Episode 16: Building AI for Life Sciences
OpenAI
0:00 Andrew Mayne: Hello, I'm Andrew Mayne, and this is the OpenAI Podcast.
0:03 On today's episode, we're talking with Andrew Mayne:
0:05 research lead Joy Jiao and product lead
0:08 Yunyun Wang about OpenAI for Life Sciences.
0:10 We'll explore Andrew Mayne:
0:11 what new models are making possible in biology and medicine
0:13 and what it takes to deploy the most advanced capabilities responsibly.
0:16 Joy Jiao: This allows it to kind of reach new levels of difficulty and Joy Jiao:
0:22 discovery that we didn't think was even possible before.
0:27 Yunyun Wang: Putting like really capable expert level knowledge
0:29 in the hands of a greater amount of people.
0:32 Joy Jiao: One of the taglines was
0:33 to scale test time compute to cure all disease.
0:34 So that is like our team tagline.
0:39 Andrew Mayne: We started off with just a basic API and then we had ChatGPT,
0:44 which is more conversational, Andrew Mayne:
0:46 was really good for text as code became a capability,
0:49 went through basically code models Andrew Mayne: and then Codex.
0:52 Now that you're getting more scientists and the life
0:54 sciences working on Andrew Mayne: these systems,
0:55 does that mean things have to evolve to help
0:58 with the way researchers might work with these tools?
1:00 Yunyun Wang: Yeah, we're really excited to build
1:03 and deploy the life sciences model series.
1:06 Yunyun Wang: So this is a new biochemistry focused model
1:09 series that's really anchored on these very complex Yunyun Wang:
1:13 life science research workflows.
1:15 And we're focused on adding new like mechanistic understanding, Yunyun Wang:
1:21 starting with genomics understanding and protein understanding,
1:23 and really focused on early Yunyun Wang: discovery use cases,
1:27 because we feel like that's one
1:28 of the core bottlenecks that greater thinking Yunyun Wang:
1:31 time, greater compute,
1:32 and really leveraging more capable AI models can help meaningfully Yunyun Wang:
1:37 scale some of these research barriers.
1:39 And I think there's also a model orchestration piece of Yunyun Wang:
1:43 actually how to embed this into workflows.
1:45 And it's been really great, first off,
1:47 having all these Yunyun Wang: different product surfaces to deploy to.
1:50 Yunyun Wang: We're seeing a lot
1:52 of really great literature synthesis Yunyun Wang:
1:54 workflows happening on ChatGPT.
1:57 Yunyun Wang: And these models really push
1:59 the frontier of long trajectory Yunyun Wang: agentic workflows.
2:02 Yunyun Wang: And we're really able to empower that on Codex.
2:05 Yunyun Wang: And more on the model orchestration piece Yunyun Wang:
2:09 is that I think for enterprise use cases,
2:12 Yunyun Wang: there's this reproducibility and repeatability element.
2:16 Yunyun Wang: And we are trying to overcome this by working
2:19 on some of the life sciences research plugins Yunyun Wang:
2:23 that we're shipping for very specific translational bio users.
2:26 Yunyun Wang: So the life sciences research plugin has over 50 skills,
2:30 which are essentially templatized Yunyun Wang:
2:33 repeatable workflows that if you need to, whether
2:35 do some sort of cross evidence match Yunyun Wang:
2:39 and search across various different papers, or do pathway analysis,
2:43 something that's like Yunyun Wang: repeatable that you often do,
2:47 we can have almost like a one-click deploy option by using our Yunyun Wang:
2:51 life sciences plugins on top.
2:53 And that's also how we're seeing
2:55 the balance between scaling for very specialized purposes.
2:57 Yunyun Wang: Something we're hoping to get into is maybe clinical purposes,
3:02 but also make Yunyun Wang:
3:03 it still very general use for all foundational biology.
3:06 Joy Jiao: I think the models can get quite far by Joy Jiao: using tools.
3:10 So for example, we can use open
3:12 source protein structure prediction algorithms Joy Jiao:
3:15 and start a research stack.
3:16 And in this case, the model is acting kind of like a regular Joy Jiao:
3:19 computational biologist.
3:20 You will kind of go run these tools on a computer.
3:22 You will look at the Joy Jiao: output.
3:24 You will tweak the input a little bit.
3:26 So I think that is something our models can already Joy Jiao: do.
3:29 I do think what will make the models even more
3:33 powerful is to start to turn them more into Joy Jiao:
3:36 kind of a biochemistry expert.
3:38 And I think with this kind of intuition and expertise, Joy Jiao:
3:42 you can use these tools even more intelligently
3:44 and get at the right answer more quickly.
3:47 Andrew Mayne: How did you get your interest in life sciences?
3:49 Joy Jiao: My, I guess, original background was actually in life sciences.
3:54 So I've always been interested in Joy Jiao: biology as a kid.
3:58 I got my PhD in systems biology around like a decade ago from Harvard.
4:03 Joy Jiao: I found academia to be very interesting,
4:06 but the pace was a little bit more slow moving than I would have liked.
4:11 Joy Jiao: And I think just the experience
4:13 of kind of like having to physically be in the lab and kind of like transferring
4:17 small amounts of liquid from one tool to another.
4:19 Joy Jiao: I think I wanted something a little bit faster paced where
4:22 I felt like I was more in direct control of my own velocity.
4:26 Joy Jiao: So I went from that to software and I ended up here at OpenAI.
4:30 Joy Jiao: And so this is kind of like a full circle moment for me Joy Jiao:
4:33 where I'm starting to look at biology again Joy Jiao:
4:35 and looking at how to accelerate my previous self with AI.
4:39 Joy Jiao: So yeah, really excited to see
4:41 what progress AI can make in this space.
4:42 Andrew Mayne: So you're like, yeah, this is too slow.
4:44 Andrew Mayne: Let me go off an AI and speed it up so I can get back into it.
4:47 Joy Jiao: Yeah, except from this end, Joy Jiao:
4:49 I don't really ever want to touch a pipette or anything again.
4:52 Joy Jiao: So I would prefer for my robots to do it for me.
4:54 Yunyun Wang: Yeah, we joke about that a lot.
4:55 Yunyun Wang: A lot of our motivation
4:57 for this is we can automate pipetting Yunyun Wang:
4:58 and never have to do that again.
5:00 Andrew Mayne: Well, that's what's interesting.
5:02 I was looking at what you all did with Ginkgo
5:04 Bioworks and the idea of taking GPT-5 and taking
5:07 an AI system and then working with a robotic
5:09 lab and how it was able to speed things up.
5:12 Could you tell us a little about that?
5:14 Joy Jiao: Yeah, the Ginkgo work is interesting because I think when it started,
5:19 I think it was like July of last year, 2025.
