00:09:23,920 --> 00:09:26,310
version of that same uh sort of gant.
I'm sorry it's so dark here, but uh I

00:09:26,310 --> 00:09:26,320
I'm sorry it's so dark here, but uh I
 

00:09:26,320 --> 00:09:28,550
I'm sorry it's so dark here, but uh I
love this because this is a screenshot

00:09:28,550 --> 00:09:28,560
love this because this is a screenshot
 

00:09:28,560 --> 00:09:31,910
love this because this is a screenshot
of theuler in our system in Boston from

00:09:31,910 --> 00:09:31,920
of theuler in our system in Boston from
 

00:09:31,920 --> 00:09:34,150
of theuler in our system in Boston from
just a few days ago. And this is it it

00:09:34,150 --> 00:09:34,160
just a few days ago. And this is it it
 

00:09:34,160 --> 00:09:36,870
just a few days ago. And this is it it
running 35 simultaneous protocols from

00:09:36,870 --> 00:09:36,880
running 35 simultaneous protocols from
 

00:09:36,880 --> 00:09:38,949
running 35 simultaneous protocols from
different users. You know, there are

00:09:38,949 --> 00:09:38,959
different users. You know, there are
 

00:09:38,959 --> 00:09:40,470
different users. You know, there are
vendors out there that can run one or

00:09:40,470 --> 00:09:40,480
vendors out there that can run one or
 

00:09:40,480 --> 00:09:42,470
vendors out there that can run one or
two things at a time, but nobody is

00:09:42,470 --> 00:09:42,480
two things at a time, but nobody is
 

00:09:42,480 --> 00:09:44,630
two things at a time, but nobody is
approaching this level of concurrency.

00:09:44,630 --> 00:09:44,640
approaching this level of concurrency.
 

00:09:44,640 --> 00:09:46,230
approaching this level of concurrency.
And I really think that allows you to

00:09:46,230 --> 00:09:46,240
And I really think that allows you to
 

00:09:46,240 --> 00:09:48,870
And I really think that allows you to
make a core lab, an autonomous lab, uh

00:09:48,870 --> 00:09:48,880
make a core lab, an autonomous lab, uh
 

00:09:48,880 --> 00:09:50,710
make a core lab, an autonomous lab, uh
that is much more effective and

00:09:50,710 --> 00:09:50,720
that is much more effective and
 

00:09:50,720 --> 00:09:52,550
that is much more effective and
efficient than you can make with any

00:09:52,550 --> 00:09:52,560
efficient than you can make with any
 

00:09:52,560 --> 00:09:54,550
efficient than you can make with any
other technology.

00:09:54,550 --> 00:09:54,560
other technology.
 

00:09:54,560 --> 00:09:58,070
other technology.
Um this is our lab just down the street.

00:09:58,070 --> 00:09:58,080
Um this is our lab just down the street.
 

00:09:58,080 --> 00:10:00,630
Um this is our lab just down the street.
Uh if you want a tour of this, you can

00:10:00,630 --> 00:10:00,640
Uh if you want a tour of this, you can
 

00:10:00,640 --> 00:10:03,190
Uh if you want a tour of this, you can
come to our booth uh and we'll give you

00:10:03,190 --> 00:10:03,200
come to our booth uh and we'll give you
 

00:10:03,200 --> 00:10:05,910
come to our booth uh and we'll give you
a Uber coupon to just drive over there

00:10:05,910 --> 00:10:05,920
a Uber coupon to just drive over there
 

00:10:05,920 --> 00:10:08,150
a Uber coupon to just drive over there
and you can get a you can go there and

00:10:08,150 --> 00:10:08,160
and you can get a you can go there and
 

00:10:08,160 --> 00:10:09,910
and you can get a you can go there and
come back in about half an hour and get

00:10:09,910 --> 00:10:09,920
come back in about half an hour and get
 

00:10:09,920 --> 00:10:12,230
come back in about half an hour and get
a quick tour of our facility and then be

00:10:12,230 --> 00:10:12,240
a quick tour of our facility and then be
 

00:10:12,240 --> 00:10:14,389
a quick tour of our facility and then be
back at SLA in no time. Uh so again,

00:10:14,389 --> 00:10:14,399
back at SLA in no time. Uh so again,
 

00:10:14,399 --> 00:10:16,550
back at SLA in no time. Uh so again,
we'll pay for the Uber. Uh also on

00:10:16,550 --> 00:10:16,560
we'll pay for the Uber. Uh also on
 

00:10:16,560 --> 00:10:17,990
we'll pay for the Uber. Uh also on
Thursday, uh we're going to be doing

00:10:17,990 --> 00:10:18,000
Thursday, uh we're going to be doing
 

00:10:18,000 --> 00:10:19,590
Thursday, uh we're going to be doing
tours all day, so you can just show up

00:10:19,590 --> 00:10:19,600
tours all day, so you can just show up
 

00:10:19,600 --> 00:10:22,790
tours all day, so you can just show up
at any time. 25 Dry Dock Avenue is the

00:10:22,790 --> 00:10:22,800
at any time. 25 Dry Dock Avenue is the
 

00:10:22,800 --> 00:10:25,030
at any time. 25 Dry Dock Avenue is the
address. So please come on by. It's one

00:10:25,030 --> 00:10:25,040
address. So please come on by. It's one
 

00:10:25,040 --> 00:10:26,790
address. So please come on by. It's one
thing to go see our booth. It's another

00:10:26,790 --> 00:10:26,800
thing to go see our booth. It's another
 

00:10:26,800 --> 00:10:30,150
thing to go see our booth. It's another
thing to see the actual racks in the lab

00:10:30,150 --> 00:10:30,160
thing to see the actual racks in the lab
 

00:10:30,160 --> 00:10:32,949
thing to see the actual racks in the lab
doing real science.

00:10:32,949 --> 00:10:32,959
doing real science.
 

00:10:32,959 --> 00:10:37,030
doing real science.
So this is the fastest part, I promise.

00:10:37,030 --> 00:10:37,040
So this is the fastest part, I promise.
 

00:10:37,040 --> 00:10:39,190
So this is the fastest part, I promise.
We are commercializing this in two ways.

00:10:39,190 --> 00:10:39,200
We are commercializing this in two ways.
 

00:10:39,200 --> 00:10:40,630
We are commercializing this in two ways.
One is we are happy to build out

00:10:40,630 --> 00:10:40,640
One is we are happy to build out
 

00:10:40,640 --> 00:10:43,110
One is we are happy to build out
autonomous labs, you know, in your lab.

00:10:43,110 --> 00:10:43,120
autonomous labs, you know, in your lab.
 

00:10:43,120 --> 00:10:45,030
autonomous labs, you know, in your lab.
Um, and that looks like a traditional

00:10:45,030 --> 00:10:45,040
Um, and that looks like a traditional
 

00:10:45,040 --> 00:10:47,350
Um, and that looks like a traditional
automation sale or you can take

00:10:47,350 --> 00:10:47,360
automation sale or you can take
 

00:10:47,360 --> 00:10:49,910
automation sale or you can take
advantage of the same technology via our

00:10:49,910 --> 00:10:49,920
advantage of the same technology via our
 

00:10:49,920 --> 00:10:53,110
advantage of the same technology via our
CRO services. So, we will run protocols

00:10:53,110 --> 00:10:53,120
CRO services. So, we will run protocols
 

00:10:53,120 --> 00:10:56,550
CRO services. So, we will run protocols
for you if you want. Uh, there is a QR

00:10:56,550 --> 00:10:56,560
for you if you want. Uh, there is a QR
 

00:10:56,560 --> 00:10:57,990
for you if you want. Uh, there is a QR
code for my LinkedIn if you want to make

00:10:57,990 --> 00:10:58,000
code for my LinkedIn if you want to make
 

00:10:58,000 --> 00:11:00,069
code for my LinkedIn if you want to make
a connection. Uh, happy to connect with

00:11:00,069 --> 00:11:00,079
a connection. Uh, happy to connect with
 

00:11:00,079 --> 00:11:01,670
a connection. Uh, happy to connect with
more folks who are interested in this.

00:11:01,670 --> 00:11:01,680
more folks who are interested in this.
 

00:11:01,680 --> 00:11:05,110
more folks who are interested in this.
Um, again, our booth 823. Stop on by.

00:11:05,110 --> 00:11:05,120
Um, again, our booth 823. Stop on by.
 

00:11:05,120 --> 00:11:08,790
Um, again, our booth 823. Stop on by.
Um, that is uh my last slide and I will

00:11:08,790 --> 00:11:08,800
Um, that is uh my last slide and I will
 

00:11:08,800 --> 00:11:10,550
Um, that is uh my last slide and I will
leave it up for just another second as I

00:11:10,550 --> 00:11:10,560
leave it up for just another second as I
 

00:11:10,560 --> 00:11:13,269
leave it up for just another second as I
see people scanning the QR code and I

00:11:13,269 --> 00:11:13,279
see people scanning the QR code and I
 

00:11:13,279 --> 00:11:15,190
see people scanning the QR code and I
will now move on and then you'll have to

00:11:15,190 --> 00:11:15,200
will now move on and then you'll have to
 

00:11:15,200 --> 00:11:18,150
will now move on and then you'll have to
find me afterwards. Sorry. Um, so uh it

00:11:18,150 --> 00:11:18,160
find me afterwards. Sorry. Um, so uh it
 

00:11:18,160 --> 00:11:20,069
find me afterwards. Sorry. Um, so uh it
is my pleasure to pass things over to

00:11:20,069 --> 00:11:20,079
is my pleasure to pass things over to
 

00:11:20,079 --> 00:11:21,990
is my pleasure to pass things over to
Joy from OpenAI who's going to tell you

00:11:21,990 --> 00:11:22,000
Joy from OpenAI who's going to tell you
 

00:11:22,000 --> 00:11:23,590
Joy from OpenAI who's going to tell you
a little bit about uh some of the closed

00:11:23,590 --> 00:11:23,600
a little bit about uh some of the closed
 

00:11:23,600 --> 00:11:25,350
a little bit about uh some of the closed
loop science we've been doing together

00:11:25,350 --> 00:11:25,360
loop science we've been doing together
 

00:11:25,360 --> 00:11:27,829
loop science we've been doing together
on top of the platform I just described.

00:11:27,829 --> 00:11:27,839
on top of the platform I just described.
 

00:11:27,839 --> 00:11:32,470
on top of the platform I just described.
Joy,

00:11:32,470 --> 00:11:32,480

 

00:11:32,480 --> 00:11:33,750

>> you can use that or you can just use

00:11:33,750 --> 00:11:33,760
>> you can use that or you can just use
 

00:11:33,760 --> 00:11:35,910
>> you can use that or you can just use
arrows.

00:11:35,910 --> 00:11:35,920
arrows.
 

00:11:35,920 --> 00:11:37,750
arrows.
Okay.

00:11:37,750 --> 00:11:37,760
Okay.
 

00:11:37,760 --> 00:11:41,030
Okay.
>> All right. Um, so I'm really excited to

00:11:41,030 --> 00:11:41,040
>> All right. Um, so I'm really excited to
 

00:11:41,040 --> 00:11:43,430
>> All right. Um, so I'm really excited to
tell you more about um, this project

00:11:43,430 --> 00:11:43,440
tell you more about um, this project
 

00:11:43,440 --> 00:11:45,590
tell you more about um, this project
that we did with GKO over the last

00:11:45,590 --> 00:11:45,600
that we did with GKO over the last
 

00:11:45,600 --> 00:11:48,550
that we did with GKO over the last
sixish months where we basically gave

00:11:48,550 --> 00:11:48,560
sixish months where we basically gave
 

00:11:48,560 --> 00:11:52,389
sixish months where we basically gave
GPD5 um, control over GKO's rack system

00:11:52,389 --> 00:11:52,399
GPD5 um, control over GKO's rack system
 

00:11:52,399 --> 00:11:55,269
GPD5 um, control over GKO's rack system
and we were able to make more protein

00:11:55,269 --> 00:11:55,279
and we were able to make more protein
 

00:11:55,279 --> 00:11:58,470
and we were able to make more protein
um, in this case SFGF at a lower cost

00:11:58,470 --> 00:11:58,480
um, in this case SFGF at a lower cost
 

00:11:58,480 --> 00:12:00,630
um, in this case SFGF at a lower cost
than the previous human state-of-the-art

00:12:00,630 --> 00:12:00,640
than the previous human state-of-the-art
 

00:12:00,640 --> 00:12:02,550
than the previous human state-of-the-art
result.

00:12:02,550 --> 00:12:02,560
result.
 

00:12:02,560 --> 00:12:04,470
result.
Um, okay. So just a little bit of

00:12:04,470 --> 00:12:04,480
Um, okay. So just a little bit of
 

00:12:04,480 --> 00:12:06,710
Um, okay. So just a little bit of
background. I think this was not quite

00:12:06,710 --> 00:12:06,720
background. I think this was not quite
 

00:12:06,720 --> 00:12:09,430
background. I think this was not quite
as obvious six months ago as it is now.

00:12:09,430 --> 00:12:09,440
as obvious six months ago as it is now.
 

00:12:09,440 --> 00:12:11,990
as obvious six months ago as it is now.
But I generally think of AI in science

00:12:11,990 --> 00:12:12,000
But I generally think of AI in science
 

00:12:12,000 --> 00:12:14,949
But I generally think of AI in science
kind of as this um different stages of

00:12:14,949 --> 00:12:14,959
kind of as this um different stages of
 

00:12:14,959 --> 00:12:17,350
kind of as this um different stages of
capability. So initially we think of AI

00:12:17,350 --> 00:12:17,360
capability. So initially we think of AI
 

00:12:17,360 --> 00:12:20,150
capability. So initially we think of AI
as tools. Um and this is you know

00:12:20,150 --> 00:12:20,160
as tools. Um and this is you know
 

00:12:20,160 --> 00:12:22,389
as tools. Um and this is you know
commonly things like alpha fold bolts

00:12:22,389 --> 00:12:22,399
commonly things like alpha fold bolts
 

00:12:22,399 --> 00:12:24,710
commonly things like alpha fold bolts
like general um biology foundation

00:12:24,710 --> 00:12:24,720
like general um biology foundation
 

00:12:24,720 --> 00:12:27,750
like general um biology foundation
models and open also had our own paper

00:12:27,750 --> 00:12:27,760
models and open also had our own paper
 

00:12:27,760 --> 00:12:30,470
models and open also had our own paper
out a few months ago about GBT4B which

00:12:30,470 --> 00:12:30,480
out a few months ago about GBT4B which
 

00:12:30,480 --> 00:12:32,949
out a few months ago about GBT4B which
is actually just a very tiny model that

00:12:32,949 --> 00:12:32,959
is actually just a very tiny model that
 

00:12:32,959 --> 00:12:36,150
is actually just a very tiny model that
we fine-tuned to generate uh Yamanaka

00:12:36,150 --> 00:12:36,160
we fine-tuned to generate uh Yamanaka
 

00:12:36,160 --> 00:12:39,430
we fine-tuned to generate uh Yamanaka
factors um variance and this actually

00:12:39,430 --> 00:12:39,440
factors um variance and this actually
 

00:12:39,440 --> 00:12:41,030
factors um variance and this actually
showed that there was like a 50x

00:12:41,030 --> 00:12:41,040
showed that there was like a 50x
 

00:12:41,040 --> 00:12:42,710
showed that there was like a 50x
reprogramming efficiency. So this is

00:12:42,710 --> 00:12:42,720
reprogramming efficiency. So this is
 

00:12:42,720 --> 00:12:44,550
reprogramming efficiency. So this is
like a interesting project kind of in

00:12:44,550 --> 00:12:44,560
like a interesting project kind of in
 

00:12:44,560 --> 00:12:47,350
like a interesting project kind of in
the AI as tools regime. Um and then we

00:12:47,350 --> 00:12:47,360
the AI as tools regime. Um and then we
 

00:12:47,360 --> 00:12:49,829
the AI as tools regime. Um and then we
see kind of AI evolving as tool users.

00:12:49,829 --> 00:12:49,839
see kind of AI evolving as tool users.
 

00:12:49,839 --> 00:12:51,990
see kind of AI evolving as tool users.
So we start giving our models access to

00:12:51,990 --> 00:12:52,000
So we start giving our models access to
 

00:12:52,000 --> 00:12:54,230
So we start giving our models access to
internet, access to computers and we

00:12:54,230 --> 00:12:54,240
internet, access to computers and we
 

00:12:54,240 --> 00:12:56,629
internet, access to computers and we
generally see a huge capability jump in

00:12:56,629 --> 00:12:56,639
generally see a huge capability jump in
 

00:12:56,639 --> 00:12:59,030
generally see a huge capability jump in
this regime. Um and then the thing that

00:12:59,030 --> 00:12:59,040
this regime. Um and then the thing that
 

00:12:59,040 --> 00:13:01,509
this regime. Um and then the thing that
I really want to talk about um here is

00:13:01,509 --> 00:13:01,519
I really want to talk about um here is
 

00:13:01,519 --> 00:13:03,829
I really want to talk about um here is
kind of AI as co-scientists working

00:13:03,829 --> 00:13:03,839
kind of AI as co-scientists working
 

00:13:03,839 --> 00:13:07,030
kind of AI as co-scientists working
alongside humans. Um and then also in

00:13:07,030 --> 00:13:07,040
alongside humans. Um and then also in
 

00:13:07,040 --> 00:13:08,550
alongside humans. Um and then also in
some cases like in this project

00:13:08,550 --> 00:13:08,560
some cases like in this project
 

00:13:08,560 --> 00:13:11,350
some cases like in this project
designing experiments independently and

00:13:11,350 --> 00:13:11,360
designing experiments independently and
 

00:13:11,360 --> 00:13:13,910
designing experiments independently and
at the start of this project GPD5 had

00:13:13,910 --> 00:13:13,920
at the start of this project GPD5 had
 

00:13:13,920 --> 00:13:15,590
at the start of this project GPD5 had
actually just finished training at

00:13:15,590 --> 00:13:15,600
actually just finished training at
 

00:13:15,600 --> 00:13:17,910
actually just finished training at
OpenAI and we knew that it was very

00:13:17,910 --> 00:13:17,920
OpenAI and we knew that it was very
 

00:13:17,920 --> 00:13:19,350
OpenAI and we knew that it was very
knowledgeable. We knew that it was very

00:13:19,350 --> 00:13:19,360
knowledgeable. We knew that it was very
 

00:13:19,360 --> 00:13:22,069
knowledgeable. We knew that it was very
good at programming and math and physics

00:13:22,069 --> 00:13:22,079
good at programming and math and physics
 

00:13:22,079 --> 00:13:24,230
good at programming and math and physics
um but we haven't really tested it in

00:13:24,230 --> 00:13:24,240
um but we haven't really tested it in
 

00:13:24,240 --> 00:13:26,470
um but we haven't really tested it in
biology and part of this is because you

00:13:26,470 --> 00:13:26,480
biology and part of this is because you
 

00:13:26,480 --> 00:13:28,389
biology and part of this is because you
know biology requires wet lab to really

00:13:28,389 --> 00:13:28,399
know biology requires wet lab to really
 

00:13:28,399 --> 00:13:30,550
know biology requires wet lab to really
validate. So I think computational

00:13:30,550 --> 00:13:30,560
validate. So I think computational
 

00:13:30,560 --> 00:13:32,790
validate. So I think computational
validation of comput computational

00:13:32,790 --> 00:13:32,800
validation of comput computational
 

00:13:32,800 --> 00:13:35,030
validation of comput computational
results is not quite as interesting. And

00:13:35,030 --> 00:13:35,040
results is not quite as interesting. And
 

00:13:35,040 --> 00:13:37,110
results is not quite as interesting. And
so with GKO we had this opportunity to

00:13:37,110 --> 00:13:37,120
so with GKO we had this opportunity to
 

00:13:37,120 --> 00:13:39,350
so with GKO we had this opportunity to
really have the model design real

00:13:39,350 --> 00:13:39,360
really have the model design real
 

00:13:39,360 --> 00:13:41,350
really have the model design real
biology wildlife experiments and we can

00:13:41,350 --> 00:13:41,360
biology wildlife experiments and we can
 

00:13:41,360 --> 00:13:44,069
biology wildlife experiments and we can
actually run them in the rack system and

00:13:44,069 --> 00:13:44,079
actually run them in the rack system and
 

00:13:44,079 --> 00:13:45,990
actually run them in the rack system and
really kind of evaluate GPD5's

00:13:45,990 --> 00:13:46,000
really kind of evaluate GPD5's
 

00:13:46,000 --> 00:13:50,710
really kind of evaluate GPD5's
capability at doing real biology. Um so

00:13:50,710 --> 00:13:50,720
capability at doing real biology. Um so
 

00:13:50,720 --> 00:13:53,110
capability at doing real biology. Um so
the experiment that we settled on is

00:13:53,110 --> 00:13:53,120
the experiment that we settled on is
 

00:13:53,120 --> 00:13:55,030
the experiment that we settled on is
self-free protein synthesis. So I have a

00:13:55,030 --> 00:13:55,040
self-free protein synthesis. So I have a
 

00:13:55,040 --> 00:13:56,550
self-free protein synthesis. So I have a
few studies kind of talking about why

00:13:56,550 --> 00:13:56,560
few studies kind of talking about why
 

00:13:56,560 --> 00:13:58,310
few studies kind of talking about why
this is important. Apologies to those in

00:13:58,310 --> 00:13:58,320
this is important. Apologies to those in
 

00:13:58,320 --> 00:13:59,670
this is important. Apologies to those in
the audience who know way more about

00:13:59,670 --> 00:13:59,680
the audience who know way more about
 

00:13:59,680 --> 00:14:02,069
the audience who know way more about
this than I do. Um, but generally, you

00:14:02,069 --> 00:14:02,079
this than I do. Um, but generally, you
 

00:14:02,079 --> 00:14:03,590
this than I do. Um, but generally, you
know, protein production is pretty

00:14:03,590 --> 00:14:03,600
know, protein production is pretty
 

00:14:03,600 --> 00:14:05,189
know, protein production is pretty
foundational to pretty much everything

00:14:05,189 --> 00:14:05,199
foundational to pretty much everything
 

00:14:05,199 --> 00:14:07,269
foundational to pretty much everything
including medicine and research. But

00:14:07,269 --> 00:14:07,279
including medicine and research. But
 

00:14:07,279 --> 00:14:09,030
including medicine and research. But
generally, traditional methods are very

00:14:09,030 --> 00:14:09,040
generally, traditional methods are very
 

00:14:09,040 --> 00:14:11,269
generally, traditional methods are very
slow and complicated. So, you have to

00:14:11,269 --> 00:14:11,279
slow and complicated. So, you have to
 

00:14:11,279 --> 00:14:13,269
slow and complicated. So, you have to
introduce DNA encoding a new protein

00:14:13,269 --> 00:14:13,279
introduce DNA encoding a new protein
 

00:14:13,279 --> 00:14:15,430
introduce DNA encoding a new protein
into cells. Often you have to create a

00:14:15,430 --> 00:14:15,440
into cells. Often you have to create a
 

00:14:15,440 --> 00:14:17,269
into cells. Often you have to create a
stable cell line, scale production of

00:14:17,269 --> 00:14:17,279
stable cell line, scale production of
 

00:14:17,279 --> 00:14:19,350
stable cell line, scale production of
bioreactors, and this often takes on the

00:14:19,350 --> 00:14:19,360
bioreactors, and this often takes on the
 

00:14:19,360 --> 00:14:21,430
bioreactors, and this often takes on the
order of weeks to months. So, the

00:14:21,430 --> 00:14:21,440
order of weeks to months. So, the
 

00:14:21,440 --> 00:14:23,110
order of weeks to months. So, the
limitations here is that it's often slow

00:14:23,110 --> 00:14:23,120
limitations here is that it's often slow
 

00:14:23,120 --> 00:14:25,350
limitations here is that it's often slow
and inflexible. There's limited protein

00:14:25,350 --> 00:14:25,360
and inflexible. There's limited protein
 

00:14:25,360 --> 00:14:26,870
and inflexible. There's limited protein
types that you can make. So for example,

00:14:26,870 --> 00:14:26,880
types that you can make. So for example,
 

00:14:26,880 --> 00:14:29,110
types that you can make. So for example,
if you have a protein that's toxic to

00:14:29,110 --> 00:14:29,120
if you have a protein that's toxic to
 

00:14:29,120 --> 00:14:31,110
if you have a protein that's toxic to
bacteria, you can't really grow it in

00:14:31,110 --> 00:14:31,120
bacteria, you can't really grow it in
 

00:14:31,120 --> 00:14:33,509
bacteria, you can't really grow it in
bacteria. Um there's potential safety

00:14:33,509 --> 00:14:33,519
bacteria. Um there's potential safety
 

00:14:33,519 --> 00:14:35,829
bacteria. Um there's potential safety
risk involved is not quite portable

00:14:35,829 --> 00:14:35,839
risk involved is not quite portable
 

00:14:35,839 --> 00:14:37,910
risk involved is not quite portable
because you require zero facilities and

00:14:37,910 --> 00:14:37,920
because you require zero facilities and
 

00:14:37,920 --> 00:14:40,069
because you require zero facilities and
large bioreactors.

