EVERY SPOKEN WORD
45 min read · 8,704 words- 0:00 – 0:39
Intro
- SPSpeaker
Hello, I'm Andrew Main, and this is the OpenAI Podcast. On today's episode, we're talking with Research Lead Joy Jiao, and Product Lead Yunyun Wang about OpenAI for Life Sciences. We'll explore what new models are making possible in biology and medicine, and what it takes to deploy the most advanced capabilities responsibly.
- JJJoy Jiao
This allows it to kind of reach new levels of difficulty and discovery that we didn't think was even possible before.
- YWYunyun Wang
Putting like really capable expert level knowledge in the hands of a greater amount of people.
- JJJoy Jiao
One of the taglines was to scale test time compute to cure all disease.
- SPSpeaker
Yes.
- JJJoy Jiao
So that is like [chuckles] our team tagline.
- 0:39 – 3:47
Introducing the Life Sciences model series
- SPSpeaker
We started off with just a basic API, and then we had ChatGPT, which is more conversational, was really good for text, as code became a capability, went through basically code models and then Codex. Now that you're getting more scientists in the life sciences working on these systems, does that mean things have to evolve to help with the way researchers might work with these tools?
- YWYunyun Wang
Yeah. We're really excited to, uh, build and deploy the Life Sciences model series. So this is a new biochemistry focused model series that's really anchored on these very complex life science research workflows, and we're focused on adding, um, new like mechanistic understanding, uh, starting with genomics understanding and protein understanding, and really focused on early discovery use cases because we feel like that's like one of the core bottlenecks that, uh, greater thinking time, greater compute, and really like leveraging like more capable AI models can help, um, meaningfully scale some of these like research barriers. And I think there's also like a, a model orchestration piece of actually how to embed this into workflows, and it's been really great, uh, first off, w- having all these different product surfaces to deploy to. We're seeing a lot of really great like literature synthesis workflows happening on, uh, ChatGPT, and, uh, these models really push the frontier of like long trajectory agentic workflows, and we're really able to empower that on, uh, Codex. And more on the model orchestration piece is that I think for enterprise use cases, there's like this reproducibility and repeatability element, and we are trying to overcome this by working on like some of the life sciences research plugins that we're shipping for very specific translational, uh, bio users. So the life sciences research plugin has over 50 skills, which are essentially templatized repeatable workflows that, um, if you need to whether do some sort of cross evidence match and search across various different papers or do, uh, pathway analysis, something that's like, uh, repeatable that you often do, you can have like almost like a one click deploy option by using our life sciences plugins on top, and that's also how we're kind of seeing, uh, the balance between, uh, scaling for very specialized, uh, purposes, uh, something we're hoping to get into is maybe clinical purposes, but also make it still very general use for all foundational biology.
- JJJoy Jiao
I think the models can get quite far by using tools. So for example, we can use open source protein structure prediction algorithms-
- SPSpeaker
Mm-hmm
- JJJoy Jiao
... inside our research stack, and in this case the model is acting kind of like a regular computational biologist.
- SPSpeaker
Mm-hmm.
- JJJoy Jiao
You will kind of go run these tools on a computer, you will look at the output, you will tweak the input a little bit. Um, so I think that is something our models can already do. I do think what will make the models even more powerful is to s- start to turn them more into kind of a biochemistry expert.
- SPSpeaker
Mm-hmm.
- JJJoy Jiao
And I think with this kind of intuition and expertise, you can use these tools, um, even more intelligently and get at the right answer more quickly.
- 3:47 – 5:00
Joy’s path into life sciences
- SPSpeaker
How did you get your interest in life sciences?
- JJJoy Jiao
My, I guess, original background was actually in life sciences. Um, so I've always been interested in biology as a kid. Um, I got my PhD in systems biology around like a decade ago from Harvard. Um, found academia to be very interesting, but the pace was a little bit more slow-moving than I would have liked. Um, and I think just the, the experience of kind of like having to physically be in the lab and kind of like transferring small amounts of liquid from like one tube to another, I think I wanted something a little bit faster paced where I felt like I was more in direct control of like my own velocity. Um, so I went from that to software, um, and I ended up here at OpenAI. And so this is kind of like a full circle moment-
- SPSpeaker
Yeah
- JJJoy Jiao
... for me where I'm like starting to look at biology again and looking at how to accelerate my previous self [chuckles] with AI. So yeah, really excited to see what, um, progress AI can make in this space.
- SPSpeaker
So you're like, "Yeah, this is too slow. Let me go off on AI and speed it up so I can get back into it."
- JJJoy Jiao
Yeah, except, you know, from this end, I, I don't really ever want to like touch a pipette or anything again.
- SPSpeaker
Right.