5:23 Joy Jiao: And at that point, GPT-5 had just finished training.
5:26 Joy Jiao: We were really not sure if the models could do any kind of biology.
5:30 Joy Jiao: We didn't really have that much biology in our training data.
5:32 Joy Jiao: It was mostly math and computer science,
5:35 which I think makes sense because these things have verifiable solutions.
5:38 Joy Jiao: And this is usually not the case in biology
5:41 unless you can go and do the experiment in a lab, right?
5:44 Joy Jiao: So when we started the collaboration with Ginkgo,
5:47 it was really, can the model do any biology at all?
5:49 Joy Jiao: Can it design experiments that actually make reactants,
5:53 like make the product that we want?
5:56 Joy Jiao: So it was actually quite surprising, I think,
5:59 when GPT-5 designed the first set of experiments with Ginkgo.
6:02 Joy Jiao: And the results came back.
6:03 Joy Jiao: Oh, we made a non-zero amount of protein.
6:05 Joy Jiao: That was actually quite surprising.
6:07 Joy Jiao: And then I think progressing from that point in time,
6:11 which is just roughly like six months ago to now,
6:14 where it actually just feels quite obvious that our models
6:17 can accelerate science is actually just really surprising.
6:20 Yunyun Wang: And it really shows the art of the possible, I think.
6:22 Yunyun Wang: I think before that experiment led
6:25 by Joy and the Ginkgo team was conducted,
6:28 I think we really didn't know for ourselves.
6:31 Yunyun Wang: And I always say, like,
6:33 we kind of learn that for ourselves when we engage in these experiments
6:36 and we have a few more in the works with others.
6:40 Yunyun Wang: And I think that is like
6:42 the type of acceleration that we're looking for.
6:45 Yunyun Wang: Ingesting high throughput experimental data is really difficult.
6:47 Yunyun Wang: It's very compute intensive.
6:49 Yunyun Wang: And I think for a lot of these scientific workflows,
6:54 like the true bottleneck for the speed and progress
6:58 of scientific acceleration is at like almost a human bottlenecks.
7:02 Yunyun Wang: And I think the future that me
7:05 and Joy see is that it's no longer human bottlenecks,
7:09 but rather maybe compute bottlenecks.
7:11 Yunyun Wang: And we're really able to deploy many sub-agents
7:15 doing parallel orchestration to divide and conquer all these tasks.
7:19 Yunyun Wang: And the researcher can now spend their time on really analyzing,
7:23 interpreting the most meaningful insights coming out of that.
7:26 Andrew Mayne: So Yunyun, how did you get into this?
7:28 Yunyun Wang: Yeah, I think reflecting back,
7:30 I've actually been working on like biology research in some shape
7:33 or form for a majority of my time here at OpenAI.
7:36 Yunyun Wang: So I first started on working
7:38 on biorisk mitigations and a lot of our biodefense initiatives.
7:42 Yunyun Wang: And so I feel like coming
7:43 to now working on the life sciences research side
7:45 gives me like just appreciation for how difficult
7:47 this problem is and tackling it from both sides.
7:52 Yunyun Wang: And my initial entries point into wet lab research
7:55 was actually through doing a lot of infectious disease and virology work.
7:59 Yunyun Wang: So I think I've always
8:01 done the interest in biosecurity in that way.
8:02 Yunyun Wang: So this just feels like a really
8:05 great moment right now to work on it,
8:07 especially when our models are getting more capable
8:10 with beneficial use and just general life sciences.
8:12 Andrew Mayne: How long has OpenAI been focused on life sciences?
8:15 Yunyun Wang: Yeah, I would say it was really the way
8:18 we design our capability evals that show us that this is possible.
8:22 Yunyun Wang: So it's been, I think, for at least two years now that we have
8:27 worked on a lot of our early research experiments.
8:31 Yunyun Wang: And now with the Ginkgo
8:33 autonomous wet lab model in the loop experiments.
8:35 Joy Jiao: I think we have a few more
8:37 research partners in the space that we're really excited about.
8:39 Joy Jiao: I think I can't actually name everyone right now,
8:42 but there's a lot of stuff kind of in the chemical design,
8:46 protein design, enzyme design space that I think is very
8:48 AI native and a lot of people are interested in.
8:51 Joy Jiao: So understanding how the world works,
8:54 understanding how chemicals react, understanding how cells interact,
8:57 how pathways inside cells interact,
8:59 all the way to can we accelerate drug discovery?
9:03 Joy Jiao: So given a disease can
9:05 a model help scientists understand the mechanism,
9:07 can we once given a target actually design a drug against that target?
9:11 Joy Jiao: Can we even accelerate the FDA approval process?
9:14 Joy Jiao: So I think there's a role for AI
9:16 to play kind of at every step of this pipeline.
9:18 Joy Jiao: And yeah, I think there's a lot of AI possible in everything.
9:23 Andrew Mayne: I've been to some of those cutting edge
9:25 labs and on the outside you have this impression of it.
9:27 Andrew Mayne: Then you walk in there and you
9:30 literally see somebody with a row of Petri dishes,
9:32 a row of samples, and just some grad student going click, click, click.
9:36 Andrew Mayne: And I'm like, oh, this is the pace of science.
9:38 Andrew Mayne: This is fast.
9:39 Andrew Mayne: With me.
9:40 Andrew Mayne: Yeah, exactly.
9:41 Andrew Mayne: You're like, enough of this.
9:42 Andrew Mayne: I got to go speed this up.
9:43 Andrew Mayne: But we forget that's often the pace of science
9:45 is just how fast the human hands can move through that.
9:48 Andrew Mayne: And with a tool like that, it's kind of exciting.
9:50 Andrew Mayne: When you start using these tools to maybe
9:54 think about new pathways for treatments or just evaluate,
9:57 you also introduce the idea that these could
9:59 be used for things that maybe are less desirable.
10:03 Andrew Mayne: Bioweapons are something that comes up a lot.
10:04 Andrew Mayne: The fact that if an AI can figure out how to do a code exploit,
10:07 might be able to figure out how to do a gene exploit.
10:10 Andrew Mayne: How are you addressing that?
10:11 Yunyun Wang: Yeah, that's a great question.
10:13 Yunyun Wang: And I think it is just probably one of the most
10:17 severe risks that we're currently really tracking for rising AI capabilities.
10:22 Yunyun Wang: Our first approach to that was really
10:25 thinking about how do we assess for information hazards?
10:28 Yunyun Wang: At what point does a model now maybe
10:32 give the final step in a synthesis of a dangerous pathogen?
10:37 Yunyun Wang: And what we found is
10:39 that the precursor steps to that really looks very benign.
10:42 Yunyun Wang: And it's really hard to distinguish between.