00:14:40,069 --> 00:14:40,079
large bioreactors.
 

00:14:40,079 --> 00:14:42,150
large bioreactors.
So cellfree protein synthesis kind of

00:14:42,150 --> 00:14:42,160
So cellfree protein synthesis kind of
 

00:14:42,160 --> 00:14:45,509
So cellfree protein synthesis kind of
turns this regime um inside out in some

00:14:45,509 --> 00:14:45,519
turns this regime um inside out in some
 

00:14:45,519 --> 00:14:47,350
turns this regime um inside out in some
ways where now instead of making

00:14:47,350 --> 00:14:47,360
ways where now instead of making
 

00:14:47,360 --> 00:14:49,910
ways where now instead of making
proteins in living cells, you actually

00:14:49,910 --> 00:14:49,920
proteins in living cells, you actually
 

00:14:49,920 --> 00:14:52,230
proteins in living cells, you actually
lice the cells. So all the protein

00:14:52,230 --> 00:14:52,240
lice the cells. So all the protein
 

00:14:52,240 --> 00:14:54,069
lice the cells. So all the protein
making machinery inside the cells is now

00:14:54,069 --> 00:14:54,079
making machinery inside the cells is now
 

00:14:54,079 --> 00:14:55,750
making machinery inside the cells is now
kind of free floating in this tube and

00:14:55,750 --> 00:14:55,760
kind of free floating in this tube and
 

00:14:55,760 --> 00:14:58,069
kind of free floating in this tube and
you can actually just spike in your DNA

00:14:58,069 --> 00:14:58,079
you can actually just spike in your DNA
 

00:14:58,079 --> 00:15:00,310
you can actually just spike in your DNA
construct encoding a protein of interest

00:15:00,310 --> 00:15:00,320
construct encoding a protein of interest
 

00:15:00,320 --> 00:15:03,189
construct encoding a protein of interest
and make it this way. Um and so this is

00:15:03,189 --> 00:15:03,199
and make it this way. Um and so this is
 

00:15:03,199 --> 00:15:05,269
and make it this way. Um and so this is
really good for kind of prototyping

00:15:05,269 --> 00:15:05,279
really good for kind of prototyping
 

00:15:05,279 --> 00:15:06,870
really good for kind of prototyping
large libraries of different protein

00:15:06,870 --> 00:15:06,880
large libraries of different protein
 

00:15:06,880 --> 00:15:08,550
large libraries of different protein
variants and just general high

00:15:08,550 --> 00:15:08,560
variants and just general high
 

00:15:08,560 --> 00:15:11,189
variants and just general high
throughput workflows. And so what this

00:15:11,189 --> 00:15:11,199
throughput workflows. And so what this
 

00:15:11,199 --> 00:15:13,670
throughput workflows. And so what this
can give us is you know faster dis uh

00:15:13,670 --> 00:15:13,680
can give us is you know faster dis uh
 

00:15:13,680 --> 00:15:16,310
can give us is you know faster dis uh
drug discovery personalized therapeutics

00:15:16,310 --> 00:15:16,320
drug discovery personalized therapeutics
 

00:15:16,320 --> 00:15:18,629
drug discovery personalized therapeutics
and we can get significant speedups

00:15:18,629 --> 00:15:18,639
and we can get significant speedups
 

00:15:18,639 --> 00:15:19,750
and we can get significant speedups
because now instead of having to make

00:15:19,750 --> 00:15:19,760
because now instead of having to make
 

00:15:19,760 --> 00:15:22,230
because now instead of having to make
these cell lines you can get proteins on

00:15:22,230 --> 00:15:22,240
these cell lines you can get proteins on
 

00:15:22,240 --> 00:15:24,790
these cell lines you can get proteins on
the order of hours or days. Um high

00:15:24,790 --> 00:15:24,800
the order of hours or days. Um high
 

00:15:24,800 --> 00:15:26,310
the order of hours or days. Um high
throughput you can make many things in

00:15:26,310 --> 00:15:26,320
throughput you can make many things in
 

00:15:26,320 --> 00:15:28,949
throughput you can make many things in
parallel. It's a lot more flexible and

00:15:28,949 --> 00:15:28,959
parallel. It's a lot more flexible and
 

00:15:28,959 --> 00:15:30,310
parallel. It's a lot more flexible and
safe because now you don't have this

00:15:30,310 --> 00:15:30,320
safe because now you don't have this
 

00:15:30,320 --> 00:15:32,949
safe because now you don't have this
kind of like um contamination concerns.

00:15:32,949 --> 00:15:32,959
kind of like um contamination concerns.
 

00:15:32,959 --> 00:15:34,790
kind of like um contamination concerns.
And the really nice thing here too is

00:15:34,790 --> 00:15:34,800
And the really nice thing here too is
 

00:15:34,800 --> 00:15:37,350
And the really nice thing here too is
that the setup itself is not super

00:15:37,350 --> 00:15:37,360
that the setup itself is not super
 

00:15:37,360 --> 00:15:39,269
that the setup itself is not super
complicated from a protocol perspective

00:15:39,269 --> 00:15:39,279
complicated from a protocol perspective
 

00:15:39,279 --> 00:15:41,590
complicated from a protocol perspective
because it's kind of at least from um

00:15:41,590 --> 00:15:41,600
because it's kind of at least from um
 

00:15:41,600 --> 00:15:43,110
because it's kind of at least from um
the models view, right? Like you add a

00:15:43,110 --> 00:15:43,120
the models view, right? Like you add a
 

00:15:43,120 --> 00:15:45,189
the models view, right? Like you add a
bunch of different volumes of reactants

00:15:45,189 --> 00:15:45,199
bunch of different volumes of reactants
 

00:15:45,199 --> 00:15:46,629
bunch of different volumes of reactants
into a tube and then that tube just

00:15:46,629 --> 00:15:46,639
into a tube and then that tube just
 

00:15:46,639 --> 00:15:49,590
into a tube and then that tube just
incubates. And so this is a good problem

00:15:49,590 --> 00:15:49,600
incubates. And so this is a good problem
 

00:15:49,600 --> 00:15:51,910
incubates. And so this is a good problem
for our models because it doesn't

00:15:51,910 --> 00:15:51,920
for our models because it doesn't
 

00:15:51,920 --> 00:15:53,910
for our models because it doesn't
require that much understanding of the

00:15:53,910 --> 00:15:53,920
require that much understanding of the
 

00:15:53,920 --> 00:15:56,230
require that much understanding of the
physicality of the experiment. It's

00:15:56,230 --> 00:15:56,240
physicality of the experiment. It's
 

00:15:56,240 --> 00:15:57,509
physicality of the experiment. It's
really just kind of thinking about

00:15:57,509 --> 00:15:57,519
really just kind of thinking about
 

00:15:57,519 --> 00:15:59,749
really just kind of thinking about
optimizing the volumes of reactants that

00:15:59,749 --> 00:15:59,759
optimizing the volumes of reactants that
 

00:15:59,759 --> 00:16:02,069
optimizing the volumes of reactants that
you're putting into the tube. So there's

00:16:02,069 --> 00:16:02,079
you're putting into the tube. So there's
 

00:16:02,079 --> 00:16:04,629
you're putting into the tube. So there's
um a nice kind of constraint on the

00:16:04,629 --> 00:16:04,639
um a nice kind of constraint on the
 

00:16:04,639 --> 00:16:06,470
um a nice kind of constraint on the
action space but there's still a lot of

00:16:06,470 --> 00:16:06,480
action space but there's still a lot of
 

00:16:06,480 --> 00:16:10,069
action space but there's still a lot of
NOS for the model to tune and explore.

00:16:10,069 --> 00:16:10,079
NOS for the model to tune and explore.
 

00:16:10,079 --> 00:16:12,230
NOS for the model to tune and explore.
Okay. So this is kind of the general

00:16:12,230 --> 00:16:12,240
Okay. So this is kind of the general
 

00:16:12,240 --> 00:16:15,189
Okay. So this is kind of the general
setup. Um on top we have GT5 that is

00:16:15,189 --> 00:16:15,199
setup. Um on top we have GT5 that is
 

00:16:15,199 --> 00:16:17,350
setup. Um on top we have GT5 that is
kind of doing the data analysis,

00:16:17,350 --> 00:16:17,360
kind of doing the data analysis,
 

00:16:17,360 --> 00:16:20,069
kind of doing the data analysis,
reasoning, generating hypothesis. It's

00:16:20,069 --> 00:16:20,079
reasoning, generating hypothesis. It's
 

00:16:20,079 --> 00:16:22,470
reasoning, generating hypothesis. It's
actually writing code to define the next

00:16:22,470 --> 00:16:22,480
actually writing code to define the next
 

00:16:22,480 --> 00:16:25,749
actually writing code to define the next
set of experiments. Um and I'll kind of

00:16:25,749 --> 00:16:25,759
set of experiments. Um and I'll kind of
 

00:16:25,759 --> 00:16:27,590
set of experiments. Um and I'll kind of
like walk through what the exper looks

00:16:27,590 --> 00:16:27,600
like walk through what the exper looks
 

00:16:27,600 --> 00:16:30,629
like walk through what the exper looks
like a few few slides out. Um but

00:16:30,629 --> 00:16:30,639
like a few few slides out. Um but
 

00:16:30,639 --> 00:16:32,470
like a few few slides out. Um but
essentially this is a pyantic object

00:16:32,470 --> 00:16:32,480
essentially this is a pyantic object
 

00:16:32,480 --> 00:16:35,110
essentially this is a pyantic object
that is like serialized into a JSON file

00:16:35,110 --> 00:16:35,120
that is like serialized into a JSON file
 

00:16:35,120 --> 00:16:37,670
that is like serialized into a JSON file
that we're sharing between the two labs

00:16:37,670 --> 00:16:37,680
that we're sharing between the two labs
 

00:16:37,680 --> 00:16:40,550
that we're sharing between the two labs
and then from the GKO side the physical

00:16:40,550 --> 00:16:40,560
and then from the GKO side the physical
 

00:16:40,560 --> 00:16:42,550
and then from the GKO side the physical
experiment is actually executed. So we

00:16:42,550 --> 00:16:42,560
experiment is actually executed. So we
 

00:16:42,560 --> 00:16:44,069
experiment is actually executed. So we
have the automated liquid handling, the

00:16:44,069 --> 00:16:44,079
have the automated liquid handling, the
 

00:16:44,079 --> 00:16:45,269
have the automated liquid handling, the
incubation, the fluoresence

00:16:45,269 --> 00:16:45,279
incubation, the fluoresence
 

00:16:45,279 --> 00:16:47,509
incubation, the fluoresence
measurements, um the standardization,

00:16:47,509 --> 00:16:47,519
measurements, um the standardization,
 

00:16:47,519 --> 00:16:48,629
measurements, um the standardization,
all of these things are happening there.

00:16:48,629 --> 00:16:48,639
all of these things are happening there.
 

00:16:48,639 --> 00:16:50,710
all of these things are happening there.
And then the metrics and the data and

00:16:50,710 --> 00:16:50,720
And then the metrics and the data and
 

00:16:50,720 --> 00:16:52,870
And then the metrics and the data and
any kind of metadata about things that

00:16:52,870 --> 00:16:52,880
any kind of metadata about things that
 

00:16:52,880 --> 00:16:54,629
any kind of metadata about things that
may have happened during the experiment

00:16:54,629 --> 00:16:54,639
may have happened during the experiment
 

00:16:54,639 --> 00:16:56,629
may have happened during the experiment
like for example there there's equipment

00:16:56,629 --> 00:16:56,639
like for example there there's equipment
 

00:16:56,639 --> 00:16:59,110
like for example there there's equipment
that eroded out on specific wells. All

00:16:59,110 --> 00:16:59,120
that eroded out on specific wells. All
 

00:16:59,120 --> 00:17:01,990
that eroded out on specific wells. All
of that data is then sent back to GPD5

00:17:01,990 --> 00:17:02,000
of that data is then sent back to GPD5
 

00:17:02,000 --> 00:17:03,670
of that data is then sent back to GPD5
and then we just kind of run this loop

00:17:03,670 --> 00:17:03,680
and then we just kind of run this loop
 

00:17:03,680 --> 00:17:06,309
and then we just kind of run this loop
over and over again. So over the last

00:17:06,309 --> 00:17:06,319
over and over again. So over the last
 

00:17:06,319 --> 00:17:09,110
over and over again. So over the last
six months we ran around 36,000

00:17:09,110 --> 00:17:09,120
six months we ran around 36,000
 

00:17:09,120 --> 00:17:12,230
six months we ran around 36,000
different reactions across 580 uh 34

00:17:12,230 --> 00:17:12,240
different reactions across 580 uh 34
 

00:17:12,240 --> 00:17:15,189
different reactions across 580 uh 34
well plates.

00:17:15,189 --> 00:17:15,199
well plates.
 

00:17:15,199 --> 00:17:19,189
well plates.
So um alongside the experiment designs

00:17:19,189 --> 00:17:19,199
So um alongside the experiment designs
 

00:17:19,199 --> 00:17:22,390
So um alongside the experiment designs
we also asked GPT5 to kind of write lab

00:17:22,390 --> 00:17:22,400
we also asked GPT5 to kind of write lab
 

00:17:22,400 --> 00:17:25,189
we also asked GPT5 to kind of write lab
notebooks and to document its kind of

00:17:25,189 --> 00:17:25,199
notebooks and to document its kind of
 

00:17:25,199 --> 00:17:26,870
notebooks and to document its kind of
biological reasoning throughout this

00:17:26,870 --> 00:17:26,880
biological reasoning throughout this
 

00:17:26,880 --> 00:17:29,110
biological reasoning throughout this
process. And part of this was because I

00:17:29,110 --> 00:17:29,120
process. And part of this was because I
 

00:17:29,120 --> 00:17:31,190
process. And part of this was because I
wanted to monitor the model to make sure

00:17:31,190 --> 00:17:31,200
wanted to monitor the model to make sure
 

00:17:31,200 --> 00:17:32,470
wanted to monitor the model to make sure
that it was actually doing something

00:17:32,470 --> 00:17:32,480
that it was actually doing something
 

00:17:32,480 --> 00:17:34,950
that it was actually doing something
sensible because one boring case you can

00:17:34,950 --> 00:17:34,960
sensible because one boring case you can
 

00:17:34,960 --> 00:17:36,549
sensible because one boring case you can
imagine is that the model just kind of

00:17:36,549 --> 00:17:36,559
imagine is that the model just kind of
 

00:17:36,559 --> 00:17:38,950
imagine is that the model just kind of
does a dumb grid search and submits that

00:17:38,950 --> 00:17:38,960
does a dumb grid search and submits that
 

00:17:38,960 --> 00:17:40,390
does a dumb grid search and submits that
as a play design that's not very

00:17:40,390 --> 00:17:40,400
as a play design that's not very
 

00:17:40,400 --> 00:17:43,669
as a play design that's not very
interesting. Um so this was actually an

00:17:43,669 --> 00:17:43,679
interesting. Um so this was actually an
 

00:17:43,679 --> 00:17:45,909
interesting. Um so this was actually an
excerpt from the first random lab

00:17:45,909 --> 00:17:45,919
excerpt from the first random lab
 

00:17:45,919 --> 00:17:48,630
excerpt from the first random lab
notebook that I opened. The part that is

00:17:48,630 --> 00:17:48,640
notebook that I opened. The part that is
 

00:17:48,640 --> 00:17:51,350
notebook that I opened. The part that is
bold um the NMP supplementation was in

00:17:51,350 --> 00:17:51,360
bold um the NMP supplementation was in
 

00:17:51,360 --> 00:17:52,870
bold um the NMP supplementation was in
situation.

00:17:52,870 --> 00:17:52,880
situation.
 

00:17:52,880 --> 00:17:55,110
situation.
This suggestion was actually the subject

00:17:55,110 --> 00:17:55,120
This suggestion was actually the subject
 

00:17:55,120 --> 00:17:57,350
This suggestion was actually the subject
of a recent paper that was told around

00:17:57,350 --> 00:17:57,360
of a recent paper that was told around
 

00:17:57,360 --> 00:17:59,990
of a recent paper that was told around
the time of our collaboration a mechan

00:17:59,990 --> 00:18:00,000
the time of our collaboration a mechan
 

00:18:00,000 --> 00:18:03,110
the time of our collaboration a mechan
lab and they had achieved a huge

00:18:03,110 --> 00:18:03,120
lab and they had achieved a huge
 

00:18:03,120 --> 00:18:06,470
lab and they had achieved a huge
reduction in the amount of cost per gram

00:18:06,470 --> 00:18:06,480
reduction in the amount of cost per gram
 

00:18:06,480 --> 00:18:07,750
reduction in the amount of cost per gram
of protein that they were able to get

00:18:07,750 --> 00:18:07,760
of protein that they were able to get
 

00:18:07,760 --> 00:18:10,310
of protein that they were able to get
from a self-free system. And so the

00:18:10,310 --> 00:18:10,320
from a self-free system. And so the
 

00:18:10,320 --> 00:18:12,070
from a self-free system. And so the
really interesting thing here is um this

00:18:12,070 --> 00:18:12,080
really interesting thing here is um this
 

00:18:12,080 --> 00:18:14,070
really interesting thing here is um this
bold line is actually one of the

00:18:14,070 --> 00:18:14,080
bold line is actually one of the
 

00:18:14,080 --> 00:18:16,150
bold line is actually one of the
critical insights that they had in this

00:18:16,150 --> 00:18:16,160
critical insights that they had in this
 

00:18:16,160 --> 00:18:18,710
critical insights that they had in this
paper but GD5 was not trained on this

00:18:18,710 --> 00:18:18,720
paper but GD5 was not trained on this
 

00:18:18,720 --> 00:18:20,470
paper but GD5 was not trained on this
paper because this paper had just come

00:18:20,470 --> 00:18:20,480
paper because this paper had just come
 

00:18:20,480 --> 00:18:23,430
paper because this paper had just come
out and also in the setting GPD5 didn't

00:18:23,430 --> 00:18:23,440
out and also in the setting GPD5 didn't
 

00:18:23,440 --> 00:18:25,750
out and also in the setting GPD5 didn't
have access to internet and so this is

00:18:25,750 --> 00:18:25,760
have access to internet and so this is
 

00:18:25,760 --> 00:18:27,510
have access to internet and so this is
actually just from knowledge that was

00:18:27,510 --> 00:18:27,520
actually just from knowledge that was
 

00:18:27,520 --> 00:18:30,870
actually just from knowledge that was
stored in the model weights and so this

00:18:30,870 --> 00:18:30,880
stored in the model weights and so this
 

00:18:30,880 --> 00:18:32,710
stored in the model weights and so this
was actually the initial sign of life to

00:18:32,710 --> 00:18:32,720
was actually the initial sign of life to
 

00:18:32,720 --> 00:18:34,950
was actually the initial sign of life to
me that the model had enough biology

00:18:34,950 --> 00:18:34,960
me that the model had enough biology
 

00:18:34,960 --> 00:18:36,710
me that the model had enough biology
knowledge in its model weights to

00:18:36,710 --> 00:18:36,720
knowledge in its model weights to
 

00:18:36,720 --> 00:18:38,549
knowledge in its model weights to
actually successfully do this

00:18:38,549 --> 00:18:38,559
actually successfully do this
 

00:18:38,559 --> 00:18:39,830
actually successfully do this
experiment.

00:18:39,830 --> 00:18:39,840
experiment.
 