10:44 Yunyun Wang: So another way to put it is like the same steps that a beneficial,
10:49 like a legitimate actor might take is looks very
10:52 similar to the ones that a dangerous, harmful actor.
10:56 Andrew Mayne: You start with something that's.
10:57 Yunyun Wang: Yes, exactly.
10:58 Yunyun Wang: So I still think that we we made the right
11:02 call for really taking a very risk averse approach to that.
11:05 Yunyun Wang: But now I'm really excited about
11:08 like differentiated access and like responsible deployment as really
11:11 a core pillar of all of our safeguards work
11:13 and really understanding that there are different user segments.
11:17 Yunyun Wang: And I almost feel like the future
11:20 we're going towards is something like models as a professions,
11:23 similar to how they models have different personalities.
11:26 Yunyun Wang: And sometimes you want to invoke the right one,
11:31 depending on the type of like workflow you're looking at.
11:35 Yunyun Wang: So I think how this translates is similar
11:40 to how biologists working on like therapeutics and their research,
11:44 they require access to data sets are often very
11:48 tightly controlled or they require access to just expert level.
11:52 Yunyun Wang: Like they all have PhDs
11:54 and have like expert level like biology, like knowledge.
11:58 Yunyun Wang: How does that compare to how
12:00 does that translate over to two models?
12:02 Yunyun Wang: I think that's why we have
12:04 to kind of similarly take the same training approach,
12:06 but also the same security approach and deploying that in like a way
12:09 where we can have those very heightened enterprise grade controls in place.
12:13 Andrew Mayne: So you just mentioned safeguards.
12:15 Andrew Mayne: Can you explain how that applies here,
12:16 where you would need them, why you would need them?
12:18 Yunyun Wang: Yeah.
12:19 Yunyun Wang: So we very thoughtfully design and design new safeguards
12:22 for pretty much all of our models across very different risk areas.
12:26 Yunyun Wang: But I think when it came to bio,
12:29 this was like the first dual use risk that is both also a capability risk.
12:32 Yunyun Wang: So it very much correlates with how we as capabilities improve,
12:37 the risk correlates.
12:39 Yunyun Wang: And I think that's why our first approach,
12:42 when we really there was no precedent for a lot of this work,
12:45 Yunyun Wang: and we were the first to really activate these high
12:49 safeguards when we saw that significant
12:51 reasoning jump in our model capabilities.
12:54 Yunyun Wang: We really wanted to make sure that we did it right.
12:57 Yunyun Wang: And I think the best way
12:59 to get it right is to incrementally deploy.
13:01 Joy Jiao: Yeah, I think it's really a fine
13:05 line between having a very capable model that's capable
13:08 of accelerating benign science and beneficial science versus
13:11 a model that could be used by a bad actor.
13:15 Joy Jiao: And I think the safest model here
13:17 would be a model just had no capability, right?
13:20 Andrew Mayne: It's not very good.
13:21 Joy Jiao: Yeah, it's not very good, but it's very safe.
13:24 Joy Jiao: And on the other hand,
13:26 if you had a model that is basically an oracle of the physical world,
13:29 it basically knows everything about every experiment,
13:31 that model could fall into the wrong hands and do
13:34 potentially very bad things because someone can go and say,
13:36 hey, design a new novel pandemic potential pathogen.
13:39 Joy Jiao: And the model can just go and do that autonomously.
13:42 Joy Jiao: So I think we need to kind
13:44 of figure out where we draw the line in between
13:47 the two and kind of think about who gets
13:49 access to a potentially very capable model and who doesn't.
13:52 Joy Jiao: And what we found in kind
13:55 of what we call general access traffic is that it's
13:59 very difficult to figure out what a user's actual
14:02 intentions are just from kind of reading a prompt.
14:06 Joy Jiao: And I think as an example of this, let's say someone says,
14:09 hey, help me clone a gene.
14:10 Joy Jiao: The model might not even be given what the gene is,
14:13 but it can come up with a protocol for it.
14:16 Joy Jiao: And so this gene could just be something
14:18 like green fluorescent protein or it could be a toxin.
14:20 Joy Jiao: And there's basically no way to figure
14:22 that out from the context of the conversation.
14:25 Joy Jiao: And so this becomes a very difficult problem in production.
14:30 Joy Jiao: And basically, I think, like you said,
14:33 we decided to kind of err on the side of safety Joy Jiao:
14:37 here and basically say that, OK,
14:39 if we think that there is a potential for misuse, we either Joy Jiao:
14:43 have the model kind of self-refuse the user,
14:45 in which case it tends to say things like, Joy Jiao:
14:48 Sorry, I can't really help you with that, but I
14:49 can give you a high-level overview of this protocol instead.
14:52 Joy Jiao: And this, unfortunately,
14:54 very much annoys our kind of professional scientists, understandably.
14:59 Joy Jiao: And we also kind of have multiple layers of mitigation on top of that.
15:03 Joy Jiao: But I think really to unlock the full capabilities of our models,
15:07 what we need is this differentiated access.
15:09 Joy Jiao: And what this means is we know who the user actually is.
15:13 Joy Jiao: They are a professional working
15:15 at a legitimate research institution or a pharma company.
15:18 Joy Jiao: And because of the regulations around these institutions,
15:22 we know that, for example, all Joy Jiao: the reagents are being tracked.
15:25 Joy Jiao: All the cell lines that they're using are being tracked.
15:27 Joy Jiao: And so this gives us confidence that this is
15:30 a legitimate user and not a random person Joy Jiao:
15:33 in a basement doing who knows what.
15:35 Joy Jiao: And that allows us to give them basically
15:38 more capabilities than we are able to provide Joy Jiao:
15:41 to the general access traffic.
15:43 Andrew Mayne: What can you do right now?
15:45 Andrew Mayne: If you're working with the models,
15:47 Andrew Mayne: you're working it within a laboratory,
15:49 Andrew Mayne: what would you say the capability is at this moment?
15:51 Joy Jiao: So I think people use the models Joy Jiao: for very different things.
15:57 Joy Jiao: I've talked to people in the Baker Lab recently Joy Jiao:
16:00 on kind of how they've been using our models on Codex.
16:03 Joy Jiao: And sometimes it's as simple as, Joy Jiao:
16:06 hey, can you write a spreadsheet for me?
16:07 Joy Jiao: I want to just minimize the number of pipetting steps Joy Jiao:
16:10 that I have to make.
16:11 Joy Jiao: And this hits me very hard because I
16:13 had done the same thing by hand in grad school.
16:16 Joy Jiao: So that's like a very simple just mathematical software operation.
16:20 Joy Jiao: And then there's much harder tasks.
16:22 Joy Jiao: Can you design a enzyme for me without these biological design tools?
16:27 Joy Jiao: So I think there's a huge range of sophistication.
16:30 Joy Jiao: Yeah.