00:18:39,840 --> 00:18:41,750
experiment.
Um something early on that we also had

00:18:41,750 --> 00:18:41,760
Um something early on that we also had
 

00:18:41,760 --> 00:18:45,029
Um something early on that we also had
to figure out too was um on almost every

00:18:45,029 --> 00:18:45,039
to figure out too was um on almost every
 

00:18:45,039 --> 00:18:47,190
to figure out too was um on almost every
roll out G5 was complaining about batch

00:18:47,190 --> 00:18:47,200
roll out G5 was complaining about batch
 

00:18:47,200 --> 00:18:48,630
roll out G5 was complaining about batch
effects and kind of just general

00:18:48,630 --> 00:18:48,640
effects and kind of just general
 

00:18:48,640 --> 00:18:52,310
effects and kind of just general
variability between replicates. So as we

00:18:52,310 --> 00:18:52,320
variability between replicates. So as we
 

00:18:52,320 --> 00:18:55,190
variability between replicates. So as we
all know biology is very noisy and so

00:18:55,190 --> 00:18:55,200
all know biology is very noisy and so
 

00:18:55,200 --> 00:18:57,110
all know biology is very noisy and so
for each experimental condition we were

00:18:57,110 --> 00:18:57,120
for each experimental condition we were
 

00:18:57,120 --> 00:18:58,470
for each experimental condition we were
testing we were doing them in

00:18:58,470 --> 00:18:58,480
testing we were doing them in
 

00:18:58,480 --> 00:19:00,150
testing we were doing them in
quadruplicates. So there were four wells

00:19:00,150 --> 00:19:00,160
quadruplicates. So there were four wells
 

00:19:00,160 --> 00:19:02,310
quadruplicates. So there were four wells
kind of in random positions on a 384

00:19:02,310 --> 00:19:02,320
kind of in random positions on a 384
 

00:19:02,320 --> 00:19:04,789
kind of in random positions on a 384
while play and G5D5 was just always

00:19:04,789 --> 00:19:04,799
while play and G5D5 was just always
 

00:19:04,799 --> 00:19:06,390
while play and G5D5 was just always
complaining about you know the same

00:19:06,390 --> 00:19:06,400
complaining about you know the same
 

00:19:06,400 --> 00:19:08,470
complaining about you know the same
experimental condition from playful play

00:19:08,470 --> 00:19:08,480
experimental condition from playful play
 

00:19:08,480 --> 00:19:11,430
experimental condition from playful play
is like not really um consistent and

00:19:11,430 --> 00:19:11,440
is like not really um consistent and
 

00:19:11,440 --> 00:19:12,950
is like not really um consistent and
then it's not consistent kind of within

00:19:12,950 --> 00:19:12,960
then it's not consistent kind of within
 

00:19:12,960 --> 00:19:15,830
then it's not consistent kind of within
plates and so there was actually a huge

00:19:15,830 --> 00:19:15,840
plates and so there was actually a huge
 

00:19:15,840 --> 00:19:17,590
plates and so there was actually a huge
amount of work from the GKO team to

00:19:17,590 --> 00:19:17,600
amount of work from the GKO team to
 

00:19:17,600 --> 00:19:19,750
amount of work from the GKO team to
reduce the coefficient of variation to

00:19:19,750 --> 00:19:19,760
reduce the coefficient of variation to
 

00:19:19,760 --> 00:19:21,990
reduce the coefficient of variation to
something that was much more usable and

00:19:21,990 --> 00:19:22,000
something that was much more usable and
 

00:19:22,000 --> 00:19:23,750
something that was much more usable and
then as said they were testing around

00:19:23,750 --> 00:19:23,760
then as said they were testing around
 

00:19:23,760 --> 00:19:26,789
then as said they were testing around
10,000 conditions per week.

00:19:26,789 --> 00:19:26,799
10,000 conditions per week.
 

00:19:26,799 --> 00:19:28,950
10,000 conditions per week.
Okay, so these were these are the

00:19:28,950 --> 00:19:28,960
Okay, so these were these are the
 

00:19:28,960 --> 00:19:30,870
Okay, so these were these are the
results um that we got out of the model.

00:19:30,870 --> 00:19:30,880
results um that we got out of the model.
 

00:19:30,880 --> 00:19:33,750
results um that we got out of the model.
The overall objective is to reduce the

00:19:33,750 --> 00:19:33,760
The overall objective is to reduce the
 

00:19:33,760 --> 00:19:35,990
The overall objective is to reduce the
cost per gram of protein that we're

00:19:35,990 --> 00:19:36,000
cost per gram of protein that we're
 

00:19:36,000 --> 00:19:37,990
cost per gram of protein that we're
making in this system. And we chose this

00:19:37,990 --> 00:19:38,000
making in this system. And we chose this
 

00:19:38,000 --> 00:19:39,830
making in this system. And we chose this
objective because I think just going for

00:19:39,830 --> 00:19:39,840
objective because I think just going for
 

00:19:39,840 --> 00:19:42,549
objective because I think just going for
tighter itself um is not that

00:19:42,549 --> 00:19:42,559
tighter itself um is not that
 

00:19:42,559 --> 00:19:43,990
tighter itself um is not that
interesting because you can imagine kind

00:19:43,990 --> 00:19:44,000
interesting because you can imagine kind
 

00:19:44,000 --> 00:19:45,430
interesting because you can imagine kind
of throwing all the most expensive

00:19:45,430 --> 00:19:45,440
of throwing all the most expensive
 

00:19:45,440 --> 00:19:47,590
of throwing all the most expensive
ingredients into a tube and that's not

00:19:47,590 --> 00:19:47,600
ingredients into a tube and that's not
 

00:19:47,600 --> 00:19:51,110
ingredients into a tube and that's not
really a scalable system. Um so here we

00:19:51,110 --> 00:19:51,120
really a scalable system. Um so here we
 

00:19:51,120 --> 00:19:53,430
really a scalable system. Um so here we
have the first round of experiments that

00:19:53,430 --> 00:19:53,440
have the first round of experiments that
 

00:19:53,440 --> 00:19:55,909
have the first round of experiments that
GBT designed. The horizontal axis is

00:19:55,909 --> 00:19:55,919
GBT designed. The horizontal axis is
 

00:19:55,919 --> 00:19:57,430
GBT designed. The horizontal axis is
amount of tighter that we were getting

00:19:57,430 --> 00:19:57,440
amount of tighter that we were getting
 

00:19:57,440 --> 00:19:59,750
amount of tighter that we were getting
out of these reactions in 3D 4 plates.

00:19:59,750 --> 00:19:59,760
out of these reactions in 3D 4 plates.
 

00:19:59,760 --> 00:20:01,990
out of these reactions in 3D 4 plates.
Um that's a grams per liter of protein

00:20:01,990 --> 00:20:02,000
Um that's a grams per liter of protein
 

00:20:02,000 --> 00:20:04,630
Um that's a grams per liter of protein
as measured by fluoresence. And on the

00:20:04,630 --> 00:20:04,640
as measured by fluoresence. And on the
 

00:20:04,640 --> 00:20:07,029
as measured by fluoresence. And on the
vertical axis we have the reaction cost

00:20:07,029 --> 00:20:07,039
vertical axis we have the reaction cost
 

00:20:07,039 --> 00:20:09,830
vertical axis we have the reaction cost
of each one of these reactions. And the

00:20:09,830 --> 00:20:09,840
of each one of these reactions. And the
 

00:20:09,840 --> 00:20:14,070
of each one of these reactions. And the
star here um is the soda 384 uh result

00:20:14,070 --> 00:20:14,080
star here um is the soda 384 uh result
 

00:20:14,080 --> 00:20:16,390
star here um is the soda 384 uh result
that came out of the dual lab. So this

00:20:16,390 --> 00:20:16,400
that came out of the dual lab. So this
 

00:20:16,400 --> 00:20:19,029
that came out of the dual lab. So this
is actually a really tiny um paper for

00:20:19,029 --> 00:20:19,039
is actually a really tiny um paper for
 

00:20:19,039 --> 00:20:21,110
is actually a really tiny um paper for
us because then we kind of had the uh

00:20:21,110 --> 00:20:21,120
us because then we kind of had the uh
 

00:20:21,120 --> 00:20:22,630
us because then we kind of had the uh
human gold standard to benchmark

00:20:22,630 --> 00:20:22,640
human gold standard to benchmark
 

00:20:22,640 --> 00:20:25,909
human gold standard to benchmark
against. And then in this setting GBD5

00:20:25,909 --> 00:20:25,919
against. And then in this setting GBD5
 

00:20:25,919 --> 00:20:28,549
against. And then in this setting GBD5
was just reasoning. It was given a

00:20:28,549 --> 00:20:28,559
was just reasoning. It was given a
 

00:20:28,559 --> 00:20:31,430
was just reasoning. It was given a
prompt about what self-free protein

00:20:31,430 --> 00:20:31,440
prompt about what self-free protein
 

00:20:31,440 --> 00:20:33,510
prompt about what self-free protein
synthesis roughly is. It was given the

00:20:33,510 --> 00:20:33,520
synthesis roughly is. It was given the
 

00:20:33,520 --> 00:20:36,710
synthesis roughly is. It was given the
pyantic model spec and we just told it

00:20:36,710 --> 00:20:36,720
pyantic model spec and we just told it
 

00:20:36,720 --> 00:20:38,870
pyantic model spec and we just told it
to write code to design 78 different

00:20:38,870 --> 00:20:38,880
to write code to design 78 different
 

00:20:38,880 --> 00:20:41,909
to write code to design 78 different
experiments for this like 384 while play

00:20:41,909 --> 00:20:41,919
experiments for this like 384 while play
 

00:20:41,919 --> 00:20:43,830
experiments for this like 384 while play
and we didn't really give it any inter

00:20:43,830 --> 00:20:43,840
and we didn't really give it any inter
 

00:20:43,840 --> 00:20:45,510
and we didn't really give it any inter
information and it didn't have access to

00:20:45,510 --> 00:20:45,520
information and it didn't have access to
 

00:20:45,520 --> 00:20:47,270
information and it didn't have access to
papers it didn't have access to internet

00:20:47,270 --> 00:20:47,280
papers it didn't have access to internet
 

00:20:47,280 --> 00:20:48,630
papers it didn't have access to internet
it were just kind of thinking in a

00:20:48,630 --> 00:20:48,640
it were just kind of thinking in a
 

00:20:48,640 --> 00:20:50,870
it were just kind of thinking in a
closed box and so these were the first

00:20:50,870 --> 00:20:50,880
closed box and so these were the first
 

00:20:50,880 --> 00:20:53,029
closed box and so these were the first
designs that I came up with. Um, I

00:20:53,029 --> 00:20:53,039
designs that I came up with. Um, I
 

00:20:53,039 --> 00:20:54,630
designs that I came up with. Um, I
remember at this point we were kind of

00:20:54,630 --> 00:20:54,640
remember at this point we were kind of
 

00:20:54,640 --> 00:20:56,710
remember at this point we were kind of
happy that it made protein at all

00:20:56,710 --> 00:20:56,720
happy that it made protein at all
 

00:20:56,720 --> 00:20:58,310
happy that it made protein at all
because I know that if you kind of gave

00:20:58,310 --> 00:20:58,320
because I know that if you kind of gave
 

00:20:58,320 --> 00:20:59,990
because I know that if you kind of gave
me, you know, put me in a closed room

00:20:59,990 --> 00:21:00,000
me, you know, put me in a closed room
 

00:21:00,000 --> 00:21:01,430
me, you know, put me in a closed room
and told me to write a reaction, I don't

00:21:01,430 --> 00:21:01,440
and told me to write a reaction, I don't
 

00:21:01,440 --> 00:21:02,630
and told me to write a reaction, I don't
think I would have gotten results this

00:21:02,630 --> 00:21:02,640
think I would have gotten results this
 

00:21:02,640 --> 00:21:04,870
think I would have gotten results this
good. Um, but that being said, you know,

00:21:04,870 --> 00:21:04,880
good. Um, but that being said, you know,
 

00:21:04,880 --> 00:21:06,710
good. Um, but that being said, you know,
kind of looking at the plot on the right

00:21:06,710 --> 00:21:06,720
kind of looking at the plot on the right
 

00:21:06,720 --> 00:21:09,110
kind of looking at the plot on the right
where we have um the minimum dollar

00:21:09,110 --> 00:21:09,120
where we have um the minimum dollar
 

00:21:09,120 --> 00:21:10,950
where we have um the minimum dollar
program that we can get. We're still, I

00:21:10,950 --> 00:21:10,960
program that we can get. We're still, I
 

00:21:10,960 --> 00:21:15,430
program that we can get. We're still, I
think, around um 14 x uh away from the

00:21:15,430 --> 00:21:15,440
think, around um 14 x uh away from the
 

00:21:15,440 --> 00:21:17,190
think, around um 14 x uh away from the
soda. So, there was still quite a bit of

00:21:17,190 --> 00:21:17,200
soda. So, there was still quite a bit of
 

00:21:17,200 --> 00:21:20,149
soda. So, there was still quite a bit of
room to go. Um, also on this stuff, we

00:21:20,149 --> 00:21:20,159
room to go. Um, also on this stuff, we
 

00:21:20,159 --> 00:21:21,830
room to go. Um, also on this stuff, we
noticed that the model was kind of

00:21:21,830 --> 00:21:21,840
noticed that the model was kind of
 

00:21:21,840 --> 00:21:23,909
noticed that the model was kind of
suggesting new reagents that we should

00:21:23,909 --> 00:21:23,919
suggesting new reagents that we should
 

00:21:23,919 --> 00:21:26,549
suggesting new reagents that we should
add. And so we actually had the model

00:21:26,549 --> 00:21:26,559
add. And so we actually had the model
 

00:21:26,559 --> 00:21:29,029
add. And so we actually had the model
rank all the reagents that it suggested

00:21:29,029 --> 00:21:29,039
rank all the reagents that it suggested
 

00:21:29,039 --> 00:21:31,270
rank all the reagents that it suggested
we should add to the pyentic reagent

00:21:31,270 --> 00:21:31,280
we should add to the pyentic reagent
 

00:21:31,280 --> 00:21:33,750
we should add to the pyentic reagent
list. Um, and then we ranked them and I

00:21:33,750 --> 00:21:33,760
list. Um, and then we ranked them and I
 

00:21:33,760 --> 00:21:36,310
list. Um, and then we ranked them and I
think we took around the first 25ish of

00:21:36,310 --> 00:21:36,320
think we took around the first 25ish of
 

00:21:36,320 --> 00:21:38,950
think we took around the first 25ish of
those reaction or sorry those reagents

00:21:38,950 --> 00:21:38,960
those reaction or sorry those reagents
 

00:21:38,960 --> 00:21:42,310
those reaction or sorry those reagents
and for the next step um, we gave it to

00:21:42,310 --> 00:21:42,320
and for the next step um, we gave it to
 

00:21:42,320 --> 00:21:44,549
and for the next step um, we gave it to
the model. So in general we basically

00:21:44,549 --> 00:21:44,559
the model. So in general we basically
 

00:21:44,559 --> 00:21:45,909
the model. So in general we basically
want to like move in the direction of

00:21:45,909 --> 00:21:45,919
want to like move in the direction of
 

00:21:45,919 --> 00:21:47,669
want to like move in the direction of
this blue arrow. We want to see this

00:21:47,669 --> 00:21:47,679
this blue arrow. We want to see this
 

00:21:47,679 --> 00:21:49,830
this blue arrow. We want to see this
kind of parto frontier move to the right

00:21:49,830 --> 00:21:49,840
kind of parto frontier move to the right
 

00:21:49,840 --> 00:21:51,270
kind of parto frontier move to the right
and downwards meaning that we're getting

00:21:51,270 --> 00:21:51,280
and downwards meaning that we're getting
 

00:21:51,280 --> 00:21:53,270
and downwards meaning that we're getting
more protein but also the reactions are

00:21:53,270 --> 00:21:53,280
more protein but also the reactions are
 

00:21:53,280 --> 00:21:55,830
more protein but also the reactions are
getting cheaper. So this is the next

00:21:55,830 --> 00:21:55,840
getting cheaper. So this is the next
 

00:21:55,840 --> 00:21:57,669
getting cheaper. So this is the next
step. Uh we see a decent amount of

00:21:57,669 --> 00:21:57,679
step. Uh we see a decent amount of
 

00:21:57,679 --> 00:21:59,430
step. Uh we see a decent amount of
improvement but it's not huge. We're

00:21:59,430 --> 00:21:59,440
improvement but it's not huge. We're
 

00:21:59,440 --> 00:22:01,110
improvement but it's not huge. We're
still many orders of magnitude away from

00:22:01,110 --> 00:22:01,120
still many orders of magnitude away from
 

00:22:01,120 --> 00:22:05,110
still many orders of magnitude away from
soda. And the difference here is um on

00:22:05,110 --> 00:22:05,120
soda. And the difference here is um on
 

00:22:05,120 --> 00:22:07,669
soda. And the difference here is um on
step two the model so as we were

00:22:07,669 --> 00:22:07,679
step two the model so as we were
 

00:22:07,679 --> 00:22:08,950
step two the model so as we were
designing these experiments the model

00:22:08,950 --> 00:22:08,960
designing these experiments the model
 

00:22:08,960 --> 00:22:11,190
designing these experiments the model
was able to use a new reagents. And so

00:22:11,190 --> 00:22:11,200
was able to use a new reagents. And so
 

00:22:11,200 --> 00:22:13,350
was able to use a new reagents. And so
we see that step three we get quite a

00:22:13,350 --> 00:22:13,360
we see that step three we get quite a
 

00:22:13,360 --> 00:22:16,630
we see that step three we get quite a
bit better but we are kind of hitting so

00:22:16,630 --> 00:22:16,640
bit better but we are kind of hitting so
 

00:22:16,640 --> 00:22:20,390
bit better but we are kind of hitting so
as you go to the left um we see that the

00:22:20,390 --> 00:22:20,400
as you go to the left um we see that the
 

00:22:20,400 --> 00:22:22,789
as you go to the left um we see that the
model is like kind of not able to go

00:22:22,789 --> 00:22:22,799
model is like kind of not able to go
 

00:22:22,799 --> 00:22:25,990
model is like kind of not able to go
down further in terms of like the cost

00:22:25,990 --> 00:22:26,000
down further in terms of like the cost
 

00:22:26,000 --> 00:22:27,669
down further in terms of like the cost
and we also want to move a little bit

00:22:27,669 --> 00:22:27,679
and we also want to move a little bit
 

00:22:27,679 --> 00:22:29,830
and we also want to move a little bit
further too in terms of tighter. And so

00:22:29,830 --> 00:22:29,840
further too in terms of tighter. And so
 

00:22:29,840 --> 00:22:32,149
further too in terms of tighter. And so
at this point we thought okay we know

00:22:32,149 --> 00:22:32,159
at this point we thought okay we know
 

00:22:32,159 --> 00:22:33,750
at this point we thought okay we know
that the model is able to make these

00:22:33,750 --> 00:22:33,760
that the model is able to make these
 

00:22:33,760 --> 00:22:36,470
that the model is able to make these
incremental improvements um just kind of

00:22:36,470 --> 00:22:36,480
incremental improvements um just kind of
 

00:22:36,480 --> 00:22:38,070
incremental improvements um just kind of
based on knowledge in this model weights

00:22:38,070 --> 00:22:38,080
based on knowledge in this model weights
 

00:22:38,080 --> 00:22:39,990
based on knowledge in this model weights
but this is not how a human scientist

00:22:39,990 --> 00:22:40,000
but this is not how a human scientist
 

00:22:40,000 --> 00:22:42,549
but this is not how a human scientist
will work. We would be able to read all

00:22:42,549 --> 00:22:42,559
will work. We would be able to read all
 

00:22:42,559 --> 00:22:43,830
will work. We would be able to read all
the previous literature. We will be able

00:22:43,830 --> 00:22:43,840
the previous literature. We will be able
 

00:22:43,840 --> 00:22:45,190
the previous literature. We will be able
to search the internet. We have access

00:22:45,190 --> 00:22:45,200
to search the internet. We have access
 

00:22:45,200 --> 00:22:46,950
to search the internet. We have access
to a computer to kind of like do

00:22:46,950 --> 00:22:46,960
to a computer to kind of like do
 

00:22:46,960 --> 00:22:49,110
to a computer to kind of like do
different kinds of data analysis. So we

00:22:49,110 --> 00:22:49,120
different kinds of data analysis. So we
 

00:22:49,120 --> 00:22:50,549
different kinds of data analysis. So we
thought let's just throw all these tools

00:22:50,549 --> 00:22:50,559
thought let's just throw all these tools
 

00:22:50,559 --> 00:22:52,390
thought let's just throw all these tools
at the model instead of kind of doing

00:22:52,390 --> 00:22:52,400
at the model instead of kind of doing
 

00:22:52,400 --> 00:22:54,789
at the model instead of kind of doing
this like artificial evaluation of like

00:22:54,789 --> 00:22:54,799
this like artificial evaluation of like
 

00:22:54,799 --> 00:22:57,110
this like artificial evaluation of like
how much internal knowledge a model has.

00:22:57,110 --> 00:22:57,120
how much internal knowledge a model has.
 

00:22:57,120 --> 00:22:59,830
how much internal knowledge a model has.
And so for the next step um this is with

00:22:59,830 --> 00:22:59,840
And so for the next step um this is with
 

00:22:59,840 --> 00:23:01,669
And so for the next step um this is with
full use and we see this huge jump. So

00:23:01,669 --> 00:23:01,679
full use and we see this huge jump. So
 

00:23:01,679 --> 00:23:04,470
full use and we see this huge jump. So
we actually immediately exceed the 384

00:23:04,470 --> 00:23:04,480
we actually immediately exceed the 384
 

00:23:04,480 --> 00:23:07,590
we actually immediately exceed the 384
wall soda um state-of-the-art. And so

00:23:07,590 --> 00:23:07,600
wall soda um state-of-the-art. And so
 

00:23:07,600 --> 00:23:08,710
wall soda um state-of-the-art. And so
now we actually kind of have to like

00:23:08,710 --> 00:23:08,720
now we actually kind of have to like
 

00:23:08,720 --> 00:23:10,789
now we actually kind of have to like
zoom in in this region. So for the next

00:23:10,789 --> 00:23:10,799
zoom in in this region. So for the next
 

00:23:10,799 --> 00:23:12,390
zoom in in this region. So for the next
part we're kind of switching from lo

00:23:12,390 --> 00:23:12,400
part we're kind of switching from lo
 

00:23:12,400 --> 00:23:14,070
part we're kind of switching from lo
scale now to linear scale to make things

00:23:14,070 --> 00:23:14,080
scale now to linear scale to make things
 

00:23:14,080 --> 00:23:17,029
scale now to linear scale to make things
a little bit easier to see. So the tools

00:23:17,029 --> 00:23:17,039
a little bit easier to see. So the tools
 

00:23:17,039 --> 00:23:18,950
a little bit easier to see. So the tools
here the model has access to a computer

00:23:18,950 --> 00:23:18,960
here the model has access to a computer
 

00:23:18,960 --> 00:23:21,190
here the model has access to a computer
and a browser. And we see kind of from

00:23:21,190 --> 00:23:21,200
and a browser. And we see kind of from
 

00:23:21,200 --> 00:23:22,710
and a browser. And we see kind of from
steps three to four the model gets

00:23:22,710 --> 00:23:22,720
steps three to four the model gets
 

00:23:22,720 --> 00:23:24,710
steps three to four the model gets
better. And then step five the model

00:23:24,710 --> 00:23:24,720
better. And then step five the model
 

00:23:24,720 --> 00:23:26,710
better. And then step five the model
gets even better. And so we can also

00:23:26,710 --> 00:23:26,720
gets even better. And so we can also
 

00:23:26,720 --> 00:23:28,789
gets even better. And so we can also
kind of see the progress um on the right

00:23:28,789 --> 00:23:28,799
kind of see the progress um on the right
 

00:23:28,799 --> 00:23:30,310
kind of see the progress um on the right
as well where we're kind of just like

00:23:30,310 --> 00:23:30,320
as well where we're kind of just like
 

00:23:30,320 --> 00:23:33,350
as well where we're kind of just like
getting much better with with each step.