16:30 Yunyun Wang: And something I'm very excited about is how we
16:33 can use our models to be a more powerful Yunyun Wang:
16:37 discriminator and like really testing and assessing like new novel ideas.
16:41 And I think something that Yunyun Wang:
16:43 I've been noticing as a trend with a lot of our research
16:46 partners and also the users of our models Yunyun Wang:
16:49 is that models for scientific research and tasks almost
16:52 require a different like persona or a Yunyun Wang: different prompting style.
16:56 So we, I often feel like, you know, like a model that is much more Yunyun Wang:
17:02 scrutinizing or a skeptic at good ideas is it's
17:04 very similar to how human scientists would go Yunyun Wang:
17:08 assess originality and feasibility.
17:09 It's really, I think, helping understand out of all the new Yunyun Wang:
17:15 papers and new publications out there that push
17:17 the frontier of a lot of these hypotheses,
17:19 Yunyun Wang: what are the ones that are really feasible
17:21 and valid for testing that's going to help Yunyun Wang:
17:24 lead to new breakthroughs?
17:25 And then translating this to something like disease target screening,
17:29 Yunyun Wang: selection, like the potentials for these drug targets are endless,
17:32 but it's really like narrowing Yunyun Wang: down the aperture.
17:34 And I feel like that's where like the assistance comes.
17:37 Like this is extremely Yunyun Wang: difficult work to do at scale and having
17:40 a model that can like empower and accelerate that process,
17:43 Yunyun Wang: I think is kind of like one of the immediate
17:46 impacts we're hoping to see by responsibly deploying this model to those users.
17:49 Andrew Mayne: It seems like it's a very interesting trajectory.
17:53 You went from Andrew Mayne: There was, you had GPT-3 on the API and GPT-3.5,
17:58 then you get ChatGPT, and now we have ChatGPT apps, and now we have Codex.
18:03 Andrew Mayne: And it sounds like these things,
18:05 just the number of things you can do with this continues to grow.
18:09 Andrew Mayne: How would you see this building?
18:11 Andrew Mayne: You know,
18:12 do you see this as basically just becoming a complete infrastructure
18:14 for kind of every kind of inquiry you might want to pursue?
18:17 Joy Jiao: Yeah, I think the dream is to have
18:20 a lot of the basic foundations of scientific workflows happen on Codex.
18:25 Joy Jiao: and I think that the goal is to have Codex
18:28 to pretty much be able to do everything that is Joy Jiao:
18:31 possible to do on the computer.
18:33 Of course, we also want to extend beyond that with kind of Joy Jiao:
18:36 hooking it up to robotics and so forth.
18:39 But I think right now we already do things, for example, Joy Jiao:
18:43 if we have a bunch of different dev boxes on our remote, on our laptop,
18:46 we can actually say, Joy Jiao: hey, Codex,
18:49 go and run this code on all of these different dev boxes
18:52 that are all remote and then Joy Jiao: Codex can do that.
18:55 We can say, monitor this for me.
18:56 And I can kind of like go away and do something Joy Jiao: else.
18:58 And the Codex are just like, they're watching all the logs for you.
19:02 It can build a lot of just Joy Jiao:
19:04 kind of fit for purpose software for analyzing specific pieces of data,
19:08 for visualizing data.
19:09 Joy Jiao: So for example,
19:10 if we have experimental biology data that we're sending each other on the team,
19:14 Joy Jiao: what I've noticed recently is instead of sending the raw data,
19:18 we've started sending HTML files Joy Jiao:
19:20 are just these kind of like beautiful UIs
19:21 that Codex has built with kind of like spinning proteins.
19:24 Joy Jiao: And it's actually just really,
19:26 it kind of changes the way that we share with each other and collaborate.
19:29 Yunyun Wang: Yeah, when we first started mapping out
19:33 how users and organizations might adopt this, Yunyun Wang:
19:36 I think we envisioned that each scientist
19:39 would get their personal assistant or their coworker.
19:43 Yunyun Wang: And this is a way that they can scale up their collective output.
19:49 Yunyun Wang: And then the next paradigm of that would
19:53 be scaling up whole research institutions where a whole
19:56 program team can actually deploy a workforce of various
19:59 agents and they can all do parallel task delegation,
20:02 kind of mimicking a lot of these existing patterns.
20:06 Yunyun Wang: And we can figure out the pieces of how
20:10 they can all collectively work together to solve larger tasks.
20:14 Andrew Mayne: It's interesting because OpenAI
20:16 has talked about the need for compute.
20:18 Andrew Mayne: And I think that sometimes we just sort of think like,
20:21 okay, so I can have more conversations and stuff.
20:23 Andrew Mayne: But when you're talking about the idea
20:26 of building these tools to become entire platforms or scientific exploration,
20:30 it sounds like the compute advantage is really critical.
20:35 Joy Jiao: Yeah, I think there's two different axises
20:37 we can think about how we are scaling compute.
20:40 Joy Jiao: The one that I think
20:41 everyone's familiar with is just getting bigger models.
20:44 Joy Jiao: And I think as we went from GPT-2 to 3,
20:47 there was a huge size increase.
20:49 Joy Jiao: And there were just these amazing emergent properties from the model.
20:53 Joy Jiao: I mean, thinking about when GPT-2 was released,
20:56 we were all kind of collectively amazed that Joy Jiao:
20:59 it was able to write a coherent article about unicorns.
21:02 Joy Jiao: And now we're in a completely different world, right?
21:04 Joy Jiao: And a lot of that is driven by model architecture,
21:07 yes, but also just the number of parameters Joy Jiao:
21:10 in the model just allows it to achieve
21:13 this incredible intelligence that we never thought Joy Jiao:
21:15 was possible before.
21:16 And then on the other axis, we have what we call test time compute scaling.
21:21 Joy Jiao: And this is when you are inferencing a model,
21:23 when it's kind of spitting out tokens.
21:25 Joy Jiao: And this is a thing that happened
21:27 fairly recently when we call these reasoning models,
21:29 Joy Jiao: is that it can think for a scalable amount of time.
21:33 And this is variable depending on how Joy Jiao:
21:36 difficult it thinks a problem is.
21:38 But we can have the model think for days where really there's kind Joy Jiao:
21:42 of ways to just kind of have it think forever about a problem.
21:45 And this allows it to kind of Joy Jiao:
21:47 reach new levels of difficulty and discovery
21:49 that we didn't think was even possible before.
21:52 Andrew Mayne: When we think about data centers,
21:54 we often just sort of think about it as generating cat pictures Andrew Mayne:
21:57 or doing text conversations.
21:58 But I think that's really the helpful framework to look at is that Andrew Mayne:
22:02 these are going to be systems for doing extremely long-term,
22:06 big, complex processes of thinking Andrew Mayne: about this.
22:09 And to me, it just makes a lot more sense when projects like Stargate saying,
22:14 Andrew Mayne: we're going to be building a lot of compute.
22:15 It's not just for what we're doing now,
22:16 but it's going to be for things like that.
22:17 Joy Jiao: When we had first announced the Teams formation on Slack, Joy Jiao:
22:23 I think one of the taglines was to scale test time compute to cure all disease.
22:27 So that is Joy Jiao: It's like our team tagline.
22:29 Yunyun Wang: It's our team motto.
22:31 Andrew Mayne: That's ambitious.
22:32 Joy Jiao: Yeah.
22:32 Andrew Mayne: I had a friend whose child
22:34 was born with one of those orphan diseases,
22:36 and she would do Andrew Mayne: fundraisers,
22:38 do everything she could to try to support.
22:40 Andrew Mayne: Some researchers were trying to find
22:41 a cure for this, but there's just not enough time,
22:43 Andrew Mayne: not enough people.
22:44 Andrew Mayne: And I'm hopeful that we're kind of in an age
22:47 now where these kinds of tools are going to Andrew Mayne:
22:49 make that maybe a thing of the past.
22:51 Joy Jiao: Yeah.
22:52 Joy Jiao: I think we're already seeing the model help a lot in these cases.
22:56 Joy Jiao: I think from things like drug repurposing.
22:59 Joy Jiao: So for example,
23:00 a drug that's already been cleared by the FDA Joy Jiao:
23:02 for use in one different indication, Joy Jiao:
23:05 but from mechanistic understandings of how that drug works, Joy Jiao:
23:10 the model has suggested in many different cases Joy Jiao:
23:12 for maybe you can use this drug to temporarily ameliorate symptoms.
23:17 Joy Jiao: We're also seeing a lot of advances in personalized medicine.
23:20 Joy Jiao: So for example,
23:21 the design of ASOs or other RNA-based treatments is very common.
23:26 Joy Jiao: And I think, yeah, we are actually very,
23:30 very close to being able to scale this up in a really vast way with AI.
23:37 Joy Jiao: I think just in the next year or two,
23:39 I think we'll see very big changes here.
23:41 Andrew Mayne: Every researcher I know,
23:43 when you ask them what they could use in their lab,
23:45 they always say more hands, more people, more people doing this kind of work.
23:49 Andrew Mayne: And you hear some people talk about,
23:51 well, is AI going to displace that?
23:53 Andrew Mayne: And I think, no,
23:55 it sounds like it's just a big accelerator Andrew Mayne:
23:56 for all the things that could be done.
23:58 Joy Jiao: Yeah, I completely agree.
24:00 Joy Jiao: I feel like when you think about lab automation,
24:04 for example, Joy Jiao: a lot of the bottleneck comes from actually
24:07 being able to translate a protocol Joy Jiao:
24:10 into something that can be run on the platform.
24:12 Joy Jiao: And we've had partners tell us
24:14 about how Codex has been helping them do this.
24:16 Joy Jiao: And this is kind of fundamentally a half coding problem,
24:19 Joy Jiao: half understanding how a lab works.
24:21 Joy Jiao: And then I think thinking about the data analysis piece,
24:25 I feel like having our models kind of walk through
24:28 a user who maybe doesn't have the deepest understanding of statistics.
24:33 Joy Jiao: They can still rigorously analyze the data that's coming in.
24:36 Joy Jiao: The model can kind of help them
24:38 probe different hypotheses or it can suggest different statistical tests.
24:41 Joy Jiao: It can point out potential issues and biases in the data.
24:44 Joy Jiao: I think these are all ways of kind
24:46 of uplifting individual scientists and helping them do better science.
24:49 Joy Jiao: But I don't think we can
24:51 ever fully replace the scientists in the loop.
24:54 Andrew Mayne: So you've been putting it into the lab.
24:56 Andrew Mayne: You've figured out how to help with automation.
24:59 Andrew Mayne: Where do you think we're going
25:00 to be six months from now, 12 months from now?
25:02 Joy Jiao: Well, I would really love to get to the point where
25:06 we can say that AI has designed a new drug or cured a disease.
25:10 Joy Jiao: I don't know if that can happen in six months,
25:12 but I will hope in the next few years that's going to happen.
25:15 Joy Jiao: I think we're seeing signs of this happening
25:16 kind of all over different stages of the pipeline.
25:19 Joy Jiao: I think obviously earlier
25:20 in the drug discovery process where you're kind
25:23 of looking at literature synthesis or the model
25:25 is kind of discovering new biology.
25:28 Joy Jiao: For that to become a new drug
25:30 on the market is going to be a very long process, possibly like a decade.
25:34 Joy Jiao: But I think there's ways that we can really
25:37 speed up this process by starting at maybe the clinical trial stage.
25:39 Joy Jiao: We're starting a little bit before then
25:42 in the safety reviews or in the drug design phase.
25:45 Joy Jiao: So I think, yeah, basically that's what I'm the most excited
25:47 about coming up in the next few years.
25:49 Yunyun Wang: Yeah, for me,
25:51 I think I'm most excited about all the possibilities Yunyun Wang:
25:55 that our users, our scientists can do on our platforms.
26:00 So for one, I think a huge win would Yunyun Wang:
26:04 be if a researcher can patent a new finding
26:07 or a new discovery on our platform and using our models.
26:11 Yunyun Wang: And that's why we really
26:12 focus on early discovery and starting with building,
26:15 like teaching the Yunyun Wang: models, like the mechanistic understanding.
26:17 So this is, again, like trying to provide the most Yunyun Wang:
26:21 powerful tools through our life sciences models to these scientists
26:23 so they can really accelerate the speed of their research.
26:26 Andrew Mayne: Do you think we'll get to a point
26:28 where the models are going to get Andrew Mayne:
26:31 really good at basically predicting the cell or predicting the outcome?
26:37 Joy Jiao: I think definitely, yes.
26:38 I think it depends on the complexity of a system.
26:40 So for example, one thing our models are already Joy Jiao:
26:42 very good at is predicting the outcome of a chemical reaction.
26:45 And I think as you increase Joy Jiao: in biochemical and biological complexity,
26:50 some of the hardest things to predict is given a drug,
26:55 Joy Jiao: will this be toxic to a specific person or to a specific system?
26:59 And I want to slowly work our way Joy Jiao:
27:02 up to that, but that is definitely on a roadmap.
27:03 That's something we want to do eventually.
27:05 Andrew Mayne: When we're looking at models that do things like language or math,
27:09 it's pretty easy to Andrew Mayne: put together evals for it?
27:11 Did it get the problem right or get it wrong?
27:12 What do evals look like for models that are doing biology?
27:14 Joy Jiao: Yeah, we have various different
27:18 ways of evaluating model Joy Jiao: performance.