00:23:33,350 --> 00:23:33,360
getting much better with with each step.
 

00:23:33,360 --> 00:23:36,230
getting much better with with each step.
Um so overall when we retested these

00:23:36,230 --> 00:23:36,240
Um so overall when we retested these
 

00:23:36,240 --> 00:23:38,149
Um so overall when we retested these
reactions in two ML tubes where now we

00:23:38,149 --> 00:23:38,159
reactions in two ML tubes where now we
 

00:23:38,159 --> 00:23:40,789
reactions in two ML tubes where now we
kind of have more oxygenation due to the

00:23:40,789 --> 00:23:40,799
kind of have more oxygenation due to the
 

00:23:40,799 --> 00:23:43,350
kind of have more oxygenation due to the
better geometry of two mel tubes uh we

00:23:43,350 --> 00:23:43,360
better geometry of two mel tubes uh we
 

00:23:43,360 --> 00:23:45,510
better geometry of two mel tubes uh we
see a 40% reduction in the total

00:23:45,510 --> 00:23:45,520
see a 40% reduction in the total
 

00:23:45,520 --> 00:23:46,950
see a 40% reduction in the total
reaction cost and then when you look at

00:23:46,950 --> 00:23:46,960
reaction cost and then when you look at
 

00:23:46,960 --> 00:23:49,750
reaction cost and then when you look at
the reagents alone it's actually 57%.

00:23:49,750 --> 00:23:49,760
the reagents alone it's actually 57%.
 

00:23:49,760 --> 00:23:52,470
the reagents alone it's actually 57%.
So it was a pretty um decent uh

00:23:52,470 --> 00:23:52,480
So it was a pretty um decent uh
 

00:23:52,480 --> 00:23:54,470
So it was a pretty um decent uh
achievement I think for the model where

00:23:54,470 --> 00:23:54,480
achievement I think for the model where
 

00:23:54,480 --> 00:23:56,230
achievement I think for the model where
from the beginning we weren't sure if we

00:23:56,230 --> 00:23:56,240
from the beginning we weren't sure if we
 

00:23:56,240 --> 00:23:58,870
from the beginning we weren't sure if we
could design even a single reaction. Um

00:23:58,870 --> 00:23:58,880
could design even a single reaction. Um
 

00:23:58,880 --> 00:24:00,630
could design even a single reaction. Um
the other interesting thing here too is

00:24:00,630 --> 00:24:00,640
the other interesting thing here too is
 

00:24:00,640 --> 00:24:02,950
the other interesting thing here too is
that because we were optimizing at this

00:24:02,950 --> 00:24:02,960
that because we were optimizing at this
 

00:24:02,960 --> 00:24:05,669
that because we were optimizing at this
384 well plate level um not sure how

00:24:05,669 --> 00:24:05,679
384 well plate level um not sure how
 

00:24:05,679 --> 00:24:07,830
384 well plate level um not sure how
easy these numbers are to see but you

00:24:07,830 --> 00:24:07,840
easy these numbers are to see but you
 

00:24:07,840 --> 00:24:09,990
easy these numbers are to see but you
can kind of see like the um RF opt which

00:24:09,990 --> 00:24:10,000
can kind of see like the um RF opt which
 

00:24:10,000 --> 00:24:11,750
can kind of see like the um RF opt which
is the result that came out of the dual

00:24:11,750 --> 00:24:11,760
is the result that came out of the dual
 

00:24:11,760 --> 00:24:15,190
is the result that came out of the dual
lab with the human gold standard um it's

00:24:15,190 --> 00:24:15,200
lab with the human gold standard um it's
 

00:24:15,200 --> 00:24:17,269
lab with the human gold standard um it's
actually much lower and tighter at the

00:24:17,269 --> 00:24:17,279
actually much lower and tighter at the
 

00:24:17,279 --> 00:24:19,510
actually much lower and tighter at the
384 well level than most of our

00:24:19,510 --> 00:24:19,520
384 well level than most of our
 

00:24:19,520 --> 00:24:21,350
384 well level than most of our
reactions and I think this is just

00:24:21,350 --> 00:24:21,360
reactions and I think this is just
 

00:24:21,360 --> 00:24:23,190
reactions and I think this is just
because the model is designing

00:24:23,190 --> 00:24:23,200
because the model is designing
 

00:24:23,200 --> 00:24:24,549
because the model is designing
experiments for this kind of like low

00:24:24,549 --> 00:24:24,559
experiments for this kind of like low
 

00:24:24,559 --> 00:24:27,510
experiments for this kind of like low
oxygenation condition in the 384 well

00:24:27,510 --> 00:24:27,520
oxygenation condition in the 384 well
 

00:24:27,520 --> 00:24:29,350
oxygenation condition in the 384 well
geometry.

00:24:29,350 --> 00:24:29,360
geometry.
 

00:24:29,360 --> 00:24:31,029
geometry.
Um, and here I'm just kind of like

00:24:31,029 --> 00:24:31,039
Um, and here I'm just kind of like
 

00:24:31,039 --> 00:24:32,789
Um, and here I'm just kind of like
showing you some concrete examples of

00:24:32,789 --> 00:24:32,799
showing you some concrete examples of
 

00:24:32,799 --> 00:24:34,390
showing you some concrete examples of
like what the code the model actually

00:24:34,390 --> 00:24:34,400
like what the code the model actually
 

00:24:34,400 --> 00:24:36,390
like what the code the model actually
saw looks like. I believe this is

00:24:36,390 --> 00:24:36,400
saw looks like. I believe this is
 

00:24:36,400 --> 00:24:38,070
saw looks like. I believe this is
actually open source. So you could also

00:24:38,070 --> 00:24:38,080
actually open source. So you could also
 

00:24:38,080 --> 00:24:39,830
actually open source. So you could also
browse this online on GitHub if you want

00:24:39,830 --> 00:24:39,840
browse this online on GitHub if you want
 

00:24:39,840 --> 00:24:42,310
browse this online on GitHub if you want
to. Um, but yeah, we basically just

00:24:42,310 --> 00:24:42,320
to. Um, but yeah, we basically just
 

00:24:42,320 --> 00:24:43,909
to. Um, but yeah, we basically just
constrained action space of the model so

00:24:43,909 --> 00:24:43,919
constrained action space of the model so
 

00:24:43,919 --> 00:24:45,669
constrained action space of the model so
it can focus on kind of the biochemistry

00:24:45,669 --> 00:24:45,679
it can focus on kind of the biochemistry
 

00:24:45,679 --> 00:24:47,510
it can focus on kind of the biochemistry
and not what's physically executable on

00:24:47,510 --> 00:24:47,520
and not what's physically executable on
 

00:24:47,520 --> 00:24:50,310
and not what's physically executable on
the racks. So it was just sending these

00:24:50,310 --> 00:24:50,320
the racks. So it was just sending these
 

00:24:50,320 --> 00:24:52,470
the racks. So it was just sending these
uh plate objects kind of back and forth

00:24:52,470 --> 00:24:52,480
uh plate objects kind of back and forth
 

00:24:52,480 --> 00:24:55,510
uh plate objects kind of back and forth
between GKO and OpenAI. And the play has

00:24:55,510 --> 00:24:55,520
between GKO and OpenAI. And the play has
 

00:24:55,520 --> 00:24:57,510
between GKO and OpenAI. And the play has
a list of sample objects. that has kind

00:24:57,510 --> 00:24:57,520
a list of sample objects. that has kind
 

00:24:57,520 --> 00:24:59,990
a list of sample objects. that has kind
of the results of the experiment. Um and

00:24:59,990 --> 00:25:00,000
of the results of the experiment. Um and
 

00:25:00,000 --> 00:25:01,269
of the results of the experiment. Um and
also you see there are things like

00:25:01,269 --> 00:25:01,279
also you see there are things like
 

00:25:01,279 --> 00:25:03,590
also you see there are things like
reagent flags of what um if like a

00:25:03,590 --> 00:25:03,600
reagent flags of what um if like a
 

00:25:03,600 --> 00:25:05,750
reagent flags of what um if like a
reagent is out of stock. Um also

00:25:05,750 --> 00:25:05,760
reagent is out of stock. Um also
 

00:25:05,760 --> 00:25:08,230
reagent is out of stock. Um also
metadata on you know different events

00:25:08,230 --> 00:25:08,240
metadata on you know different events
 

00:25:08,240 --> 00:25:09,510
metadata on you know different events
that might have occurred during the

00:25:09,510 --> 00:25:09,520
that might have occurred during the
 

00:25:09,520 --> 00:25:12,310
that might have occurred during the
execution of an experiment. Um and then

00:25:12,310 --> 00:25:12,320
execution of an experiment. Um and then
 

00:25:12,320 --> 00:25:15,110
execution of an experiment. Um and then
we also have validation for all the

00:25:15,110 --> 00:25:15,120
we also have validation for all the
 

00:25:15,120 --> 00:25:16,950
we also have validation for all the
different parts of the plates. And this

00:25:16,950 --> 00:25:16,960
different parts of the plates. And this
 

00:25:16,960 --> 00:25:18,549
different parts of the plates. And this
just you know make sure that for example

00:25:18,549 --> 00:25:18,559
just you know make sure that for example
 

00:25:18,559 --> 00:25:19,990
just you know make sure that for example
like a plate has the right number of

00:25:19,990 --> 00:25:20,000
like a plate has the right number of
 

00:25:20,000 --> 00:25:21,990
like a plate has the right number of
samples. It has a sufficient number of

00:25:21,990 --> 00:25:22,000
samples. It has a sufficient number of
 

00:25:22,000 --> 00:25:26,070
samples. It has a sufficient number of
replicates. And what this um gives us is

00:25:26,070 --> 00:25:26,080
replicates. And what this um gives us is
 

00:25:26,080 --> 00:25:29,669
replicates. And what this um gives us is
that oops uh we know that if a plate

00:25:29,669 --> 00:25:29,679
that oops uh we know that if a plate
 

00:25:29,679 --> 00:25:31,830
that oops uh we know that if a plate
passes all these validations then this

00:25:31,830 --> 00:25:31,840
passes all these validations then this
 

00:25:31,840 --> 00:25:33,750
passes all these validations then this
experiment is executable on the rack

00:25:33,750 --> 00:25:33,760
experiment is executable on the rack
 

00:25:33,760 --> 00:25:36,789
experiment is executable on the rack
system. So this gives us a really nice

00:25:36,789 --> 00:25:36,799
system. So this gives us a really nice
 

00:25:36,799 --> 00:25:38,549
system. So this gives us a really nice
way to kind of programmatically check

00:25:38,549 --> 00:25:38,559
way to kind of programmatically check
 

00:25:38,559 --> 00:25:40,310
way to kind of programmatically check
all the rollouts and the outputs are

00:25:40,310 --> 00:25:40,320
all the rollouts and the outputs are
 

00:25:40,320 --> 00:25:42,230
all the rollouts and the outputs are
coming from the model without you know

00:25:42,230 --> 00:25:42,240
coming from the model without you know
 

00:25:42,240 --> 00:25:43,750
coming from the model without you know
trying to run it on the rack and then

00:25:43,750 --> 00:25:43,760
trying to run it on the rack and then
 

00:25:43,760 --> 00:25:45,350
trying to run it on the rack and then
only later to realize that it was

00:25:45,350 --> 00:25:45,360
only later to realize that it was
 

00:25:45,360 --> 00:25:46,710
only later to realize that it was
broken.

00:25:46,710 --> 00:25:46,720
broken.
 

00:25:46,720 --> 00:25:49,430
broken.
Um let's see I have a few more of these

00:25:49,430 --> 00:25:49,440
Um let's see I have a few more of these
 

00:25:49,440 --> 00:25:52,070
Um let's see I have a few more of these
where yeah we're validating that the

00:25:52,070 --> 00:25:52,080
where yeah we're validating that the
 

00:25:52,080 --> 00:25:54,070
where yeah we're validating that the
reagents are available. We're validating

00:25:54,070 --> 00:25:54,080
reagents are available. We're validating
 

00:25:54,080 --> 00:25:56,230
reagents are available. We're validating
that a model can only add things in 25

00:25:56,230 --> 00:25:56,240
that a model can only add things in 25
 

00:25:56,240 --> 00:25:57,990
that a model can only add things in 25
nanol increments because this is a

00:25:57,990 --> 00:25:58,000
nanol increments because this is a
 

00:25:58,000 --> 00:25:59,430
nanol increments because this is a
limitation the physical limitation of

00:25:59,430 --> 00:25:59,440
limitation the physical limitation of
 

00:25:59,440 --> 00:26:02,390
limitation the physical limitation of
the rack system. um non-zero default

00:26:02,390 --> 00:26:02,400
the rack system. um non-zero default
 

00:26:02,400 --> 00:26:04,390
the rack system. um non-zero default
values is is interesting because I think

00:26:04,390 --> 00:26:04,400
values is is interesting because I think
 

00:26:04,400 --> 00:26:06,789
values is is interesting because I think
we added this after we noticed um there

00:26:06,789 --> 00:26:06,799
we added this after we noticed um there
 

00:26:06,799 --> 00:26:08,710
we added this after we noticed um there
was one plate where the model cheated

00:26:08,710 --> 00:26:08,720
was one plate where the model cheated
 

00:26:08,720 --> 00:26:10,070
was one plate where the model cheated
where it was trying to cram a lot of

00:26:10,070 --> 00:26:10,080
where it was trying to cram a lot of
 

00:26:10,080 --> 00:26:12,549
where it was trying to cram a lot of
reagents into a single reaction and it

00:26:12,549 --> 00:26:12,559
reagents into a single reaction and it
 

00:26:12,559 --> 00:26:13,990
reagents into a single reaction and it
was kind of running out of volume and so

00:26:13,990 --> 00:26:14,000
was kind of running out of volume and so
 

00:26:14,000 --> 00:26:16,549
was kind of running out of volume and so
what it did was that it um indicated a

00:26:16,549 --> 00:26:16,559
what it did was that it um indicated a
 

00:26:16,559 --> 00:26:18,149
what it did was that it um indicated a
negative volume from the reagents and

00:26:18,149 --> 00:26:18,159
negative volume from the reagents and
 

00:26:18,159 --> 00:26:19,669
negative volume from the reagents and
this way it was able to kind of add more

00:26:19,669 --> 00:26:19,679
this way it was able to kind of add more
 

00:26:19,679 --> 00:26:21,990
this way it was able to kind of add more
things in and so we made it stop doing

00:26:21,990 --> 00:26:22,000
things in and so we made it stop doing
 

00:26:22,000 --> 00:26:23,269
things in and so we made it stop doing
that by kind of adding this extra

00:26:23,269 --> 00:26:23,279
that by kind of adding this extra
 

00:26:23,279 --> 00:26:28,390
that by kind of adding this extra
pyantic check. So um we also had to do

00:26:28,390 --> 00:26:28,400
pyantic check. So um we also had to do
 

00:26:28,400 --> 00:26:30,630
pyantic check. So um we also had to do
some kind of ranking on the designs that

00:26:30,630 --> 00:26:30,640
some kind of ranking on the designs that
 

00:26:30,640 --> 00:26:33,350
some kind of ranking on the designs that
we're sending to GKO and this is because

00:26:33,350 --> 00:26:33,360
we're sending to GKO and this is because
 

00:26:33,360 --> 00:26:35,110
we're sending to GKO and this is because
not all plates that the model design

00:26:35,110 --> 00:26:35,120
not all plates that the model design
 

00:26:35,120 --> 00:26:37,029
not all plates that the model design
path of identical validation. So we were

00:26:37,029 --> 00:26:37,039
path of identical validation. So we were
 

00:26:37,039 --> 00:26:38,470
path of identical validation. So we were
generating kind of 2x the number of

00:26:38,470 --> 00:26:38,480
generating kind of 2x the number of
 

00:26:38,480 --> 00:26:41,190
generating kind of 2x the number of
target plates. So if we're targeting 128

00:26:41,190 --> 00:26:41,200
target plates. So if we're targeting 128
 

00:26:41,200 --> 00:26:44,070
target plates. So if we're targeting 128
would generate 256 plates and then this

00:26:44,070 --> 00:26:44,080
would generate 256 plates and then this
 

00:26:44,080 --> 00:26:45,510
would generate 256 plates and then this
will give us kind of a surplus of plates

00:26:45,510 --> 00:26:45,520
will give us kind of a surplus of plates
 

00:26:45,520 --> 00:26:47,190
will give us kind of a surplus of plates
and we had to rank them somehow instead

00:26:47,190 --> 00:26:47,200
and we had to rank them somehow instead
 

00:26:47,200 --> 00:26:50,149
and we had to rank them somehow instead
of um randomly sending them to go. So we

00:26:50,149 --> 00:26:50,159
of um randomly sending them to go. So we
 

00:26:50,159 --> 00:26:52,310
of um randomly sending them to go. So we
tried a different her uh heristics for

00:26:52,310 --> 00:26:52,320
tried a different her uh heristics for
 

00:26:52,320 --> 00:26:54,470
tried a different her uh heristics for
ranking. I don't think um any of them

00:26:54,470 --> 00:26:54,480
ranking. I don't think um any of them
 

00:26:54,480 --> 00:26:56,310
ranking. I don't think um any of them
was really super correlated to the

00:26:56,310 --> 00:26:56,320
was really super correlated to the
 

00:26:56,320 --> 00:26:57,750
was really super correlated to the
outcome, but I'll just kind of talk

00:26:57,750 --> 00:26:57,760
outcome, but I'll just kind of talk
 

00:26:57,760 --> 00:26:59,909
outcome, but I'll just kind of talk
about what we did. Anyway, um the first

00:26:59,909 --> 00:26:59,919
about what we did. Anyway, um the first
 

00:26:59,919 --> 00:27:01,269
about what we did. Anyway, um the first
one we did was kind of like this like

00:27:01,269 --> 00:27:01,279
one we did was kind of like this like
 

00:27:01,279 --> 00:27:03,669
one we did was kind of like this like
vibrating regime where we had GPD5 then

00:27:03,669 --> 00:27:03,679
vibrating regime where we had GPD5 then
 

00:27:03,679 --> 00:27:05,590
vibrating regime where we had GPD5 then
act as a PI and we told it here's an

00:27:05,590 --> 00:27:05,600
act as a PI and we told it here's an
 

00:27:05,600 --> 00:27:07,110
act as a PI and we told it here's an
experimental design from a postto in

00:27:07,110 --> 00:27:07,120
experimental design from a postto in
 

00:27:07,120 --> 00:27:09,190
experimental design from a postto in
your lab. Um the experimental design was

00:27:09,190 --> 00:27:09,200
your lab. Um the experimental design was
 

00:27:09,200 --> 00:27:10,630
your lab. Um the experimental design was
just the play object and the lab

00:27:10,630 --> 00:27:10,640
just the play object and the lab
 

00:27:10,640 --> 00:27:12,470
just the play object and the lab
notebook and you can kind of like grade

00:27:12,470 --> 00:27:12,480
notebook and you can kind of like grade
 

00:27:12,480 --> 00:27:14,630
notebook and you can kind of like grade
it on how good it is kind of going from

00:27:14,630 --> 00:27:14,640
it on how good it is kind of going from
 

00:27:14,640 --> 00:27:16,390
it on how good it is kind of going from
like a scale of like blind unguided

00:27:16,390 --> 00:27:16,400
like a scale of like blind unguided
 

00:27:16,400 --> 00:27:18,149
like a scale of like blind unguided
search to like a very good sound

00:27:18,149 --> 00:27:18,159
search to like a very good sound
 

00:27:18,159 --> 00:27:20,789
search to like a very good sound
scientifically rigorous experiment. Um

00:27:20,789 --> 00:27:20,799
scientifically rigorous experiment. Um
 

00:27:20,799 --> 00:27:22,870
scientifically rigorous experiment. Um
so that was a one way of ranking. The

00:27:22,870 --> 00:27:22,880
so that was a one way of ranking. The
 

00:27:22,880 --> 00:27:25,510
so that was a one way of ranking. The
other one was um this idea of kind of

00:27:25,510 --> 00:27:25,520
other one was um this idea of kind of
 

00:27:25,520 --> 00:27:27,830
other one was um this idea of kind of
like checking your work instead of the

00:27:27,830 --> 00:27:27,840
like checking your work instead of the
 

00:27:27,840 --> 00:27:29,110
like checking your work instead of the
final result because we don't know what

00:27:29,110 --> 00:27:29,120
final result because we don't know what
 

00:27:29,120 --> 00:27:31,190
final result because we don't know what
the final result is. So we kind of

00:27:31,190 --> 00:27:31,200
the final result is. So we kind of
 

00:27:31,200 --> 00:27:33,669
the final result is. So we kind of
looked at all the different actions the

00:27:33,669 --> 00:27:33,679
looked at all the different actions the
 

00:27:33,679 --> 00:27:35,590
looked at all the different actions the
model was taking on the computer as it

00:27:35,590 --> 00:27:35,600
model was taking on the computer as it
 

00:27:35,600 --> 00:27:37,830
model was taking on the computer as it
was designing the plate and then we had

00:27:37,830 --> 00:27:37,840
was designing the plate and then we had
 

00:27:37,840 --> 00:27:40,070
was designing the plate and then we had
a separate model come in look at its

00:27:40,070 --> 00:27:40,080
a separate model come in look at its
 

00:27:40,080 --> 00:27:42,549
a separate model come in look at its
work and say okay did it do kind of the

00:27:42,549 --> 00:27:42,559
work and say okay did it do kind of the
 

00:27:42,559 --> 00:27:44,870
work and say okay did it do kind of the
right things that make sense. So some of

00:27:44,870 --> 00:27:44,880
right things that make sense. So some of
 

00:27:44,880 --> 00:27:46,710
right things that make sense. So some of
these are pretty obvious. Did you add

00:27:46,710 --> 00:27:46,720
these are pretty obvious. Did you add
 

00:27:46,720 --> 00:27:49,029
these are pretty obvious. Did you add
like at least look at the raw files from

00:27:49,029 --> 00:27:49,039
like at least look at the raw files from
 

00:27:49,039 --> 00:27:51,110
like at least look at the raw files from
the experiments? Um did you look at the

00:27:51,110 --> 00:27:51,120
the experiments? Um did you look at the
 

00:27:51,120 --> 00:27:52,950
the experiments? Um did you look at the
reagent loss to think about batch

00:27:52,950 --> 00:27:52,960
reagent loss to think about batch
 

00:27:52,960 --> 00:27:55,029
reagent loss to think about batch
effects in the reagents? Um there's also

00:27:55,029 --> 00:27:55,039
effects in the reagents? Um there's also
 

00:27:55,039 --> 00:27:56,549
effects in the reagents? Um there's also
a time point field and experimental

00:27:56,549 --> 00:27:56,559
a time point field and experimental
 

00:27:56,559 --> 00:27:58,149
a time point field and experimental
results metadata. Did you look at that

00:27:58,149 --> 00:27:58,159
results metadata. Did you look at that
 

00:27:58,159 --> 00:28:00,549
results metadata. Did you look at that
that field? And this really guards