27:21 A really nice way to do this is kind of with experimental data.
27:26 So someone has already Joy Jiao: done the experiment and then you ask the model,
27:29 can it kind of predict the outcome of these Joy Jiao: experiments?
27:32 So a lot of the kind of virtual cell work, basically it looks like this, right?
27:39 So Joy Jiao: model and then you try to get it to predict a unseen perturbation.
27:44 We can also do a lot with Joy Jiao: synthetic data.
27:48 And this means that maybe you have generated a set
27:51 of data and you put very specific Joy Jiao:
27:54 characteristics in this data that could be kind of like foot guns for the model.
27:58 Joy Jiao: And these are things that maybe a typical
28:01 computational biologist might encounter day to Joy Jiao: day.
28:04 So this could be some weird bias in the data.
28:06 It could be some QC thing that you have to do or Joy Jiao:
28:08 statistical correction.
28:09 And because we generated the data ourselves,
28:11 then we can actually go and Joy Jiao:
28:14 test the model's capability as a computational biologist
28:16 that doesn't catch all of these Joy Jiao: different mistakes.
28:18 So there's a lot of different ways you can be creative with evaluation.
28:22 Joy Jiao: But that being said,
28:24 I think YLAB is still kind of the final real evaluation of the model, right?
28:27 Joy Jiao: And as you like to say, nothing in biology is really real until you
28:31 can prove it in the real Joy Jiao: world.
28:33 And so we do have a lot of research collaborations where we try to do just that.
28:37 Yunyun Wang: Yeah, evals have really
28:38 become more complex and sophisticated over time.
28:40 Yunyun Wang: And I think that's especially true
28:43 for designing evals that can really capture both value creation,
28:46 but solving complex problems for life sciences.
28:48 Yunyun Wang: So I think we really try
28:50 to focus on examples that are not like toy problems, but really capture that.
28:53 Yunyun Wang: Like, for example,
28:55 like the the messiness of like pre-processed site data.
28:58 Yunyun Wang: And when we design these new evaluations,
29:01 a starting point is often just trying to recreate an existing experiment.
29:05 Yunyun Wang: So something that has already a baseline.
29:08 So we already know what the either current state
29:10 of the art looks like or the current ground truth looks like.
29:15 Yunyun Wang: So a evaluation I'm really excited
29:18 about is looking at if our models can assess
29:22 like the antibody binding predictions and looking at how
29:25 that's been done for an existing virus variant.
29:30 Yunyun Wang: And then once we have already done that baseline,
29:32 we can push forward and say,
29:33 can we do this with something that hasn't been done before?
29:37 Yunyun Wang: And I think that is like some
29:40 of the precursor steps to de novo antibody design,
29:43 maybe expanding the neutralization for new viral variants.
29:46 Yunyun Wang: And that's also on the path
29:48 to new treatments and potentially developing new vaccines.
29:51 Andrew Mayne: What has been the reception in the life sciences,
29:54 particularly at conferences in the community, people you know?
29:57 Andrew Mayne: Have you seen a lot of willingness to embrace
30:00 this or skepticism or people who just don't think it's helpful?
30:04 Joy Jiao: I think it probably depends on what part of the country you're in.
30:08 Joy Jiao: I feel like kind of being on the West Coast,
30:13 everyone is pretty AI-pilled.
30:14 Joy Jiao: And so they really embrace this AI scientist,
30:18 the agentic workflows, and they really kind of see the future for AI.
30:23 Joy Jiao: When I'm at a conference on the East Coast, this changes a lot.
30:27 Joy Jiao: I think people are generally a bit more skeptical.
30:29 Maybe there's a little bit more Joy Jiao: doubt around the AI capabilities.
30:33 And yeah, I think it's just maybe like a cultural difference.
30:37 Joy Jiao: I think most of the big AI labs are here.
30:39 And so we kind of have a firsthand experience of what the Joy Jiao:
30:42 models are capable of.
30:43 And this kind of changes our perspective a little bit.
30:46 Andrew Mayne: How do you bridge that gap?
30:48 How do you get more scientists to understand?
30:50 Because it sounds like Andrew Mayne: the more people contributing, the better,
30:53 because there are weaknesses or areas need to be improved Andrew Mayne:
30:55 upon and the more you get people who are maybe
30:57 skeptical about this to sort of figure out how to participate?
30:59 Joy Jiao: Yeah, I think there's a few different ways.
31:02 The easiest way is by launching our models Joy Jiao:
31:05 through different platforms like Chat or Codex.
31:08 And I think just by kind of showing individual Joy Jiao:
31:11 scientists how useful this could be,
31:13 maybe just making a serial dilution spreadsheet for someone Joy Jiao:
31:15 who's pipetting, but that has real value, right?
31:17 And I think you can kind of build up from there.
31:20 Joy Jiao: I think coming from the other end,
31:22 we do have these more deep research collaborations with labs Joy Jiao:
31:26 for example, like antibody design or enzyme design.
31:28 And these sort of things are kind of more, Joy Jiao: you know,
31:31 they result in publications and then people will read and say, okay,
31:33 you know, Joy Jiao: a AI system did a lot of work.
31:36 It has biological novelty.
31:37 It's been proven out in the wet lab.
31:39 And Joy Jiao: so I think that also lends credibility to the system.
31:43 Yeah.
31:43 I think the simple answer is you Yunyun Wang: show by doing and you show
31:47 by publishing and engaging with the scientific community.
31:50 Yunyun Wang: And I think the skepticism
31:52 is really healthy and should be welcomed.
31:55 Yunyun Wang: I think it's just really great to see people
31:58 get really excited because and also trying to disprove maybe because
32:01 the potential for this technology is so great if we get
32:04 it right and if we can actually really leverage its full capability.
32:08 Yunyun Wang: So I feel like the carefulness about how do we
32:11 actually make this work for real problems is like very much warranted.
32:16 Yunyun Wang: But yeah, I think when we publish,
32:19 and I think that just also shows a need for more rigorous Yunyun Wang:
32:23 evaluations that represent these life science workflows
32:26 and research problems so people can Yunyun Wang: look at and eval and say, yes,
32:30 I feel like now I have 100 different ideas for how I can Yunyun Wang:
32:34 implement this into my lab and solve some of the current bottlenecks I'm facing.
32:39 Joy Jiao: I actually think there's a certain amount
32:42 of stress I've encountered from people who Joy Jiao:
32:45 are worried that, you know, AI is really powerful,
32:47 but they don't know how to use it the right way.
32:49 Joy Jiao: And so there is this general feeling of like,
32:51 I need more AI in my workflow,
32:53 in my life, but they Joy Jiao: don't know where AI should come in.
32:56 And I think part of the product vision is to just make it so Joy Jiao:
33:00 simple that it just works.