00:28:00,549 --> 00:28:00,559
that field? And this really guards
 

00:28:00,559 --> 00:28:03,110
that field? And this really guards
against kind of um designs from the

00:28:03,110 --> 00:28:03,120
against kind of um designs from the
 

00:28:03,120 --> 00:28:04,549
against kind of um designs from the
model where it's actually just like a

00:28:04,549 --> 00:28:04,559
model where it's actually just like a
 

00:28:04,559 --> 00:28:06,950
model where it's actually just like a
grid search and it's like not very not

00:28:06,950 --> 00:28:06,960
grid search and it's like not very not
 

00:28:06,960 --> 00:28:08,870
grid search and it's like not very not
thinking really hard. So this kind of

00:28:08,870 --> 00:28:08,880
thinking really hard. So this kind of
 

00:28:08,880 --> 00:28:11,669
thinking really hard. So this kind of
filters for the smart designs. Um and I

00:28:11,669 --> 00:28:11,679
filters for the smart designs. Um and I
 

00:28:11,679 --> 00:28:12,710
filters for the smart designs. Um and I
think the fact that this is not

00:28:12,710 --> 00:28:12,720
think the fact that this is not
 

00:28:12,720 --> 00:28:14,230
think the fact that this is not
correlated that well with the final

00:28:14,230 --> 00:28:14,240
correlated that well with the final
 

00:28:14,240 --> 00:28:15,909
correlated that well with the final
titers is just maybe sometimes in

00:28:15,909 --> 00:28:15,919
titers is just maybe sometimes in
 

00:28:15,919 --> 00:28:17,430
titers is just maybe sometimes in
science you do all the right things but

00:28:17,430 --> 00:28:17,440
science you do all the right things but
 

00:28:17,440 --> 00:28:18,950
science you do all the right things but
you don't get the wild results that you

00:28:18,950 --> 00:28:18,960
you don't get the wild results that you
 

00:28:18,960 --> 00:28:23,190
you don't get the wild results that you
hope for. Um and then finally 5 wrote

00:28:23,190 --> 00:28:23,200
hope for. Um and then finally 5 wrote
 

00:28:23,200 --> 00:28:25,350
hope for. Um and then finally 5 wrote
some interesting uh lab notebook

00:28:25,350 --> 00:28:25,360
some interesting uh lab notebook
 

00:28:25,360 --> 00:28:27,430
some interesting uh lab notebook
entries. I think these slides will be

00:28:27,430 --> 00:28:27,440
entries. I think these slides will be
 

00:28:27,440 --> 00:28:28,950
entries. I think these slides will be
online so maybe people can read them a

00:28:28,950 --> 00:28:28,960
online so maybe people can read them a
 

00:28:28,960 --> 00:28:29,990
online so maybe people can read them a
little bit later but it does like

00:28:29,990 --> 00:28:30,000
little bit later but it does like
 

00:28:30,000 --> 00:28:31,510
little bit later but it does like
consider a lot of biochemistry of these

00:28:31,510 --> 00:28:31,520
consider a lot of biochemistry of these
 

00:28:31,520 --> 00:28:33,750
consider a lot of biochemistry of these
reactions. Um I don't understand most of

00:28:33,750 --> 00:28:33,760
reactions. Um I don't understand most of
 

00:28:33,760 --> 00:28:35,510
reactions. Um I don't understand most of
these. They kind of like vaguely make

00:28:35,510 --> 00:28:35,520
these. They kind of like vaguely make
 

00:28:35,520 --> 00:28:37,750
these. They kind of like vaguely make
sense to me. So it thinks about core

00:28:37,750 --> 00:28:37,760
sense to me. So it thinks about core
 

00:28:37,760 --> 00:28:41,430
sense to me. So it thinks about core
ions and buffers um energy modules that

00:28:41,430 --> 00:28:41,440
ions and buffers um energy modules that
 

00:28:41,440 --> 00:28:44,789
ions and buffers um energy modules that
rivals glucose and MPS um different

00:28:44,789 --> 00:28:44,799
rivals glucose and MPS um different
 

00:28:44,799 --> 00:28:47,669
rivals glucose and MPS um different
co-actors that it was adding in. I think

00:28:47,669 --> 00:28:47,679
co-actors that it was adding in. I think
 

00:28:47,679 --> 00:28:50,310
co-actors that it was adding in. I think
we had some interesting high title

00:28:50,310 --> 00:28:50,320
we had some interesting high title
 

00:28:50,320 --> 00:28:52,389
we had some interesting high title
results with both spermadine addition

00:28:52,389 --> 00:28:52,399
results with both spermadine addition
 

00:28:52,399 --> 00:28:54,310
results with both spermadine addition
also catalace. So the model had this

00:28:54,310 --> 00:28:54,320
also catalace. So the model had this
 

00:28:54,320 --> 00:28:55,990
also catalace. So the model had this
theory that okay the catal is like good

00:28:55,990 --> 00:28:56,000
theory that okay the catal is like good
 

00:28:56,000 --> 00:28:59,029
theory that okay the catal is like good
for gen um consuming radical free

00:28:59,029 --> 00:28:59,039
for gen um consuming radical free
 

00:28:59,039 --> 00:29:00,549
for gen um consuming radical free
radicals but then also generating

00:29:00,549 --> 00:29:00,559
radicals but then also generating
 

00:29:00,559 --> 00:29:02,549
radicals but then also generating
additional oxygen at the liquid air

00:29:02,549 --> 00:29:02,559
additional oxygen at the liquid air
 

00:29:02,559 --> 00:29:04,149
additional oxygen at the liquid air
interface. unclear if that's a real

00:29:04,149 --> 00:29:04,159
interface. unclear if that's a real
 

00:29:04,159 --> 00:29:06,870
interface. unclear if that's a real
mechanism, but we did have a very good

00:29:06,870 --> 00:29:06,880
mechanism, but we did have a very good
 

00:29:06,880 --> 00:29:08,630
mechanism, but we did have a very good
um experimental condition come out of

00:29:08,630 --> 00:29:08,640
um experimental condition come out of
 

00:29:08,640 --> 00:29:13,029
um experimental condition come out of
this one. So yeah, um overall what we

00:29:13,029 --> 00:29:13,039
this one. So yeah, um overall what we
 

00:29:13,039 --> 00:29:15,510
this one. So yeah, um overall what we
learned was that I think by using the

00:29:15,510 --> 00:29:15,520
learned was that I think by using the
 

00:29:15,520 --> 00:29:17,590
learned was that I think by using the
pyic model to constrain the action space

00:29:17,590 --> 00:29:17,600
pyic model to constrain the action space
 

00:29:17,600 --> 00:29:19,590
pyic model to constrain the action space
of GPD5, it actually made very few

00:29:19,590 --> 00:29:19,600
of GPD5, it actually made very few
 

00:29:19,600 --> 00:29:22,230
of GPD5, it actually made very few
mistakes. So from the 36,000 different

00:29:22,230 --> 00:29:22,240
mistakes. So from the 36,000 different
 

00:29:22,240 --> 00:29:23,909
mistakes. So from the 36,000 different
experiments we did, it there were only

00:29:23,909 --> 00:29:23,919
experiments we did, it there were only
 

00:29:23,919 --> 00:29:25,350
experiments we did, it there were only
two bad plays. There was a negative

00:29:25,350 --> 00:29:25,360
two bad plays. There was a negative
 

00:29:25,360 --> 00:29:27,269
two bad plays. There was a negative
volume one that I talked about. Um, and

00:29:27,269 --> 00:29:27,279
volume one that I talked about. Um, and
 

00:29:27,279 --> 00:29:28,950
volume one that I talked about. Um, and
one that was kind of silly was a simple

00:29:28,950 --> 00:29:28,960
one that was kind of silly was a simple
 

00:29:28,960 --> 00:29:30,789
one that was kind of silly was a simple
math error where it was kind of

00:29:30,789 --> 00:29:30,799
math error where it was kind of
 

00:29:30,799 --> 00:29:32,710
math error where it was kind of
converting nanol liters to microl liters

00:29:32,710 --> 00:29:32,720
converting nanol liters to microl liters
 

00:29:32,720 --> 00:29:35,669
converting nanol liters to microl liters
incorrectly and so it basically

00:29:35,669 --> 00:29:35,679
incorrectly and so it basically
 

00:29:35,679 --> 00:29:38,230
incorrectly and so it basically
accidentally set all the other reagents

00:29:38,230 --> 00:29:38,240
accidentally set all the other reagents
 

00:29:38,240 --> 00:29:40,230
accidentally set all the other reagents
to zero and so we had a play that was

00:29:40,230 --> 00:29:40,240
to zero and so we had a play that was
 

00:29:40,240 --> 00:29:43,669
to zero and so we had a play that was
just like a sugar gradient. Um, and then

00:29:43,669 --> 00:29:43,679
just like a sugar gradient. Um, and then
 

00:29:43,679 --> 00:29:45,430
just like a sugar gradient. Um, and then
human oversight was still required for

00:29:45,430 --> 00:29:45,440
human oversight was still required for
 

00:29:45,440 --> 00:29:47,110
human oversight was still required for
kind of improving the protocols and

00:29:47,110 --> 00:29:47,120
kind of improving the protocols and
 

00:29:47,120 --> 00:29:49,510
kind of improving the protocols and
handling the reagents on off the rack

00:29:49,510 --> 00:29:49,520
handling the reagents on off the rack
 

00:29:49,520 --> 00:29:51,909
handling the reagents on off the rack
system. So we saw that system can design

00:29:51,909 --> 00:29:51,919
system. So we saw that system can design
 

00:29:51,919 --> 00:29:54,389
system. So we saw that system can design
and interpret experiments but we see

00:29:54,389 --> 00:29:54,399
and interpret experiments but we see
 

00:29:54,399 --> 00:29:55,909
and interpret experiments but we see
that lab work still requires kind of

00:29:55,909 --> 00:29:55,919
that lab work still requires kind of
 

00:29:55,919 --> 00:29:57,750
that lab work still requires kind of
these practical details and we do need

00:29:57,750 --> 00:29:57,760
these practical details and we do need
 

00:29:57,760 --> 00:29:59,830
these practical details and we do need
experienced operators and kind of

00:29:59,830 --> 00:29:59,840
experienced operators and kind of
 

00:29:59,840 --> 00:30:01,110
experienced operators and kind of
looking to the future we're really

00:30:01,110 --> 00:30:01,120
looking to the future we're really
 

00:30:01,120 --> 00:30:03,269
looking to the future we're really
excited to apply AI and more of these

00:30:03,269 --> 00:30:03,279
excited to apply AI and more of these
 

00:30:03,279 --> 00:30:05,990
excited to apply AI and more of these
lab in the loop optimization workflows.

00:30:05,990 --> 00:30:06,000
lab in the loop optimization workflows.
 

00:30:06,000 --> 00:30:07,990
lab in the loop optimization workflows.
Um and we really see kind of autonomous

00:30:07,990 --> 00:30:08,000
Um and we really see kind of autonomous
 

00:30:08,000 --> 00:30:10,149
Um and we really see kind of autonomous
labs as being complimentary to AI models

00:30:10,149 --> 00:30:10,159
labs as being complimentary to AI models
 

00:30:10,159 --> 00:30:12,070
labs as being complimentary to AI models
where the models can design the

00:30:12,070 --> 00:30:12,080
where the models can design the
 

00:30:12,080 --> 00:30:14,630
where the models can design the
experiments. Um but ultimately biology

00:30:14,630 --> 00:30:14,640
experiments. Um but ultimately biology
 

00:30:14,640 --> 00:30:16,630
experiments. Um but ultimately biology
really still requires testing iteration

00:30:16,630 --> 00:30:16,640
really still requires testing iteration
 

00:30:16,640 --> 00:30:17,990
really still requires testing iteration
kind of validation in the physical

00:30:17,990 --> 00:30:18,000
kind of validation in the physical
 

00:30:18,000 --> 00:30:19,510
kind of validation in the physical
world.

00:30:19,510 --> 00:30:19,520
world.
 

00:30:19,520 --> 00:30:22,549
world.
So yeah, that is all from me. Um, I

00:30:22,549 --> 00:30:22,559
So yeah, that is all from me. Um, I
 

00:30:22,559 --> 00:30:24,149
So yeah, that is all from me. Um, I
guess next up is Todd or are we taking

00:30:24,149 --> 00:30:24,159
guess next up is Todd or are we taking
 

00:30:24,159 --> 00:30:24,710
guess next up is Todd or are we taking
questions?

00:30:24,710 --> 00:30:24,720
questions?
 

00:30:24,720 --> 00:30:26,389
questions?
>> Oh, no, no, that's great. So I think uh

00:30:26,389 --> 00:30:26,399
>> Oh, no, no, that's great. So I think uh
 

00:30:26,399 --> 00:30:27,510
>> Oh, no, no, that's great. So I think uh
Todd will speak and then we'll do

00:30:27,510 --> 00:30:27,520
Todd will speak and then we'll do
 

00:30:27,520 --> 00:30:28,470
Todd will speak and then we'll do
questions at the end.

00:30:28,470 --> 00:30:28,480
questions at the end.
 

00:30:28,480 --> 00:30:31,190
questions at the end.
>> Okay.

00:30:31,190 --> 00:30:31,200
>> Okay.
 

00:30:31,200 --> 00:30:34,310
>> Okay.
Okay.

00:30:34,310 --> 00:30:34,320

 

00:30:34,320 --> 00:30:37,510

>> All right. Hello. I'm Todd Edwards.

00:30:37,510 --> 00:30:37,520
>> All right. Hello. I'm Todd Edwards.
 

00:30:37,520 --> 00:30:40,389
>> All right. Hello. I'm Todd Edwards.
Here we are. Uh, so I am with Emil,

00:30:40,389 --> 00:30:40,399
Here we are. Uh, so I am with Emil,
 

00:30:40,399 --> 00:30:41,750
Here we are. Uh, so I am with Emil,
which is the Environmental Molecular

00:30:41,750 --> 00:30:41,760
which is the Environmental Molecular
 

00:30:41,760 --> 00:30:44,149
which is the Environmental Molecular
Sciences Laboratory over at the Pacific

00:30:44,149 --> 00:30:44,159
Sciences Laboratory over at the Pacific
 

00:30:44,159 --> 00:30:46,789
Sciences Laboratory over at the Pacific
Northwest National Lab. Uh what I'm

00:30:46,789 --> 00:30:46,799
Northwest National Lab. Uh what I'm
 

00:30:46,799 --> 00:30:49,110
Northwest National Lab. Uh what I'm
going to talk to you about today is what

00:30:49,110 --> 00:30:49,120
going to talk to you about today is what
 

00:30:49,120 --> 00:30:51,110
going to talk to you about today is what
we are and kind of what we're trying to

00:30:51,110 --> 00:30:51,120
we are and kind of what we're trying to
 

00:30:51,120 --> 00:30:52,789
we are and kind of what we're trying to
achieve and the steps we're taking along

00:30:52,789 --> 00:30:52,799
achieve and the steps we're taking along
 

00:30:52,799 --> 00:30:56,470
achieve and the steps we're taking along
the way to get there. So EMS is a user

00:30:56,470 --> 00:30:56,480
the way to get there. So EMS is a user
 

00:30:56,480 --> 00:30:59,430
the way to get there. So EMS is a user
facility which is a a type of lab within

00:30:59,430 --> 00:30:59,440
facility which is a a type of lab within
 

00:30:59,440 --> 00:31:00,789
facility which is a a type of lab within
the department of energy office of

00:31:00,789 --> 00:31:00,799
the department of energy office of
 

00:31:00,799 --> 00:31:03,029
the department of energy office of
science where we don't do our own

00:31:03,029 --> 00:31:03,039
science where we don't do our own
 

00:31:03,039 --> 00:31:04,950
science where we don't do our own
research but we have scientists and

00:31:04,950 --> 00:31:04,960
research but we have scientists and
 

00:31:04,960 --> 00:31:08,389
research but we have scientists and
equipment and we partner with folks

00:31:08,389 --> 00:31:08,399
equipment and we partner with folks
 

00:31:08,399 --> 00:31:11,190
equipment and we partner with folks
around the country to take their samples

00:31:11,190 --> 00:31:11,200
around the country to take their samples
 

00:31:11,200 --> 00:31:13,990
around the country to take their samples
do whatever kind of analysis they need.

00:31:13,990 --> 00:31:14,000
do whatever kind of analysis they need.
 

00:31:14,000 --> 00:31:16,389
do whatever kind of analysis they need.
Then they get the data for a year and

00:31:16,389 --> 00:31:16,399
Then they get the data for a year and
 

00:31:16,399 --> 00:31:18,789
Then they get the data for a year and
then the data becomes public. So we're

00:31:18,789 --> 00:31:18,799
then the data becomes public. So we're
 

00:31:18,799 --> 00:31:20,950
then the data becomes public. So we're
able to build a large public repository

00:31:20,950 --> 00:31:20,960
able to build a large public repository
 

00:31:20,960 --> 00:31:24,310
able to build a large public repository
of interesting data that um hopefully is

00:31:24,310 --> 00:31:24,320
of interesting data that um hopefully is
 

00:31:24,320 --> 00:31:26,630
of interesting data that um hopefully is
all done consistently so that it can

00:31:26,630 --> 00:31:26,640
all done consistently so that it can
 

00:31:26,640 --> 00:31:29,510
all done consistently so that it can
contribute to larger projects. Uh you

00:31:29,510 --> 00:31:29,520
contribute to larger projects. Uh you
 

00:31:29,520 --> 00:31:30,950
contribute to larger projects. Uh you
can see we have functional systems

00:31:30,950 --> 00:31:30,960
can see we have functional systems
 

00:31:30,960 --> 00:31:33,269
can see we have functional systems
biology. This is traditional studying

00:31:33,269 --> 00:31:33,279
biology. This is traditional studying
 

00:31:33,279 --> 00:31:35,750
biology. This is traditional studying
microbes in culture. we have

00:31:35,750 --> 00:31:35,760
microbes in culture. we have
 

00:31:35,760 --> 00:31:37,190
microbes in culture. we have
environmental transformation and

00:31:37,190 --> 00:31:37,200
environmental transformation and
 

00:31:37,200 --> 00:31:39,909
environmental transformation and
interactions and this is looking at the

00:31:39,909 --> 00:31:39,919
interactions and this is looking at the
 

00:31:39,919 --> 00:31:41,990
interactions and this is looking at the
interactions between microbes and soil

00:31:41,990 --> 00:31:42,000
interactions between microbes and soil
 

00:31:42,000 --> 00:31:44,870
interactions between microbes and soil
or atmosphere plant roots. Uh so looking

00:31:44,870 --> 00:31:44,880
or atmosphere plant roots. Uh so looking
 

00:31:44,880 --> 00:31:46,389
or atmosphere plant roots. Uh so looking
at it in the context and that's for

00:31:46,389 --> 00:31:46,399
at it in the context and that's for
 

00:31:46,399 --> 00:31:49,190
at it in the context and that's for
building earth system models and we have

00:31:49,190 --> 00:31:49,200
building earth system models and we have
 

00:31:49,200 --> 00:31:51,430
building earth system models and we have
a computing group that helps build

00:31:51,430 --> 00:31:51,440
a computing group that helps build
 

00:31:51,440 --> 00:31:54,870
a computing group that helps build
models and use models in our research.

00:31:54,870 --> 00:31:54,880
models and use models in our research.
 

00:31:54,880 --> 00:31:57,590
models and use models in our research.
And so there's a a QR code here so you

00:31:57,590 --> 00:31:57,600
And so there's a a QR code here so you
 

00:31:57,600 --> 00:31:59,509
And so there's a a QR code here so you
can check us out, find out how you could

00:31:59,509 --> 00:31:59,519
can check us out, find out how you could
 

00:31:59,519 --> 00:32:01,990
can check us out, find out how you could
collaborate with us. Uh and then science

00:32:01,990 --> 00:32:02,000
collaborate with us. Uh and then science
 

00:32:02,000 --> 00:32:03,350
collaborate with us. Uh and then science
central is up at the top. That's where

00:32:03,350 --> 00:32:03,360
central is up at the top. That's where
 

00:32:03,360 --> 00:32:06,950
central is up at the top. That's where
you can get access to the data.

00:32:06,950 --> 00:32:06,960
you can get access to the data.
 

00:32:06,960 --> 00:32:09,990
you can get access to the data.
Now, because we have been very

00:32:09,990 --> 00:32:10,000
Now, because we have been very
 

00:32:10,000 --> 00:32:13,029
Now, because we have been very
traditionally working individually with

00:32:13,029 --> 00:32:13,039
traditionally working individually with
 

00:32:13,039 --> 00:32:15,750
traditionally working individually with
PIs, very small scale specific

00:32:15,750 --> 00:32:15,760
PIs, very small scale specific
 

00:32:15,760 --> 00:32:18,070
PIs, very small scale specific
experiments, but we have a vision where

00:32:18,070 --> 00:32:18,080
experiments, but we have a vision where
 

00:32:18,080 --> 00:32:20,630
experiments, but we have a vision where
we want to automate in order to improve

00:32:20,630 --> 00:32:20,640
we want to automate in order to improve
 

00:32:20,640 --> 00:32:23,669
we want to automate in order to improve
the rate at which we can generate data

00:32:23,669 --> 00:32:23,679
the rate at which we can generate data
 

00:32:23,679 --> 00:32:25,269
the rate at which we can generate data
and improve the consistency that we're

00:32:25,269 --> 00:32:25,279
and improve the consistency that we're
 

00:32:25,279 --> 00:32:28,310
and improve the consistency that we're
generating data. And but we're not doing

00:32:28,310 --> 00:32:28,320
generating data. And but we're not doing
 

00:32:28,320 --> 00:32:29,669
generating data. And but we're not doing
production, right? We're not doing the

00:32:29,669 --> 00:32:29,679
production, right? We're not doing the
 

00:32:29,679 --> 00:32:31,190
production, right? We're not doing the
traditional high throughput doing the

00:32:31,190 --> 00:32:31,200
traditional high throughput doing the
 

00:32:31,200 --> 00:32:34,549
traditional high throughput doing the
same few uh processes over and over. We

00:32:34,549 --> 00:32:34,559
same few uh processes over and over. We
 

00:32:34,559 --> 00:32:36,389
same few uh processes over and over. We
have to really tailor every time we work

00:32:36,389 --> 00:32:36,399
have to really tailor every time we work
 

00:32:36,399 --> 00:32:38,710
have to really tailor every time we work
with a new collaborator. Tailor the

00:32:38,710 --> 00:32:38,720
with a new collaborator. Tailor the
 

00:32:38,720 --> 00:32:41,430
with a new collaborator. Tailor the
experiments to what they need. Uh but we

00:32:41,430 --> 00:32:41,440
experiments to what they need. Uh but we
 

00:32:41,440 --> 00:32:42,870
experiments to what they need. Uh but we
still want to generate that high quality

00:32:42,870 --> 00:32:42,880
still want to generate that high quality
 

00:32:42,880 --> 00:32:45,750
still want to generate that high quality
consistent data. So we need modularity

00:32:45,750 --> 00:32:45,760
consistent data. So we need modularity
 

00:32:45,760 --> 00:32:47,830
consistent data. So we need modularity
in our platforms both in the hardware

00:32:47,830 --> 00:32:47,840
in our platforms both in the hardware
 

00:32:47,840 --> 00:32:50,950
in our platforms both in the hardware
and the software side. And then you know

00:32:50,950 --> 00:32:50,960
and the software side. And then you know
 

00:32:50,960 --> 00:32:53,029
and the software side. And then you know
as as things have grown autonomous

00:32:53,029 --> 00:32:53,039
as as things have grown autonomous
 

00:32:53,039 --> 00:32:55,590
as as things have grown autonomous
experimentation has cropped up and

00:32:55,590 --> 00:32:55,600
experimentation has cropped up and
 

00:32:55,600 --> 00:32:57,909
experimentation has cropped up and
proven to be a very efficient way to

00:32:57,909 --> 00:32:57,919
proven to be a very efficient way to
 

00:32:57,919 --> 00:33:00,470
proven to be a very efficient way to
build out the data that we need to make

00:33:00,470 --> 00:33:00,480
build out the data that we need to make
 

00:33:00,480 --> 00:33:03,110
build out the data that we need to make
the models or improve models over time.

00:33:03,110 --> 00:33:03,120
the models or improve models over time.
 

00:33:03,120 --> 00:33:05,669
the models or improve models over time.
Uh and again we have access to limbs

00:33:05,669 --> 00:33:05,679
Uh and again we have access to limbs
 

00:33:05,679 --> 00:33:07,750
Uh and again we have access to limbs
high performance computing data storage

00:33:07,750 --> 00:33:07,760
high performance computing data storage
 

00:33:07,760 --> 00:33:09,750
high performance computing data storage
so that we can generate the data analyze

00:33:09,750 --> 00:33:09,760
so that we can generate the data analyze
 

00:33:09,760 --> 00:33:13,029
so that we can generate the data analyze
it and make it open.

00:33:13,029 --> 00:33:13,039
it and make it open.
 

00:33:13,039 --> 00:33:15,350
it and make it open.
uh the challenge that we're specifically

00:33:15,350 --> 00:33:15,360
uh the challenge that we're specifically
 

00:33:15,360 --> 00:33:18,389
uh the challenge that we're specifically
trying to attack here at Emil is the the

00:33:18,389 --> 00:33:18,399
trying to attack here at Emil is the the
 

00:33:18,399 --> 00:33:21,110
trying to attack here at Emil is the the
notion of the genomes is is not enough

00:33:21,110 --> 00:33:21,120
notion of the genomes is is not enough
 

00:33:21,120 --> 00:33:23,350
notion of the genomes is is not enough
like if you understand the sequence of a

00:33:23,350 --> 00:33:23,360
like if you understand the sequence of a
 

00:33:23,360 --> 00:33:25,029
like if you understand the sequence of a
microbe that doesn't really tell you

00:33:25,029 --> 00:33:25,039
microbe that doesn't really tell you
 

00:33:25,039 --> 00:33:27,029
microbe that doesn't really tell you
enough to be able to predict how it's

00:33:27,029 --> 00:33:27,039
enough to be able to predict how it's
 

00:33:27,039 --> 00:33:28,950
enough to be able to predict how it's
going to behave. you need to study the

00:33:28,950 --> 00:33:28,960
going to behave. you need to study the
 

00:33:28,960 --> 00:33:30,950
going to behave. you need to study the
the phenotypes and so you need to study

00:33:30,950 --> 00:33:30,960
the phenotypes and so you need to study
 

00:33:30,960 --> 00:33:33,750
the phenotypes and so you need to study
the microbes under different conditions

00:33:33,750 --> 00:33:33,760
the microbes under different conditions
 

00:33:33,760 --> 00:33:36,710
the microbes under different conditions
and um in order to build models so that

00:33:36,710 --> 00:33:36,720
and um in order to build models so that
 

00:33:36,720 --> 00:33:38,789
and um in order to build models so that
you can predict this microbe this

00:33:38,789 --> 00:33:38,799
you can predict this microbe this
 

00:33:38,799 --> 00:33:40,950
you can predict this microbe this
genetic sequence under these conditions

00:33:40,950 --> 00:33:40,960
genetic sequence under these conditions
 

00:33:40,960 --> 00:33:43,430
genetic sequence under these conditions
how will it behave and that's relevant

00:33:43,430 --> 00:33:43,440
how will it behave and that's relevant
 

00:33:43,440 --> 00:33:46,870
how will it behave and that's relevant
for industry um this sin 3.0 know this

00:33:46,870 --> 00:33:46,880
for industry um this sin 3.0 know this
 

00:33:46,880 --> 00:33:48,549
for industry um this sin 3.0 know this
is an example we like to share. Uh

00:33:48,549 --> 00:33:48,559
is an example we like to share. Uh
 

00:33:48,559 --> 00:33:51,269
is an example we like to share. Uh
someone has created a microbe with a

00:33:51,269 --> 00:33:51,279
someone has created a microbe with a
 

00:33:51,279 --> 00:33:54,789
someone has created a microbe with a
minimal amount of genes and but still be

00:33:54,789 --> 00:33:54,799
minimal amount of genes and but still be
 

00:33:54,799 --> 00:33:57,430
minimal amount of genes and but still be
viable to grow and 30% of those genes

00:33:57,430 --> 00:33:57,440
viable to grow and 30% of those genes
 

00:33:57,440 --> 00:33:59,750
viable to grow and 30% of those genes
they don't know what they do. So it's a

00:33:59,750 --> 00:33:59,760
they don't know what they do. So it's a
 

00:33:59,760 --> 00:34:04,389
they don't know what they do. So it's a
very large problem and as in terms of

00:34:04,389 --> 00:34:04,399
very large problem and as in terms of
 

00:34:04,399 --> 00:34:06,630
very large problem and as in terms of
understanding what possible proteins

00:34:06,630 --> 00:34:06,640
understanding what possible proteins
 

00:34:06,640 --> 00:34:08,869
understanding what possible proteins
might be expressed and what proteins are

00:34:08,869 --> 00:34:08,879
might be expressed and what proteins are
 

00:34:08,879 --> 00:34:11,510
might be expressed and what proteins are
actually expressed and how they interact

00:34:11,510 --> 00:34:11,520
actually expressed and how they interact
 

00:34:11,520 --> 00:34:13,589
actually expressed and how they interact
and uh you know the problem is just

00:34:13,589 --> 00:34:13,599
and uh you know the problem is just
 

00:34:13,599 --> 00:34:15,109
and uh you know the problem is just
growing that the disparity between the

00:34:15,109 --> 00:34:15,119
growing that the disparity between the
 

00:34:15,119 --> 00:34:16,710
growing that the disparity between the
genomic sequences and the knowledge

00:34:16,710 --> 00:34:16,720
genomic sequences and the knowledge
 

00:34:16,720 --> 00:34:18,950
genomic sequences and the knowledge
about the genome is getting bigger and

00:34:18,950 --> 00:34:18,960
about the genome is getting bigger and
 

00:34:18,960 --> 00:34:21,270
about the genome is getting bigger and
bigger because uh more and more microbes

00:34:21,270 --> 00:34:21,280
bigger because uh more and more microbes
 

00:34:21,280 --> 00:34:23,270
bigger because uh more and more microbes
are being discovered and more sequences

00:34:23,270 --> 00:34:23,280
are being discovered and more sequences
 

00:34:23,280 --> 00:34:26,069
are being discovered and more sequences
are being produced. So our efforts have

00:34:26,069 --> 00:34:26,079
are being produced. So our efforts have
 

00:34:26,079 --> 00:34:28,149
are being produced. So our efforts have
been really to to try to catch up with

00:34:28,149 --> 00:34:28,159
been really to to try to catch up with
 

00:34:28,159 --> 00:34:30,389
been really to to try to catch up with
the ability to generate genomic data. We

00:34:30,389 --> 00:34:30,399
the ability to generate genomic data. We
 

00:34:30,399 --> 00:34:31,829
the ability to generate genomic data. We
want to be able to generate phenomic

00:34:31,829 --> 00:34:31,839
want to be able to generate phenomic
 

00:34:31,839 --> 00:34:34,310
want to be able to generate phenomic
data.

00:34:34,310 --> 00:34:34,320
data.
 

00:34:34,320 --> 00:34:36,389
data.
Now Emil like I said on the right hand

00:34:36,389 --> 00:34:36,399
Now Emil like I said on the right hand
 

00:34:36,399 --> 00:34:37,909
Now Emil like I said on the right hand
side is our traditional approach and

00:34:37,909 --> 00:34:37,919
side is our traditional approach and
 

00:34:37,919 --> 00:34:39,669
side is our traditional approach and
this is where we work with an individual

00:34:39,669 --> 00:34:39,679
this is where we work with an individual
 

00:34:39,679 --> 00:34:42,230
this is where we work with an individual
PI. We do a very targeted experiments

00:34:42,230 --> 00:34:42,240
PI. We do a very targeted experiments
 

00:34:42,240 --> 00:34:44,389
PI. We do a very targeted experiments
for them. Uh it's you know lower

00:34:44,389 --> 00:34:44,399
for them. Uh it's you know lower
 

00:34:44,399 --> 00:34:46,149
for them. Uh it's you know lower
throughput but it's really getting them

00:34:46,149 --> 00:34:46,159
throughput but it's really getting them
 

00:34:46,159 --> 00:34:47,750
throughput but it's really getting them
what they need. But because we're

00:34:47,750 --> 00:34:47,760
what they need. But because we're
 

00:34:47,760 --> 00:34:50,790
what they need. But because we're
generating it we're able to pull all the

00:34:50,790 --> 00:34:50,800
generating it we're able to pull all the
 

00:34:50,800 --> 00:34:54,069
generating it we're able to pull all the
data so it can be used for other folks.

00:34:54,069 --> 00:34:54,079
data so it can be used for other folks.
 

00:34:54,079 --> 00:34:56,149
data so it can be used for other folks.
What we want to do is do that on a much

00:34:56,149 --> 00:34:56,159
What we want to do is do that on a much
 

00:34:56,159 --> 00:34:58,790
What we want to do is do that on a much
larger scale uh with this automation

00:34:58,790 --> 00:34:58,800
larger scale uh with this automation
 

00:34:58,800 --> 00:35:01,030
larger scale uh with this automation
centric approach but we we don't want to

00:35:01,030 --> 00:35:01,040
centric approach but we we don't want to
 

00:35:01,040 --> 00:35:04,230
centric approach but we we don't want to
go directly from uh this traditional

00:35:04,230 --> 00:35:04,240
go directly from uh this traditional
 

00:35:04,240 --> 00:35:07,589
go directly from uh this traditional
approach to the the high throughput

00:35:07,589 --> 00:35:07,599
approach to the the high throughput
 

00:35:07,599 --> 00:35:09,670
approach to the the high throughput
you know forcing folks to do a narrow

00:35:09,670 --> 00:35:09,680
you know forcing folks to do a narrow
 

00:35:09,680 --> 00:35:12,150
you know forcing folks to do a narrow
set of experimental procedures

00:35:12,150 --> 00:35:12,160
set of experimental procedures
 

00:35:12,160 --> 00:35:14,150
set of experimental procedures
uh we want to jump from traditional

00:35:14,150 --> 00:35:14,160
uh we want to jump from traditional
 

00:35:14,160 --> 00:35:18,150
uh we want to jump from traditional
straight to a large scale R&D

00:35:18,150 --> 00:35:18,160
straight to a large scale R&D
 

00:35:18,160 --> 00:35:22,310
straight to a large scale R&D
uh let's see uh so this is our dream we

00:35:22,310 --> 00:35:22,320
uh let's see uh so this is our dream we
 

00:35:22,320 --> 00:35:25,109
uh let's see uh so this is our dream we
want to produce A we wanted we wanted to

00:35:25,109 --> 00:35:25,119
want to produce A we wanted we wanted to
 

00:35:25,119 --> 00:35:28,150
want to produce A we wanted we wanted to
commission a device that would be able

00:35:28,150 --> 00:35:28,160
commission a device that would be able
 

00:35:28,160 --> 00:35:31,349
commission a device that would be able
to do a very wide range of possible

00:35:31,349 --> 00:35:31,359
to do a very wide range of possible
 

00:35:31,359 --> 00:35:32,950
to do a very wide range of possible
experiments

00:35:32,950 --> 00:35:32,960
experiments
 

00:35:32,960 --> 00:35:34,790
experiments
uh at high throughput and with that

00:35:34,790 --> 00:35:34,800
uh at high throughput and with that
 

00:35:34,800 --> 00:35:37,670
uh at high throughput and with that
consistency and treat it like a

00:35:37,670 --> 00:35:37,680
consistency and treat it like a
 

00:35:37,680 --> 00:35:39,190
consistency and treat it like a
something like a James Webb telescope,

00:35:39,190 --> 00:35:39,200
something like a James Webb telescope,
 

00:35:39,200 --> 00:35:42,390
something like a James Webb telescope,
right? We want to make it able to do a

00:35:42,390 --> 00:35:42,400
right? We want to make it able to do a
 

00:35:42,400 --> 00:35:44,870
right? We want to make it able to do a
campaign of science, study a particular

00:35:44,870 --> 00:35:44,880
campaign of science, study a particular
 

00:35:44,880 --> 00:35:47,510
campaign of science, study a particular
thing and then once that campaign is

00:35:47,510 --> 00:35:47,520
thing and then once that campaign is
 

00:35:47,520 --> 00:35:49,430
thing and then once that campaign is
over and those folks have their data,

00:35:49,430 --> 00:35:49,440
over and those folks have their data,
 

00:35:49,440 --> 00:35:51,589
over and those folks have their data,
shift and pointed at something else and

00:35:51,589 --> 00:35:51,599
shift and pointed at something else and
 

00:35:51,599 --> 00:35:54,390
shift and pointed at something else and
study a whole new problem. So when we

00:35:54,390 --> 00:35:54,400
study a whole new problem. So when we
 

00:35:54,400 --> 00:35:56,069
study a whole new problem. So when we
when we set about this, we couldn't

00:35:56,069 --> 00:35:56,079
when we set about this, we couldn't
 

00:35:56,079 --> 00:35:59,430
when we set about this, we couldn't
really plan uh what what throughput we

00:35:59,430 --> 00:35:59,440
really plan uh what what throughput we
 

00:35:59,440 --> 00:36:01,430
really plan uh what what throughput we
needed or what experiments we needed. We

00:36:01,430 --> 00:36:01,440
needed or what experiments we needed. We
 

00:36:01,440 --> 00:36:04,470
needed or what experiments we needed. We
figured we focused more on the range of

00:36:04,470 --> 00:36:04,480
figured we focused more on the range of
 

00:36:04,480 --> 00:36:06,870
figured we focused more on the range of
capabilities that we wanted and kind of

00:36:06,870 --> 00:36:06,880
capabilities that we wanted and kind of
 

00:36:06,880 --> 00:36:08,550
capabilities that we wanted and kind of
a capacity that we wanted to be able to

00:36:08,550 --> 00:36:08,560
a capacity that we wanted to be able to
 

00:36:08,560 --> 00:36:11,589
a capacity that we wanted to be able to
do and then we went out and took bids

00:36:11,589 --> 00:36:11,599
do and then we went out and took bids
 

00:36:11,599 --> 00:36:13,910
do and then we went out and took bids
and came up with a solution which we we

00:36:13,910 --> 00:36:13,920
and came up with a solution which we we
 

00:36:13,920 --> 00:36:17,670
and came up with a solution which we we
are working with GKO uh to produce and

00:36:17,670 --> 00:36:17,680
are working with GKO uh to produce and
 

00:36:17,680 --> 00:36:21,030
are working with GKO uh to produce and
we call it a capability because M2PC is

00:36:21,030 --> 00:36:21,040
we call it a capability because M2PC is
 

00:36:21,040 --> 00:36:22,950
we call it a capability because M2PC is
both the equipment that's going to go in

00:36:22,950 --> 00:36:22,960
both the equipment that's going to go in
 

00:36:22,960 --> 00:36:24,790
both the equipment that's going to go in
the room but it's also this whole new

00:36:24,790 --> 00:36:24,800
the room but it's also this whole new
 

00:36:24,800 --> 00:36:27,190
the room but it's also this whole new
lab. So, we have a whole new building,

00:36:27,190 --> 00:36:27,200
lab. So, we have a whole new building,
 

00:36:27,200 --> 00:36:30,230
lab. So, we have a whole new building,
new support labs space, uh, and a big

00:36:30,230 --> 00:36:30,240
new support labs space, uh, and a big
 

00:36:30,240 --> 00:36:32,390
new support labs space, uh, and a big
automation space, plus this this rack

00:36:32,390 --> 00:36:32,400
automation space, plus this this rack
 

00:36:32,400 --> 00:36:34,950
automation space, plus this this rack
system. And you can see it's it may be

00:36:34,950 --> 00:36:34,960
system. And you can see it's it may be
 

00:36:34,960 --> 00:36:36,630
system. And you can see it's it may be
hard to see from the back, but we have

00:36:36,630 --> 00:36:36,640
hard to see from the back, but we have
 

00:36:36,640 --> 00:36:38,550
hard to see from the back, but we have
different pods, and each one is focused

00:36:38,550 --> 00:36:38,560
different pods, and each one is focused
 

00:36:38,560 --> 00:36:40,230
different pods, and each one is focused
on a different type of capability. So,

00:36:40,230 --> 00:36:40,240
on a different type of capability. So,
 

00:36:40,240 --> 00:36:42,390
on a different type of capability. So,
there's the ability to produce media, so

00:36:42,390 --> 00:36:42,400
there's the ability to produce media, so
 

00:36:42,400 --> 00:36:45,349
there's the ability to produce media, so
we can do a wide variety of plates. We

00:36:45,349 --> 00:36:45,359
we can do a wide variety of plates. We
 

00:36:45,359 --> 00:36:47,750
we can do a wide variety of plates. We
have a a cultivation section, we have

00:36:47,750 --> 00:36:47,760
have a a cultivation section, we have
 

00:36:47,760 --> 00:36:50,150
have a a cultivation section, we have
analytical tools, we have massspec, we

00:36:50,150 --> 00:36:50,160
analytical tools, we have massspec, we
 

00:36:50,160 --> 00:36:51,670
analytical tools, we have massspec, we
have sample prep, and then we have a

00:36:51,670 --> 00:36:51,680
have sample prep, and then we have a
 

00:36:51,680 --> 00:36:54,470
have sample prep, and then we have a
BSL2 section. And so this is all in the

00:36:54,470 --> 00:36:54,480
BSL2 section. And so this is all in the
 

00:36:54,480 --> 00:36:56,630
BSL2 section. And so this is all in the
design process right now. Still settling

00:36:56,630 --> 00:36:56,640
design process right now. Still settling
 

00:36:56,640 --> 00:36:59,030
design process right now. Still settling
on what we're going to actually get. But

00:36:59,030 --> 00:36:59,040
on what we're going to actually get. But
 

00:36:59,040 --> 00:37:01,109
on what we're going to actually get. But
the idea is that instead of now working

00:37:01,109 --> 00:37:01,119
the idea is that instead of now working
 

00:37:01,119 --> 00:37:03,990
the idea is that instead of now working
with individuals and doing experiments

00:37:03,990 --> 00:37:04,000
with individuals and doing experiments
 

00:37:04,000 --> 00:37:06,710
with individuals and doing experiments
on say a single massspec, we're going to

00:37:06,710 --> 00:37:06,720
on say a single massspec, we're going to
 

00:37:06,720 --> 00:37:08,150
on say a single massspec, we're going to
work with a campaign. So we'll get a

00:37:08,150 --> 00:37:08,160
work with a campaign. So we'll get a
 

00:37:08,160 --> 00:37:09,589
work with a campaign. So we'll get a
group of scientists together that are

00:37:09,589 --> 00:37:09,599
group of scientists together that are
 

00:37:09,599 --> 00:37:11,750
group of scientists together that are
all focused on the similar problem and

00:37:11,750 --> 00:37:11,760
all focused on the similar problem and
 

00:37:11,760 --> 00:37:13,910
all focused on the similar problem and
we'll plan out a series of experiments

00:37:13,910 --> 00:37:13,920
we'll plan out a series of experiments
 

00:37:13,920 --> 00:37:17,190
we'll plan out a series of experiments
and then run it on MTPC. And then while

00:37:17,190 --> 00:37:17,200
and then run it on MTPC. And then while
 

00:37:17,200 --> 00:37:18,950
and then run it on MTPC. And then while
that's running, we're planning the next

00:37:18,950 --> 00:37:18,960
that's running, we're planning the next
 

00:37:18,960 --> 00:37:20,710
that's running, we're planning the next
campaign. So when they get their data,

00:37:20,710 --> 00:37:20,720
campaign. So when they get their data,
 

00:37:20,720 --> 00:37:22,069
campaign. So when they get their data,
we can move on and start the second

00:37:22,069 --> 00:37:22,079
we can move on and start the second
 

00:37:22,079 --> 00:37:24,710
we can move on and start the second
campaign and so on and so on.

00:37:24,710 --> 00:37:24,720
campaign and so on and so on.
 