33:01 So you can just go to something like Codex and say,
33:03 hey, I want to do Joy Jiao: whatever I'm doing today.
33:05 And then Codex can figure out all the different pieces,
33:07 Joy Jiao: the multi-agent workflows, the tool calling, all of that.
33:10 And so, yeah, basically you don't have to Joy Jiao:
33:13 stress about how to get uplift from AI, and it just happens naturally.
33:17 Andrew Mayne: We do see those step changes every time
33:21 these models become smarter and understand users Andrew Mayne: better.
33:24 Andrew Mayne: You get more utility because some people go,
33:26 I don't have to spend a lot of time trying Andrew Mayne:
33:27 to prompt it or figure out all the tricks to it.
33:30 Andrew Mayne: If you were talking to somebody
33:32 who was considering getting into the life sciences,
33:34 maybe a high Andrew Mayne: school student right now,
33:36 what advice would you give them?
33:37 Joy Jiao: I feel like when I was in high school,
33:41 so I did the USA Biology Olympiad back when I was a high school student.
33:46 Joy Jiao: And I think out of all the different Olympias,
33:50 I think biology was seen as kind of the most
33:52 like memorization heavy one versus like math, right?
33:55 Joy Jiao: Where it's kind of more, I don't know,
33:58 test time compute scaling, whereas biology is more kind of memory and retrieval.
34:01 Joy Jiao: I think my hope is that with AI having kind
34:06 of like learned all the relationships
34:09 between all the different research pieces is
34:12 that it can really uplift human creativity and just make the process less
34:17 memorization and more kind of helping
34:20 people connect different fields of research together.
34:24 Joy Jiao: And just kind of, I guess,
34:26 furthering the frontiers of what people are able to explore in biology.
34:32 Joy Jiao: So, yeah, I feel like my advice to, I guess,
34:36 a high school student would be that maybe you don't
34:39 have to kind of go and memorize all the biology books.
34:42 Joy Jiao: You should just do more exploration with AI.
34:45 Joy Jiao: I think you can definitely read papers and just ask questions.
34:50 Joy Jiao: And I think you can do
34:52 both deeper dives and broader overviews this way.
34:55 Joy Jiao: And I think just the way of learning really changes.
34:59 Yunyun Wang: I found that when I was in the lab,
35:02 there was like a real like solo,
35:04 like individual aspect of doing biology research compared to like,
35:07 for example, when I went to my first like CS hackathon,
35:11 there was some excitement about just like the collaborative
35:13 nature when we first like built our app together.
35:17 Yunyun Wang: So I feel like that's really the future
35:20 I hope to see for early adopters and students
35:22 using our models and maybe using it in the Codex
35:25 runtime because there is a collaborative nature to it, too.
35:29 Yunyun Wang: I think, for example,
35:31 sending your scripts or sending your conversations or maybe
35:34 one day we all have our own co-scientists
35:36 or agent and we can deploy our agent to now work with a teammate in that way.
35:42 Yunyun Wang: I think there's just like
35:43 new like interactions and new modalities for us.
35:45 Yunyun Wang: So I would just encourage students to adopt early and just
35:48 to like also pioneer their own path for how they would like to use it.
35:53 Yunyun Wang: For me personally,
35:54 I always actually kind of felt like I got into wet lab a little bit too early.
35:58 Yunyun Wang: And like we mentioned earlier, I did not enjoy pipetting.
36:02 Andrew Mayne: That's the theme here.
36:03 Andrew Mayne: Nobody likes pipetting.
36:04 Yunyun Wang: Yeah, there's a lot of like very intense manual tasks involved.
36:08 Yunyun Wang: And so I hope that like, you know,
36:12 when our AI models can connect with physical devices that, yeah,
36:16 we can just like make a lot of like the learning
36:19 curve more fun for students so that they can kind
36:23 of like learn with the models and then kind of like
36:27 maximize their time with like the really interesting interactions spaces.
36:31 Andrew Mayne: So I've been working with a student.
36:33 I like to help students come with projects.
36:35 And one of them is we've taken Codex
36:37 and he's connected to a greenhouse and basically
36:39 using it to get photos back and to look at it and to evaluate it.
36:42 And I think it's been fun to see how he's
36:45 been taking both AI technology and then something traditional like
36:47 a greenhouse and combining them two and basically building up
36:49 the skill set of learning how to use the two of them.
36:53 Andrew Mayne: When you talk to your peers,
36:55 you talk to people who are running labs or running experiments,
36:59 researchers, what advice do you have for them?
37:01 Because the problem I see is that a lot of them go, that's great.
37:05 I just don't have the time.
37:07 But ultimately, what we're trying to do is save them time.
37:10 So do you have any kind of quick advice that you
37:13 give them or any ways you try to maybe inspire them?
37:16 Joy Jiao: Most people that I know,
37:19 I think in academia use AI in, I think, two main ways that I've seen.
37:27 Joy Jiao: One is to kind of talk to AI
37:30 about an existing piece of research paper or something
37:33 and just kind of make sure that you're understanding
37:35 things the right way or kind of fact checking.
37:38 Joy Jiao: And this is personally what I really like to use AI
37:41 for because you can ask really dumb questions and you don't feel any judgment.
37:45 Joy Jiao: actually just really wonderful for learning.
37:47 And I think people use a lot for analyzing Joy Jiao: experimental results.
37:51 And I think this comes back to the statistics piece that I learned, Joy Jiao:
37:56 what I mentioned before,
37:57 where sometimes you don't know what the right way to analyze your data is,
38:01 Joy Jiao: or there's just kind of so many
38:05 different interdisciplinary fields that your data might Joy Jiao:
38:08 touch on something in chemistry or something in like
38:11 a random niche field of like protein biology.
38:14 Joy Jiao: And the really nice thing is that a model can kind of like pull
38:18 those different ways of data analysis in for you
38:20 and kind of explore all of these different paths.
38:24 Joy Jiao: I feel like both of those are pretty low lift ways to try things out.
38:28 Joy Jiao: So you could just kind of like
38:30 throw a PDF file at AI and just be like,
38:32 hey, help me understand this paper and just have a natural conversation.
38:35 Joy Jiao: Or you can boot up Codex
38:37 and do some data analysis directly on your laptop.
38:39 Yunyun Wang: Yeah, I would say that you'd have to start
38:43 with making sure it doesn't feel like work right away.
38:46 Yunyun Wang: So maybe it'll be easier when you're focusing on AI adoption
38:49 to just like work on like a hobby project or a passion project.
38:53 Yunyun Wang: For me, for example,
38:55 I actually started working on like more like literature synthesis
38:59 tasks when I was doing creative writing projects, Yunyun Wang:
39:02 which are kind of like just something that was
39:05 not at all related to like our day to day,
39:08 Yunyun Wang: even though biology is a very creative space.
39:11 Yunyun Wang: I was just exploring that through a different, different medium.
39:13 Yunyun Wang: And I think that's actually when I started unlocking like a lot
39:17 of different ways to either prompt the model
39:18 or to actually access different data sources.
39:20 Yunyun Wang: So I think that just gave me a lot
39:23 of like pattern matching abilities for when I was trying to apply it,
39:26 Yunyun Wang: because we're not going to get it right in the first try.
39:28 Yunyun Wang: And it is really hard.
39:29 Yunyun Wang: And I feel like the progress and pace
39:32 of this field moves so fast that every week or month,
39:35 Yunyun Wang: there is a new pretty exciting development that might
39:39 change how we engage with models Yunyun Wang: or AI systems.
39:42 Yunyun Wang: So I think it's just important to get started somewhere.
39:45 Yunyun Wang: And I think another theme is the collaboration element.
39:49 Yunyun Wang: I feel like it's more powerful when you
39:51 have a recommendation from either somebody on your Yunyun Wang:
39:54 direct team who is doing the same day-to-day tasks as you.
39:57 Yunyun Wang: That happens a lot on our team as well,
40:00 where somebody will say, oh, I got Codex to like now touch these three different
40:04 like internal like databases that we weren't able to connect before.
40:07 Yunyun Wang: And I don't even like the latent space,
40:10 the latent capabilities are just so vast that there's a lot
40:13 that we just don't know until again, we can do it.
40:16 Yunyun Wang: So I think just having conversations with your friends,
40:18 your lab mates, your teammates will, I think,
40:20 spark a lot of those conversations, a lot of those creative juices and then
40:24 help you help you with your own adoption.
40:27 Andrew Mayne: What does science look like 10 years from now?
40:30 Yunyun Wang: I think when we started this team,
40:33 we do have like really just ambitious targets.
40:37 Yunyun Wang: And one of those is like,
40:39 I think we do want to make meaningful strides towards or even Yunyun Wang:
40:42 if like assist with like curing a disease.
40:45 Yunyun Wang: And I think there's just so many rare
40:47 like orphan diseases that doesn't really have the Yunyun Wang:
40:50 attention and the resources that it warrants because it's
40:54 just such a difficult field to Yunyun Wang: actually,
40:58 for example, clinical research is so difficult to actually
41:00 bring that to patients Yunyun Wang: and to market.
41:03 Yunyun Wang: So while 10 years, I feel like it's just a really long timeline,
41:09 I'm really excited about Yunyun Wang: some of the progress that we can make.
41:12 Yunyun Wang: And I think it's good to be carefully
41:14 optimistic that we're going to see some of those Yunyun Wang:
41:17 breakthroughs pretty soon.
41:18 Joy Jiao: Yeah, I think maybe this is a bit of a sci-fi
41:21 vision that I have of the world that I really hope becomes reality,
41:24 which is that you have these autonomous labs.
41:27 Joy Jiao: They are just mostly robots and you have them all hooked up to AI
41:32 and you just have autonomous research institutes
41:34 that are constantly running and curing human disease.
41:37 Joy Jiao: It's maybe making new materials, making new drugs.
41:42 Joy Jiao: It's maybe solving personalized medicine.
41:44 Joy Jiao: There's a lot of NF1 or just ultra rare diseases where people
41:49 without vast monetary and research scientific resources
41:52 can even begin to think about solving.
41:56 Joy Jiao: But we can solve that with AI.
41:58 Joy Jiao: And I think we can kind of almost break
42:02 through the financial and regulatory and monetary constraints with the system.
42:07 Joy Jiao: So I think that that's kind of the dream.
42:08 Joy Jiao: And I think also even separately thinking
42:10 kind of more about the biosecurity side of things.
42:14 Joy Jiao: these systems can be
42:15 kind of constantly sampling our environment, right?
42:17 It can be sampling Joy Jiao: wastewater.
42:19 It can be sampling the air and constantly
42:22 detecting potential threats or even Joy Jiao:
42:24 just better predictions for the flu and getting better flu vaccines.
42:28 But just generally, these Joy Jiao: different medical countermeasures,
42:30 I think should be happening autonomously in 10 years.
42:32 And I think that that's basically something, yeah, I'm really excited about.
42:35 Andrew Mayne: The AI lab is exciting because I Andrew Mayne:
42:40 think if people really understand what it means is it's not,
42:43 there aren't scientists, it's there Andrew Mayne: are more scientists,
42:45 but they sit at home and they go into Codex and say,
42:47 can you go run this Andrew Mayne: experiment for me?
42:49 And like, you have a data center, you have a science center doing that.
42:52 Joy Jiao: Right, exactly.
42:53 Yeah.
42:53 And I think I didn't talk about the scientists
42:55 in the vision I was just Joy Jiao:
42:58 describing, but obviously there are people involved in here.
43:00 And I think it's really kind of Joy Jiao:
43:03 high level direction setting from the humans.
43:05 We're saying, here's a patient with this disease.
43:09 Joy Jiao: here are some potential solutions
43:11 or things that maybe you can look at.
43:13 Joy Jiao: And I think that AI can then go off and explore different ideas.
43:17 Joy Jiao: It can design experiments and then come back to the humans and say,
43:19 Joy Jiao: here's what I found.
43:20 Joy Jiao: What do you think we should do next?
43:23 Joy Jiao: And this can be kind of an academic discussion.
43:27 Joy Jiao: It's a little bit similar to kind
43:29 of the way that people interact with codex today,
43:31 Joy Jiao: where you say, here, go write a function or go write a piece of code.
43:34 Joy Jiao: And it writes it and says, here's the code.
43:36 Joy Jiao: And then the person tells the next thing to do.
43:37 Joy Jiao: So I think it's a little bit similar to that kind of interaction,
43:41 but on a much grander scale and on a much longer time horizon.
43:44 Yunyun Wang: I think it's really like
43:47 the democratizing science aspect and putting like
43:49 really capable expert level knowledge in the hands
43:52 of a greater amount of people.
43:54 Yunyun Wang: And I think what that can mean for personalized medicine,
43:57 for bolstering our societal defenses.
44:00 Yunyun Wang: There's just like so many
44:03 naturally occurring new like variants every year, new like influenza strains.
44:08 Yunyun Wang: So I think it's really just like securing defenses
44:10 and feeling like we actually have more agency to counter all that.
44:14 Yunyun Wang: And I think I'm really excited about
44:16 a lot of like the medical countermeasure acceleration work as well.
44:20 Andrew Mayne: Well, it's very exciting.
44:21 Thank you for sharing this with us.
44:23 Yunyun Wang: Thank you for having us.
44:24 Joy Jiao: Yeah, thank you so much.