00:37:24,720 --> 00:37:26,390
campaign and so on and so on.
Uh but this is a you know this is the

00:37:26,390 --> 00:37:26,400
Uh but this is a you know this is the
 

00:37:26,400 --> 00:37:27,829
Uh but this is a you know this is the
goal. This is the dream. It's going to

00:37:27,829 --> 00:37:27,839
goal. This is the dream. It's going to
 

00:37:27,839 --> 00:37:30,710
goal. This is the dream. It's going to
take a few years. Uh so how do we get

00:37:30,710 --> 00:37:30,720
take a few years. Uh so how do we get
 

00:37:30,720 --> 00:37:32,870
take a few years. Uh so how do we get
there? So in the meantime, we've been

00:37:32,870 --> 00:37:32,880
there? So in the meantime, we've been
 

00:37:32,880 --> 00:37:34,390
there? So in the meantime, we've been
taking steps along the way to build out

00:37:34,390 --> 00:37:34,400
taking steps along the way to build out
 

00:37:34,400 --> 00:37:37,270
taking steps along the way to build out
all the infrastructure, all the uh

00:37:37,270 --> 00:37:37,280
all the infrastructure, all the uh
 

00:37:37,280 --> 00:37:39,109
all the infrastructure, all the uh
adjacent things that we need to

00:37:39,109 --> 00:37:39,119
adjacent things that we need to
 

00:37:39,119 --> 00:37:41,270
adjacent things that we need to
accomplish in order to get there. And

00:37:41,270 --> 00:37:41,280
accomplish in order to get there. And
 

00:37:41,280 --> 00:37:43,270
accomplish in order to get there. And
I'll walk you through that. Our first

00:37:43,270 --> 00:37:43,280
I'll walk you through that. Our first
 

00:37:43,280 --> 00:37:45,829
I'll walk you through that. Our first
one is automation. Like I said, we have

00:37:45,829 --> 00:37:45,839
one is automation. Like I said, we have
 

00:37:45,839 --> 00:37:47,910
one is automation. Like I said, we have
been a fairly traditional lab, not much

00:37:47,910 --> 00:37:47,920
been a fairly traditional lab, not much
 

00:37:47,920 --> 00:37:51,270
been a fairly traditional lab, not much
automation. and we wanted to kind of dip

00:37:51,270 --> 00:37:51,280
automation. and we wanted to kind of dip
 

00:37:51,280 --> 00:37:54,310
automation. and we wanted to kind of dip
our toes first. So we we did a similar

00:37:54,310 --> 00:37:54,320
our toes first. So we we did a similar
 

00:37:54,320 --> 00:37:57,270
our toes first. So we we did a similar
kind of commissioning process uh to

00:37:57,270 --> 00:37:57,280
kind of commissioning process uh to
 

00:37:57,280 --> 00:38:00,870
kind of commissioning process uh to
contract a a system to do what M2PC is

00:38:00,870 --> 00:38:00,880
contract a a system to do what M2PC is
 

00:38:00,880 --> 00:38:02,950
contract a a system to do what M2PC is
going to do but at a smaller scale and

00:38:02,950 --> 00:38:02,960
going to do but at a smaller scale and
 

00:38:02,960 --> 00:38:06,870
going to do but at a smaller scale and
in an anorobic environment. And so and

00:38:06,870 --> 00:38:06,880
in an anorobic environment. And so and
 

00:38:06,880 --> 00:38:09,030
in an anorobic environment. And so and
we also worked with Genko on this one

00:38:09,030 --> 00:38:09,040
we also worked with Genko on this one
 

00:38:09,040 --> 00:38:11,349
we also worked with Genko on this one
and instead of doing racks they built

00:38:11,349 --> 00:38:11,359
and instead of doing racks they built
 

00:38:11,359 --> 00:38:13,109
and instead of doing racks they built
what are called tiles which are very

00:38:13,109 --> 00:38:13,119
what are called tiles which are very
 

00:38:13,119 --> 00:38:15,270
what are called tiles which are very
similar to a rack but they're enclosed

00:38:15,270 --> 00:38:15,280
similar to a rack but they're enclosed
 

00:38:15,280 --> 00:38:18,150
similar to a rack but they're enclosed
in these anorobic pods. So we can do all

00:38:18,150 --> 00:38:18,160
in these anorobic pods. So we can do all
 

00:38:18,160 --> 00:38:20,230
in these anorobic pods. So we can do all
of the cultivation, the growth

00:38:20,230 --> 00:38:20,240
of the cultivation, the growth
 

00:38:20,240 --> 00:38:22,150
of the cultivation, the growth
experiments. We can do analysis with

00:38:22,150 --> 00:38:22,160
experiments. We can do analysis with
 

00:38:22,160 --> 00:38:25,270
experiments. We can do analysis with
HBLC and some other tools and produce

00:38:25,270 --> 00:38:25,280
HBLC and some other tools and produce
 

00:38:25,280 --> 00:38:27,670
HBLC and some other tools and produce
plates and do those kind of studies on

00:38:27,670 --> 00:38:27,680
plates and do those kind of studies on
 

00:38:27,680 --> 00:38:29,990
plates and do those kind of studies on
anorobic microbes while keeping them in

00:38:29,990 --> 00:38:30,000
anorobic microbes while keeping them in
 

00:38:30,000 --> 00:38:32,630
anorobic microbes while keeping them in
an anorobic conditions the whole way so

00:38:32,630 --> 00:38:32,640
an anorobic conditions the whole way so
 

00:38:32,640 --> 00:38:35,589
an anorobic conditions the whole way so
we don't have uh impacts of working in

00:38:35,589 --> 00:38:35,599
we don't have uh impacts of working in
 

00:38:35,599 --> 00:38:38,790
we don't have uh impacts of working in
oxygen or trying to get rid of oxygen.

00:38:38,790 --> 00:38:38,800
oxygen or trying to get rid of oxygen.
 

00:38:38,800 --> 00:38:41,030
oxygen or trying to get rid of oxygen.
Um, and this is also forcing us to

00:38:41,030 --> 00:38:41,040
Um, and this is also forcing us to
 

00:38:41,040 --> 00:38:44,230
Um, and this is also forcing us to
figure out how do we manage this kind of

00:38:44,230 --> 00:38:44,240
figure out how do we manage this kind of
 

00:38:44,240 --> 00:38:47,109
figure out how do we manage this kind of
level of throughput in the lab that

00:38:47,109 --> 00:38:47,119
level of throughput in the lab that
 

00:38:47,119 --> 00:38:49,030
level of throughput in the lab that
we're not used to. So, there's a lot of

00:38:49,030 --> 00:38:49,040
we're not used to. So, there's a lot of
 

00:38:49,040 --> 00:38:50,470
we're not used to. So, there's a lot of
consumables that have to go in and

00:38:50,470 --> 00:38:50,480
consumables that have to go in and
 

00:38:50,480 --> 00:38:51,990
consumables that have to go in and
reagents and there's a lot of waste

00:38:51,990 --> 00:38:52,000
reagents and there's a lot of waste
 

00:38:52,000 --> 00:38:54,390
reagents and there's a lot of waste
coming out and there's we need kind of a

00:38:54,390 --> 00:38:54,400
coming out and there's we need kind of a
 

00:38:54,400 --> 00:38:56,870
coming out and there's we need kind of a
new group of people to support it. Uh,

00:38:56,870 --> 00:38:56,880
new group of people to support it. Uh,
 

00:38:56,880 --> 00:38:59,109
new group of people to support it. Uh,
and so it's better to do this now

00:38:59,109 --> 00:38:59,119
and so it's better to do this now
 

00:38:59,119 --> 00:39:00,630
and so it's better to do this now
because we're going to need that when we

00:39:00,630 --> 00:39:00,640
because we're going to need that when we
 

00:39:00,640 --> 00:39:04,390
because we're going to need that when we
get to MTPC. We're gonna need even more.

00:39:04,390 --> 00:39:04,400
get to MTPC. We're gonna need even more.
 

00:39:04,400 --> 00:39:06,550
get to MTPC. We're gonna need even more.
We also have a a computing analytics and

00:39:06,550 --> 00:39:06,560
We also have a a computing analytics and
 

00:39:06,560 --> 00:39:08,069
We also have a a computing analytics and
modeling team, CAM. I mentioned them

00:39:08,069 --> 00:39:08,079
modeling team, CAM. I mentioned them
 

00:39:08,079 --> 00:39:10,150
modeling team, CAM. I mentioned them
before. They have a a section called

00:39:10,150 --> 00:39:10,160
before. They have a a section called
 

00:39:10,160 --> 00:39:12,390
before. They have a a section called
MIDAS, the modeling integration data

00:39:12,390 --> 00:39:12,400
MIDAS, the modeling integration data
 

00:39:12,400 --> 00:39:16,150
MIDAS, the modeling integration data
agents for science. And their goal is to

00:39:16,150 --> 00:39:16,160
agents for science. And their goal is to
 

00:39:16,160 --> 00:39:19,670
agents for science. And their goal is to
build AI agents that will help all along

00:39:19,670 --> 00:39:19,680
build AI agents that will help all along
 

00:39:19,680 --> 00:39:22,710
build AI agents that will help all along
the way. And this is this is kind of a

00:39:22,710 --> 00:39:22,720
the way. And this is this is kind of a
 

00:39:22,720 --> 00:39:25,030
the way. And this is this is kind of a
one typical experiment of how we how we

00:39:25,030 --> 00:39:25,040
one typical experiment of how we how we
 

00:39:25,040 --> 00:39:27,430
one typical experiment of how we how we
work. Users might come in and they go to

00:39:27,430 --> 00:39:27,440
work. Users might come in and they go to
 

00:39:27,440 --> 00:39:30,310
work. Users might come in and they go to
the JGI to get their sequencing. And

00:39:30,310 --> 00:39:30,320
the JGI to get their sequencing. And
 

00:39:30,320 --> 00:39:33,430
the JGI to get their sequencing. And
that sequence data can be used to create

00:39:33,430 --> 00:39:33,440
that sequence data can be used to create
 

00:39:33,440 --> 00:39:38,069
that sequence data can be used to create
models which are um not so useful when

00:39:38,069 --> 00:39:38,079
models which are um not so useful when
 

00:39:38,079 --> 00:39:39,430
models which are um not so useful when
they first come out because they're not

00:39:39,430 --> 00:39:39,440
they first come out because they're not
 

00:39:39,440 --> 00:39:41,430
they first come out because they're not
constrained. So they don't predict very

00:39:41,430 --> 00:39:41,440
constrained. So they don't predict very
 

00:39:41,440 --> 00:39:44,069
constrained. So they don't predict very
well. But we can do experiments doing

00:39:44,069 --> 00:39:44,079
well. But we can do experiments doing
 

00:39:44,079 --> 00:39:46,630
well. But we can do experiments doing
the phenomics to constrain those models

00:39:46,630 --> 00:39:46,640
the phenomics to constrain those models
 

00:39:46,640 --> 00:39:48,790
the phenomics to constrain those models
and improve them so they can do better

00:39:48,790 --> 00:39:48,800
and improve them so they can do better
 

00:39:48,800 --> 00:39:51,030
and improve them so they can do better
predictions. The goal eventually being

00:39:51,030 --> 00:39:51,040
predictions. The goal eventually being
 

00:39:51,040 --> 00:39:53,190
predictions. The goal eventually being
to build a digital phenome of the of the

00:39:53,190 --> 00:39:53,200
to build a digital phenome of the of the
 

00:39:53,200 --> 00:39:56,390
to build a digital phenome of the of the
microbe so that we can predict the the

00:39:56,390 --> 00:39:56,400
microbe so that we can predict the the
 

00:39:56,400 --> 00:39:58,790
microbe so that we can predict the the
outcome the functional parameters of the

00:39:58,790 --> 00:39:58,800
outcome the functional parameters of the
 

00:39:58,800 --> 00:40:00,790
outcome the functional parameters of the
microbes. And so we're building tools

00:40:00,790 --> 00:40:00,800
microbes. And so we're building tools
 

00:40:00,800 --> 00:40:04,470
microbes. And so we're building tools
now to help build those models, improve

00:40:04,470 --> 00:40:04,480
now to help build those models, improve
 

00:40:04,480 --> 00:40:07,190
now to help build those models, improve
those models with data. And then the

00:40:07,190 --> 00:40:07,200
those models with data. And then the
 

00:40:07,200 --> 00:40:09,190
those models with data. And then the
next step is to build models to build

00:40:09,190 --> 00:40:09,200
next step is to build models to build
 

00:40:09,200 --> 00:40:12,230
next step is to build models to build
agents to help design the experiments

00:40:12,230 --> 00:40:12,240
agents to help design the experiments
 

00:40:12,240 --> 00:40:14,790
agents to help design the experiments
that would be run on the systems so that

00:40:14,790 --> 00:40:14,800
that would be run on the systems so that
 

00:40:14,800 --> 00:40:16,550
that would be run on the systems so that
we can eventually get to that closed

00:40:16,550 --> 00:40:16,560
we can eventually get to that closed
 

00:40:16,560 --> 00:40:19,430
we can eventually get to that closed
loop autonomous experimentation where we

00:40:19,430 --> 00:40:19,440
loop autonomous experimentation where we
 

00:40:19,440 --> 00:40:24,390
loop autonomous experimentation where we
are using AI to efficiently plan

00:40:24,390 --> 00:40:24,400
are using AI to efficiently plan
 

00:40:24,400 --> 00:40:27,030
are using AI to efficiently plan
experiments to uh sample that whole

00:40:27,030 --> 00:40:27,040
experiments to uh sample that whole
 

00:40:27,040 --> 00:40:30,710
experiments to uh sample that whole
parameter space because the

00:40:30,710 --> 00:40:30,720
parameter space because the
 

00:40:30,720 --> 00:40:33,109
parameter space because the
u the growth conditions we saw in the

00:40:33,109 --> 00:40:33,119
u the growth conditions we saw in the
 

00:40:33,119 --> 00:40:34,310
u the growth conditions we saw in the
last talk there are many different

00:40:34,310 --> 00:40:34,320
last talk there are many different
 

00:40:34,320 --> 00:40:36,470
last talk there are many different
variables that you can adjust a person

00:40:36,470 --> 00:40:36,480
variables that you can adjust a person
 

00:40:36,480 --> 00:40:38,390
variables that you can adjust a person
trying to do that is not efficient

00:40:38,390 --> 00:40:38,400
trying to do that is not efficient
 

00:40:38,400 --> 00:40:40,470
trying to do that is not efficient
whereas AI can do that much more

00:40:40,470 --> 00:40:40,480
whereas AI can do that much more
 

00:40:40,480 --> 00:40:42,950
whereas AI can do that much more
efficiently. So we can get a model that

00:40:42,950 --> 00:40:42,960
efficiently. So we can get a model that
 

00:40:42,960 --> 00:40:45,510
efficiently. So we can get a model that
works.

00:40:45,510 --> 00:40:45,520
works.
 

00:40:45,520 --> 00:40:47,910
works.
Uh the next thing we're working on is to

00:40:47,910 --> 00:40:47,920
Uh the next thing we're working on is to
 

00:40:47,920 --> 00:40:49,910
Uh the next thing we're working on is to
standardize our workflows. So like I

00:40:49,910 --> 00:40:49,920
standardize our workflows. So like I
 

00:40:49,920 --> 00:40:52,870
standardize our workflows. So like I
said we we get all sorts of input. Uh

00:40:52,870 --> 00:40:52,880
said we we get all sorts of input. Uh
 

00:40:52,880 --> 00:40:54,069
said we we get all sorts of input. Uh
different people have different

00:40:54,069 --> 00:40:54,079
different people have different
 

00:40:54,079 --> 00:40:56,950
different people have different
workflows needs but from our standpoint

00:40:56,950 --> 00:40:56,960
workflows needs but from our standpoint
 

00:40:56,960 --> 00:40:58,630
workflows needs but from our standpoint
we need to standardize right? So we're

00:40:58,630 --> 00:40:58,640
we need to standardize right? So we're
 

00:40:58,640 --> 00:40:59,750
we need to standardize right? So we're
trying to come up with a way to

00:40:59,750 --> 00:40:59,760
trying to come up with a way to
 

00:40:59,760 --> 00:41:01,990
trying to come up with a way to
represent workflows, automated biology

00:41:01,990 --> 00:41:02,000
represent workflows, automated biology
 

00:41:02,000 --> 00:41:05,030
represent workflows, automated biology
workflows in a standard way so that we

00:41:05,030 --> 00:41:05,040
workflows in a standard way so that we
 

00:41:05,040 --> 00:41:07,829
workflows in a standard way so that we
have a definition of them so that we can

00:41:07,829 --> 00:41:07,839
have a definition of them so that we can
 

00:41:07,839 --> 00:41:10,150
have a definition of them so that we can
compare data that comes out of these

00:41:10,150 --> 00:41:10,160
compare data that comes out of these
 

00:41:10,160 --> 00:41:12,630
compare data that comes out of these
workflows apples to apples instead of

00:41:12,630 --> 00:41:12,640
workflows apples to apples instead of
 

00:41:12,640 --> 00:41:14,630
workflows apples to apples instead of
not being able to compare them. Right?

00:41:14,630 --> 00:41:14,640
not being able to compare them. Right?
 

00:41:14,640 --> 00:41:16,710
not being able to compare them. Right?
Because if your experiment is basically

00:41:16,710 --> 00:41:16,720
Because if your experiment is basically
 

00:41:16,720 --> 00:41:18,710
Because if your experiment is basically
the same as your experiment, but it's

00:41:18,710 --> 00:41:18,720
the same as your experiment, but it's
 

00:41:18,720 --> 00:41:19,990
the same as your experiment, but it's
different enough that you can't compare

00:41:19,990 --> 00:41:20,000
different enough that you can't compare
 

00:41:20,000 --> 00:41:23,510
different enough that you can't compare
your results unless you know exactly how

00:41:23,510 --> 00:41:23,520
your results unless you know exactly how
 

00:41:23,520 --> 00:41:26,470
your results unless you know exactly how
they are different. And as uh anyone has

00:41:26,470 --> 00:41:26,480
they are different. And as uh anyone has
 

00:41:26,480 --> 00:41:29,430
they are different. And as uh anyone has
tried to work with a biologist who has a

00:41:29,430 --> 00:41:29,440
tried to work with a biologist who has a
 

00:41:29,440 --> 00:41:31,829
tried to work with a biologist who has a
you know they ask for a cultivation and

00:41:31,829 --> 00:41:31,839
you know they ask for a cultivation and
 

00:41:31,839 --> 00:41:33,829
you know they ask for a cultivation and
then you have to start probing what are

00:41:33,829 --> 00:41:33,839
then you have to start probing what are
 

00:41:33,839 --> 00:41:35,990
then you have to start probing what are
all the details eventually you can sort

00:41:35,990 --> 00:41:36,000
all the details eventually you can sort
 

00:41:36,000 --> 00:41:38,710
all the details eventually you can sort
it out but there's a lot of work. So we

00:41:38,710 --> 00:41:38,720
it out but there's a lot of work. So we
 

00:41:38,720 --> 00:41:40,069
it out but there's a lot of work. So we
are building tools that will help us

00:41:40,069 --> 00:41:40,079
are building tools that will help us
 

00:41:40,079 --> 00:41:42,309
are building tools that will help us
onboard these workflows convert them

00:41:42,309 --> 00:41:42,319
onboard these workflows convert them
 

00:41:42,319 --> 00:41:45,109
onboard these workflows convert them
into our standard format and then you

00:41:45,109 --> 00:41:45,119
into our standard format and then you
 

00:41:45,119 --> 00:41:46,710
into our standard format and then you
convert that standard format into

00:41:46,710 --> 00:41:46,720
convert that standard format into
 

00:41:46,720 --> 00:41:50,550
convert that standard format into
instructions for our automation.

00:41:50,550 --> 00:41:50,560
instructions for our automation.
 

00:41:50,560 --> 00:41:51,829
instructions for our automation.
And then the last little bit, this is

00:41:51,829 --> 00:41:51,839
And then the last little bit, this is
 

00:41:51,839 --> 00:41:53,349
And then the last little bit, this is
where I've been more involved is is

00:41:53,349 --> 00:41:53,359
where I've been more involved is is
 

00:41:53,359 --> 00:41:54,950
where I've been more involved is is
trying to pull it all together. So we've

00:41:54,950 --> 00:41:54,960
trying to pull it all together. So we've
 

00:41:54,960 --> 00:41:56,630
trying to pull it all together. So we've
got the automation, we've got the

00:41:56,630 --> 00:41:56,640
got the automation, we've got the
 

00:41:56,640 --> 00:41:59,430
got the automation, we've got the
workflow, we got the the tools to design

00:41:59,430 --> 00:41:59,440
workflow, we got the the tools to design
 

00:41:59,440 --> 00:42:01,270
workflow, we got the the tools to design
experiments, but we hadn't put it all

00:42:01,270 --> 00:42:01,280
experiments, but we hadn't put it all
 

00:42:01,280 --> 00:42:04,309
experiments, but we hadn't put it all
together. And uh if you've done

00:42:04,309 --> 00:42:04,319
together. And uh if you've done
 

00:42:04,319 --> 00:42:05,910
together. And uh if you've done
interdisciplinary work, you know, it's

00:42:05,910 --> 00:42:05,920
interdisciplinary work, you know, it's
 

00:42:05,920 --> 00:42:09,430
interdisciplinary work, you know, it's
not easy. So we we decided to do a

00:42:09,430 --> 00:42:09,440
not easy. So we we decided to do a
 

00:42:09,440 --> 00:42:12,069
not easy. So we we decided to do a
minimum viable product to show this this

00:42:12,069 --> 00:42:12,079
minimum viable product to show this this
 

00:42:12,079 --> 00:42:14,230
minimum viable product to show this this
loop uh going all the way around and

00:42:14,230 --> 00:42:14,240
loop uh going all the way around and
 

00:42:14,240 --> 00:42:16,390
loop uh going all the way around and
then forcing the biologists and the

00:42:16,390 --> 00:42:16,400
then forcing the biologists and the
 

00:42:16,400 --> 00:42:18,230
then forcing the biologists and the
systems biologists and the automation

00:42:18,230 --> 00:42:18,240
systems biologists and the automation
 

00:42:18,240 --> 00:42:20,309
systems biologists and the automation
people and the coders, the modelers all

00:42:20,309 --> 00:42:20,319
people and the coders, the modelers all
 

00:42:20,319 --> 00:42:21,990
people and the coders, the modelers all
to work together and understand really

00:42:21,990 --> 00:42:22,000
to work together and understand really
 

00:42:22,000 --> 00:42:25,109
to work together and understand really
what the other teams need uh and sort

00:42:25,109 --> 00:42:25,119
what the other teams need uh and sort
 

00:42:25,119 --> 00:42:27,670
what the other teams need uh and sort
out those interfaces and we were able to

00:42:27,670 --> 00:42:27,680
out those interfaces and we were able to
 

00:42:27,680 --> 00:42:30,550
out those interfaces and we were able to
do that. So we have a an a tool that

00:42:30,550 --> 00:42:30,560
do that. So we have a an a tool that
 

00:42:30,560 --> 00:42:32,309
do that. So we have a an a tool that
orchestrates this all and it takes an

00:42:32,309 --> 00:42:32,319
orchestrates this all and it takes an
 

00:42:32,319 --> 00:42:35,270
orchestrates this all and it takes an
input file to uh that defines what the

00:42:35,270 --> 00:42:35,280
input file to uh that defines what the
 

00:42:35,280 --> 00:42:37,270
input file to uh that defines what the
experiment is. We have another tool that

00:42:37,270 --> 00:42:37,280
experiment is. We have another tool that
 

00:42:37,280 --> 00:42:39,990
experiment is. We have another tool that
designs the plates. Uh and then another

00:42:39,990 --> 00:42:40,000
designs the plates. Uh and then another
 

00:42:40,000 --> 00:42:43,030
designs the plates. Uh and then another
tool that turns that plate design into

00:42:43,030 --> 00:42:43,040
tool that turns that plate design into
 

00:42:43,040 --> 00:42:45,030
tool that turns that plate design into
pipe heading instructions. We used a TCA

00:42:45,030 --> 00:42:45,040
pipe heading instructions. We used a TCA
 

00:42:45,040 --> 00:42:47,109
pipe heading instructions. We used a TCA
fluent. Uh and we did a growth

00:42:47,109 --> 00:42:47,119
fluent. Uh and we did a growth
 

00:42:47,119 --> 00:42:48,790
fluent. Uh and we did a growth
experiment where we measured the OD

00:42:48,790 --> 00:42:48,800
experiment where we measured the OD
 

00:42:48,800 --> 00:42:51,750
experiment where we measured the OD
every hour and then took that raw data

00:42:51,750 --> 00:42:51,760
every hour and then took that raw data
 

00:42:51,760 --> 00:42:55,270
every hour and then took that raw data
got the growth rates and then did a

00:42:55,270 --> 00:42:55,280
got the growth rates and then did a
 

00:42:55,280 --> 00:42:57,190
got the growth rates and then did a
check. Do have we met our conditions to

00:42:57,190 --> 00:42:57,200
check. Do have we met our conditions to
 

00:42:57,200 --> 00:42:59,670
check. Do have we met our conditions to
exit? If we have it generates a report

00:42:59,670 --> 00:42:59,680
exit? If we have it generates a report
 

00:42:59,680 --> 00:43:02,309
exit? If we have it generates a report
if we haven't it takes that data and it

00:43:02,309 --> 00:43:02,319
if we haven't it takes that data and it
 

00:43:02,319 --> 00:43:04,710
if we haven't it takes that data and it
adds it in and plans the next round of

00:43:04,710 --> 00:43:04,720
adds it in and plans the next round of
 

00:43:04,720 --> 00:43:06,470
adds it in and plans the next round of
experiments.

00:43:06,470 --> 00:43:06,480
experiments.
 

00:43:06,480 --> 00:43:07,829
experiments.
And again this was only a proof of

00:43:07,829 --> 00:43:07,839
And again this was only a proof of
 

00:43:07,839 --> 00:43:10,470
And again this was only a proof of
concept. Um we call it Aurelia uh

00:43:10,470 --> 00:43:10,480
concept. Um we call it Aurelia uh
 

00:43:10,480 --> 00:43:13,109
concept. Um we call it Aurelia uh
because everyone was using MVP for other

00:43:13,109 --> 00:43:13,119
because everyone was using MVP for other
 

00:43:13,119 --> 00:43:15,349
because everyone was using MVP for other
things. So that is the autonomously

00:43:15,349 --> 00:43:15,359
things. So that is the autonomously
 

00:43:15,359 --> 00:43:18,069
things. So that is the autonomously
repeating experimentation leveraging

00:43:18,069 --> 00:43:18,079
repeating experimentation leveraging
 

00:43:18,079 --> 00:43:20,790
repeating experimentation leveraging
integrated AI.

00:43:20,790 --> 00:43:20,800
integrated AI.
 

00:43:20,800 --> 00:43:23,190
integrated AI.
Uh but here's the here's kind of the

00:43:23,190 --> 00:43:23,200
Uh but here's the here's kind of the
 

00:43:23,200 --> 00:43:25,589
Uh but here's the here's kind of the
short version. So we're using uraria

00:43:25,589 --> 00:43:25,599
short version. So we're using uraria
 

00:43:25,599 --> 00:43:27,990
short version. So we're using uraria
lipolytica which is a industrial

00:43:27,990 --> 00:43:28,000
lipolytica which is a industrial
 

00:43:28,000 --> 00:43:31,750
lipolytica which is a industrial
relevant yeast. uh and it grows on a

00:43:31,750 --> 00:43:31,760
relevant yeast. uh and it grows on a
 

00:43:31,760 --> 00:43:33,349
relevant yeast. uh and it grows on a
carbon source as well as a nitrogen

00:43:33,349 --> 00:43:33,359
carbon source as well as a nitrogen
 

00:43:33,359 --> 00:43:36,069
carbon source as well as a nitrogen
source. And the nitrogen can be

00:43:36,069 --> 00:43:36,079
source. And the nitrogen can be
 

00:43:36,079 --> 00:43:37,990
source. And the nitrogen can be
expensive. So the experiment we're doing

00:43:37,990 --> 00:43:38,000
expensive. So the experiment we're doing
 

00:43:38,000 --> 00:43:40,630
expensive. So the experiment we're doing
here is to try to grow on a cheap

00:43:40,630 --> 00:43:40,640
here is to try to grow on a cheap
 

00:43:40,640 --> 00:43:42,870
here is to try to grow on a cheap
nitrogen source, which it doesn't grow

00:43:42,870 --> 00:43:42,880
nitrogen source, which it doesn't grow
 

00:43:42,880 --> 00:43:44,069
nitrogen source, which it doesn't grow
very well. But if you have a little bit

00:43:44,069 --> 00:43:44,079
very well. But if you have a little bit
 

00:43:44,079 --> 00:43:45,990
very well. But if you have a little bit
of the expensive one, you can get it to

00:43:45,990 --> 00:43:46,000
of the expensive one, you can get it to
 

00:43:46,000 --> 00:43:47,510
of the expensive one, you can get it to
grow. So we're trying to figure out

00:43:47,510 --> 00:43:47,520
grow. So we're trying to figure out
 

00:43:47,520 --> 00:43:49,829
grow. So we're trying to figure out
what's the the ratio that gives us the

00:43:49,829 --> 00:43:49,839
what's the the ratio that gives us the
 

00:43:49,839 --> 00:43:52,390
what's the the ratio that gives us the
best growth while trying to minimize the

00:43:52,390 --> 00:43:52,400
best growth while trying to minimize the
 

00:43:52,400 --> 00:43:55,190
best growth while trying to minimize the
amount of the expensive one.

00:43:55,190 --> 00:43:55,200
amount of the expensive one.
 

00:43:55,200 --> 00:43:57,349
amount of the expensive one.
Now in A we have to do some manual work

00:43:57,349 --> 00:43:57,359
Now in A we have to do some manual work
 

00:43:57,359 --> 00:43:59,349
Now in A we have to do some manual work
to get the the cells in the right spots

00:43:59,349 --> 00:43:59,359
to get the the cells in the right spots
 

00:43:59,359 --> 00:44:00,950
to get the the cells in the right spots
so that we can do the growth experiment

00:44:00,950 --> 00:44:00,960
so that we can do the growth experiment
 

00:44:00,960 --> 00:44:03,750
so that we can do the growth experiment
consistently. Uh B it shows the plate

00:44:03,750 --> 00:44:03,760
consistently. Uh B it shows the plate
 

00:44:03,760 --> 00:44:05,190
consistently. Uh B it shows the plate
that was the plate map that was

00:44:05,190 --> 00:44:05,200
that was the plate map that was
 

00:44:05,200 --> 00:44:08,069
that was the plate map that was
designed. We have uh repeats we checked

00:44:08,069 --> 00:44:08,079
designed. We have uh repeats we checked
 

00:44:08,079 --> 00:44:11,510
designed. We have uh repeats we checked
the edge effects uh and then it came up

00:44:11,510 --> 00:44:11,520
the edge effects uh and then it came up
 

00:44:11,520 --> 00:44:13,349
the edge effects uh and then it came up
with these conditions and then we built

00:44:13,349 --> 00:44:13,359
with these conditions and then we built
 

00:44:13,359 --> 00:44:16,550
with these conditions and then we built
the plates, we grew them and C shows a

00:44:16,550 --> 00:44:16,560
the plates, we grew them and C shows a
 

00:44:16,560 --> 00:44:19,510
the plates, we grew them and C shows a
few of the um the different conditions

00:44:19,510 --> 00:44:19,520
few of the um the different conditions
 

00:44:19,520 --> 00:44:21,750
few of the um the different conditions
that we tested and you can see some are

00:44:21,750 --> 00:44:21,760
that we tested and you can see some are
 

00:44:21,760 --> 00:44:24,230
that we tested and you can see some are
more consistent than others. uh but it

00:44:24,230 --> 00:44:24,240
more consistent than others. uh but it
 

00:44:24,240 --> 00:44:26,309
more consistent than others. uh but it
the tool is doing the statistics under

00:44:26,309 --> 00:44:26,319
the tool is doing the statistics under
 

00:44:26,319 --> 00:44:29,030
the tool is doing the statistics under
the hood and using that to produce D

00:44:29,030 --> 00:44:29,040
the hood and using that to produce D
 

00:44:29,040 --> 00:44:31,510
the hood and using that to produce D
which is the the map of the two

00:44:31,510 --> 00:44:31,520
which is the the map of the two
 

00:44:31,520 --> 00:44:33,990
which is the the map of the two
different nitrogen sources and the the

00:44:33,990 --> 00:44:34,000
different nitrogen sources and the the
 

00:44:34,000 --> 00:44:36,790
different nitrogen sources and the the
darkness of the blue says how how fast

00:44:36,790 --> 00:44:36,800
darkness of the blue says how how fast
 

00:44:36,800 --> 00:44:39,589
darkness of the blue says how how fast
the growth the maximum growth rate and

00:44:39,589 --> 00:44:39,599
the growth the maximum growth rate and
 

00:44:39,599 --> 00:44:40,950
the growth the maximum growth rate and
you can see it didn't have enough

00:44:40,950 --> 00:44:40,960
you can see it didn't have enough
 

00:44:40,960 --> 00:44:43,109
you can see it didn't have enough
information there to keep going. So the

00:44:43,109 --> 00:44:43,119
information there to keep going. So the
 

00:44:43,119 --> 00:44:45,829
information there to keep going. So the
red dots are the the uh ratios that it

00:44:45,829 --> 00:44:45,839
red dots are the the uh ratios that it
 

00:44:45,839 --> 00:44:48,309
red dots are the the uh ratios that it
picked for the second round. And at the

00:44:48,309 --> 00:44:48,319
picked for the second round. And at the
 

00:44:48,319 --> 00:44:50,630
picked for the second round. And at the
end we got at the second round we you

00:44:50,630 --> 00:44:50,640
end we got at the second round we you
 

00:44:50,640 --> 00:44:52,230
end we got at the second round we you
can see there's kind of a plateau in the

00:44:52,230 --> 00:44:52,240
can see there's kind of a plateau in the
 

00:44:52,240 --> 00:44:55,109
can see there's kind of a plateau in the
lower left. And so because we were just

00:44:55,109 --> 00:44:55,119
lower left. And so because we were just
 

00:44:55,119 --> 00:44:58,550
lower left. And so because we were just
doing proof of concept uh we said it's

00:44:58,550 --> 00:44:58,560
doing proof of concept uh we said it's
 

00:44:58,560 --> 00:45:00,950
doing proof of concept uh we said it's
done found it it found its confidence

00:45:00,950 --> 00:45:00,960
done found it it found its confidence
 

00:45:00,960 --> 00:45:04,309
done found it it found its confidence
well enough to produce that uh to

00:45:04,309 --> 00:45:04,319
well enough to produce that uh to
 

00:45:04,319 --> 00:45:07,750
well enough to produce that uh to
produce the exit and it worked. But you

00:45:07,750 --> 00:45:07,760
produce the exit and it worked. But you
 

00:45:07,760 --> 00:45:09,190
produce the exit and it worked. But you
can see this obviously needs a lot of

00:45:09,190 --> 00:45:09,200
can see this obviously needs a lot of
 

00:45:09,200 --> 00:45:10,710
can see this obviously needs a lot of
improvement, right? didn't didn't look

00:45:10,710 --> 00:45:10,720
improvement, right? didn't didn't look
 

00:45:10,720 --> 00:45:13,349
improvement, right? didn't didn't look
at anything in the upper right. Uh we

00:45:13,349 --> 00:45:13,359
at anything in the upper right. Uh we
 

00:45:13,359 --> 00:45:15,430
at anything in the upper right. Uh we
could tighten this up. We can show more

00:45:15,430 --> 00:45:15,440
could tighten this up. We can show more
 

00:45:15,440 --> 00:45:18,550
could tighten this up. We can show more
about how the um how it decided to maybe

00:45:18,550 --> 00:45:18,560
about how the um how it decided to maybe
 

00:45:18,560 --> 00:45:20,390
about how the um how it decided to maybe
ignore that one dark spot up in the the

00:45:20,390 --> 00:45:20,400
ignore that one dark spot up in the the
 

00:45:20,400 --> 00:45:22,790
ignore that one dark spot up in the the
upper left. But fundamentally it works

00:45:22,790 --> 00:45:22,800
upper left. But fundamentally it works
 

00:45:22,800 --> 00:45:26,550
upper left. But fundamentally it works
and now we are working to improve this

00:45:26,550 --> 00:45:26,560
and now we are working to improve this
 

00:45:26,560 --> 00:45:28,870
and now we are working to improve this
tool and also to extend it. So this was

00:45:28,870 --> 00:45:28,880
tool and also to extend it. So this was
 

00:45:28,880 --> 00:45:30,470
tool and also to extend it. So this was
done with a fluent that had a plate

00:45:30,470 --> 00:45:30,480
done with a fluent that had a plate
 

00:45:30,480 --> 00:45:33,109
done with a fluent that had a plate
reader and a single incubator shaker.

00:45:33,109 --> 00:45:33,119
reader and a single incubator shaker.
 

00:45:33,119 --> 00:45:36,630
reader and a single incubator shaker.
Now we have AMP 2. We have access to a

00:45:36,630 --> 00:45:36,640
Now we have AMP 2. We have access to a
 

00:45:36,640 --> 00:45:38,470
Now we have AMP 2. We have access to a
larger and better incubator. we have

00:45:38,470 --> 00:45:38,480
larger and better incubator. we have
 

00:45:38,480 --> 00:45:40,069
larger and better incubator. we have
more analysis tools that we can rope

00:45:40,069 --> 00:45:40,079
more analysis tools that we can rope
 

00:45:40,079 --> 00:45:42,710
more analysis tools that we can rope
into this. Uh so the team is working to

00:45:42,710 --> 00:45:42,720
into this. Uh so the team is working to
 

00:45:42,720 --> 00:45:45,109
into this. Uh so the team is working to
expand it

00:45:45,109 --> 00:45:45,119
expand it
 

00:45:45,119 --> 00:45:48,870
expand it
but um overall we're getting there. Uh

00:45:48,870 --> 00:45:48,880
but um overall we're getting there. Uh
 

00:45:48,880 --> 00:45:51,910
but um overall we're getting there. Uh
and it's a process that we need to

00:45:51,910 --> 00:45:51,920
and it's a process that we need to
 

00:45:51,920 --> 00:45:53,910
and it's a process that we need to
accomplish by the time MTPC gets here.

00:45:53,910 --> 00:45:53,920
accomplish by the time MTPC gets here.
 

00:45:53,920 --> 00:45:57,030
accomplish by the time MTPC gets here.
So this is our our grand vision. Uh on

00:45:57,030 --> 00:45:57,040
So this is our our grand vision. Uh on
 

00:45:57,040 --> 00:45:58,950
So this is our our grand vision. Uh on
the upper upper side we've got the

00:45:58,950 --> 00:45:58,960
the upper upper side we've got the
 

00:45:58,960 --> 00:46:01,430
the upper upper side we've got the
green. This is onboarding workflows

00:46:01,430 --> 00:46:01,440
green. This is onboarding workflows
 

00:46:01,440 --> 00:46:03,829
green. This is onboarding workflows
getting into that standardized format.

00:46:03,829 --> 00:46:03,839
getting into that standardized format.
 

00:46:03,839 --> 00:46:05,910
getting into that standardized format.
The blue is the the platform. So

00:46:05,910 --> 00:46:05,920
The blue is the the platform. So
 

00:46:05,920 --> 00:46:07,510
The blue is the the platform. So
generating taking that standard format

00:46:07,510 --> 00:46:07,520
generating taking that standard format
 

00:46:07,520 --> 00:46:10,630
generating taking that standard format
and generating scripts that can run. And

00:46:10,630 --> 00:46:10,640
and generating scripts that can run. And
 

00:46:10,640 --> 00:46:14,069
and generating scripts that can run. And
by separating those two, we can

00:46:14,069 --> 00:46:14,079
by separating those two, we can
 

00:46:14,079 --> 00:46:16,069
by separating those two, we can
build these tools out in a modular way.

00:46:16,069 --> 00:46:16,079
build these tools out in a modular way.
 

00:46:16,079 --> 00:46:19,190
build these tools out in a modular way.
So folks who are focused on um the

00:46:19,190 --> 00:46:19,200
So folks who are focused on um the
 

00:46:19,200 --> 00:46:20,790
So folks who are focused on um the
onboarding tools, they don't have to

00:46:20,790 --> 00:46:20,800
onboarding tools, they don't have to
 

00:46:20,800 --> 00:46:23,190
onboarding tools, they don't have to
worry about also making it work with the

00:46:23,190 --> 00:46:23,200
worry about also making it work with the
 

00:46:23,200 --> 00:46:25,349
worry about also making it work with the
Fluent and with the AMP 2 and with every

00:46:25,349 --> 00:46:25,359
Fluent and with the AMP 2 and with every
 

00:46:25,359 --> 00:46:27,109
Fluent and with the AMP 2 and with every
other system. They can just get to the

00:46:27,109 --> 00:46:27,119
other system. They can just get to the
 

00:46:27,119 --> 00:46:28,470
other system. They can just get to the
standard format. And then we have

00:46:28,470 --> 00:46:28,480
standard format. And then we have
 

00:46:28,480 --> 00:46:30,069
standard format. And then we have
specialists that work with the

00:46:30,069 --> 00:46:30,079
specialists that work with the
 

00:46:30,079 --> 00:46:31,589
specialists that work with the
automation that can take from the

00:46:31,589 --> 00:46:31,599
automation that can take from the
 

00:46:31,599 --> 00:46:33,670
automation that can take from the
standard format and get it down to the

00:46:33,670 --> 00:46:33,680
standard format and get it down to the
 

00:46:33,680 --> 00:46:36,069
standard format and get it down to the
the the scripting level. We have our

00:46:36,069 --> 00:46:36,079
the the scripting level. We have our
 

00:46:36,079 --> 00:46:38,230
the the scripting level. We have our
science central where the data comes

00:46:38,230 --> 00:46:38,240
science central where the data comes
 

00:46:38,240 --> 00:46:41,589
science central where the data comes
out. And then MIDAS on the orange side

00:46:41,589 --> 00:46:41,599
out. And then MIDAS on the orange side
 

00:46:41,599 --> 00:46:43,510
out. And then MIDAS on the orange side
doing the the tools that interact with

00:46:43,510 --> 00:46:43,520
doing the the tools that interact with
 

00:46:43,520 --> 00:46:45,990
doing the the tools that interact with
the models to take the data, improve the

00:46:45,990 --> 00:46:46,000
the models to take the data, improve the
 

00:46:46,000 --> 00:46:47,829
the models to take the data, improve the
models, use the models to make

00:46:47,829 --> 00:46:47,839
models, use the models to make
 

00:46:47,839 --> 00:46:50,309
models, use the models to make
predictions, to plan experiments and do

00:46:50,309 --> 00:46:50,319
predictions, to plan experiments and do
 

00:46:50,319 --> 00:46:52,870
predictions, to plan experiments and do
that autonomous loop.

00:46:52,870 --> 00:46:52,880
that autonomous loop.
 

00:46:52,880 --> 00:46:55,349
that autonomous loop.
And the um

00:46:55,349 --> 00:46:55,359
And the um
 

00:46:55,359 --> 00:46:57,109
And the um
the the last piece I would say about

00:46:57,109 --> 00:46:57,119
the the last piece I would say about
 

00:46:57,119 --> 00:47:00,309
the the last piece I would say about
this is that this this model side can be

00:47:00,309 --> 00:47:00,319
this is that this this model side can be
 

00:47:00,319 --> 00:47:03,030
this is that this this model side can be
as as complex as it needs to be. So it

00:47:03,030 --> 00:47:03,040
as as complex as it needs to be. So it
 

00:47:03,040 --> 00:47:04,630
as as complex as it needs to be. So it
could be a simple design of experiment

00:47:04,630 --> 00:47:04,640
could be a simple design of experiment
 

00:47:04,640 --> 00:47:08,230
could be a simple design of experiment
tool uh or maybe we can incorporate open

00:47:08,230 --> 00:47:08,240
tool uh or maybe we can incorporate open
 

00:47:08,240 --> 00:47:11,670
tool uh or maybe we can incorporate open
AI and do a much more complex design but

00:47:11,670 --> 00:47:11,680
AI and do a much more complex design but
 

00:47:11,680 --> 00:47:14,150
AI and do a much more complex design but
that still doesn't have to worry about

00:47:14,150 --> 00:47:14,160
that still doesn't have to worry about
 

00:47:14,160 --> 00:47:15,910
that still doesn't have to worry about
how do we implement that design on our

00:47:15,910 --> 00:47:15,920
how do we implement that design on our
 

00:47:15,920 --> 00:47:17,430
how do we implement that design on our
automation. They can just focus on that

00:47:17,430 --> 00:47:17,440
automation. They can just focus on that
 

00:47:17,440 --> 00:47:19,670
automation. They can just focus on that
tool.

00:47:19,670 --> 00:47:19,680
tool.
 

00:47:19,680 --> 00:47:23,349
tool.
And so that's our vision. And uh I

00:47:23,349 --> 00:47:23,359
And so that's our vision. And uh I
 

00:47:23,359 --> 00:47:25,109
And so that's our vision. And uh I
realize now after watching Will I forgot

00:47:25,109 --> 00:47:25,119
realize now after watching Will I forgot
 

00:47:25,119 --> 00:47:27,030
realize now after watching Will I forgot
to put my contact information up here,

00:47:27,030 --> 00:47:27,040
to put my contact information up here,
 

00:47:27,040 --> 00:47:29,349
to put my contact information up here,
but you can find me at Hemsel uh Todd

00:47:29,349 --> 00:47:29,359
but you can find me at Hemsel uh Todd
 

00:47:29,359 --> 00:47:31,589
but you can find me at Hemsel uh Todd
Edwards. Find me on LinkedIn. I'm happy

00:47:31,589 --> 00:47:31,599
Edwards. Find me on LinkedIn. I'm happy
 

00:47:31,599 --> 00:47:34,470
Edwards. Find me on LinkedIn. I'm happy
to chat. Uh and then yeah, definitely if

00:47:34,470 --> 00:47:34,480
to chat. Uh and then yeah, definitely if
 

00:47:34,480 --> 00:47:35,990
to chat. Uh and then yeah, definitely if
you're a data person, check out Science

00:47:35,990 --> 00:47:36,000
you're a data person, check out Science
 

00:47:36,000 --> 00:47:37,589
you're a data person, check out Science
Central and make use of that data

00:47:37,589 --> 00:47:37,599
Central and make use of that data
 

00:47:37,599 --> 00:47:40,950
Central and make use of that data
because taxpayer money funded it. So

00:47:40,950 --> 00:47:40,960
because taxpayer money funded it. So
 

00:47:40,960 --> 00:47:42,950
because taxpayer money funded it. So
it's free. Thank you.

00:47:42,950 --> 00:47:42,960
it's free. Thank you.
 

00:47:42,960 --> 00:47:47,670
it's free. Thank you.
>> Let's thank Joy and Todd.

00:47:47,670 --> 00:47:47,680

 

00:47:47,680 --> 00:47:50,920

>> It's GKO.