- JJJoy Jiao
So I, I would prefer for like robots to do it for me. [chuckles]
- YWYunyun Wang
Yeah, we joke about that a lot. A lot of our motivation for this is we can automate pipetting and never have to do that again.
- 5:00 – 7:27
Autonomous lab with Ginkgo Bioworks
- YWYunyun Wang
[chuckles]
- SPSpeaker
Well, that's what's interesting is I was looking at what you all did with, uh, Ginkgo Bioworks, and the idea of taking GPT-5 and taking an AI system and then working with a robotic lab, and how it was able to speed things up. Could you tell us a little about that?
- JJJoy Jiao
Yeah. Um, the Ginkgo work is interesting because I think when it started, I think it was like through July of last year-
- SPSpeaker
Mm-hmm
- JJJoy Jiao
... uh, 2025, and at that point, GPT-5 had just finished training. We were really not sure if the models could do any kind of biology. We didn't really have that much biology in our training data. It was mostly math and computer science, which I think makes sense because these things have verifiable solutions, and this is usually not the case in biology, unless you have... you can go and do the experiment in a lab, right? Um, so when we started the collaboration with Ginkgo, it was really can the model do any biology at all? Can it design experiments that-Actually make, um, reactants, like make-
- SPSpeaker
Mm-hmm
- JJJoy Jiao
... the product that we want. So it was actually quite surprising, I think, when GPT-5 designed the first set of experiments with Ginkgo, and the results came back and we're, "Oh, we made a non-zero amount of protein," and that was actually quite surprising. And then I think progressing from that point in time, which is just roughly like six months ago to now-
- SPSpeaker
Mm-hmm
- JJJoy Jiao
... where it actually just feels quite obvious that our models can accelerate science is actually just really surprising.
- YWYunyun Wang
And it, it really shows, like, the art of the possible, I think. I think before that experiment led by Joy and the, the bio- uh, Ginkgo team was conducted, I think we really didn't know, like, for ourselves, and I always say, like, we kind of learn that for ourselves when we, like, engage in these experiments, and we have a few more, um, in the, in the works with others. And I think that is, like, the type of, like, acceleration that we're looking for. Um, ingesting high throughput experimental data is really difficult. It's very compute intensive.
- SPSpeaker
Mm-hmm.
- YWYunyun Wang
And I think for a lot of these scientific workflows, like, the true bottleneck for, for the speed and progress of scientific acceleration is at, like, um, how... Like, almost a human bottlenecks.
- SPSpeaker
Mm-hmm.
- YWYunyun Wang
And I think the future that me and Joy see is that in- it's no longer, um, human bottlenecks, but rather maybe compute bottlenecks, and we're really able to, like, deploy many sub-agents doing parallel orchestration to, uh, divide and conquer all these tasks, and the researcher can now spend their time on, like, really analyzing, interpreting, like, the most meaningful insights coming out of that.
- 7:27 – 8:12
Yunyun’s path into life sciences
- SPSpeaker
So Yunyun, how did you get into this?
- YWYunyun Wang
Yeah. I think reflecting back, I've actually been working on, like, biology research in some shape or form for a majority of my time here at OpenAI.
- SPSpeaker
Mm.
- YWYunyun Wang
So I first started on, uh, working on biorisk mitigations-
- SPSpeaker
Mm
- YWYunyun Wang
... and a lot of our biodefense initiatives, and so I feel like coming to now working on the life sciences research side gives me, like, just appreciation for how difficult this problem is and tackling it from both sides. And I've, uh... My initial entries point into wet lab research was actually through, um, doing a lot of infectious disease and virology work. So I think I've always gotten, like, the interest in biosecurity-
- SPSpeaker
Mm
- YWYunyun Wang
... in that way, so this just feels like a really great moment, like, right now to work on it, especially when our models are getting more capable at, um, beneficial use and just general
- 8:12 – 9:48
OpenAI’s life sciences work
- YWYunyun Wang
life sciences.
- SPSpeaker
How long has OpenAI been focused on life sciences?
- YWYunyun Wang
Yeah. I would say, uh, it was really the way we design our capability evals that show us that this is possible, so it's been, I think, for, um, at least, like, two years now that we have worked on a lot of our early, um, research, research experiments, and now with the Ginkgo, um, autonomous, like, wet lab, uh, model-in-the-loop experiments.
- JJJoy Jiao
I think we, we have a few more research partners in the space that we're really excited about. I think I'm, I can't actually name everyone right now.
- SPSpeaker
Mm.
- JJJoy Jiao
But there's a lot of stuff kind of in the chemical design, protein design, enzyme design space that I think is very AI native and a lot of people are interested in. So understanding how the world works, um, understanding how chemicals react, understanding how cells interact, how pathways inside cells interact, all the way to can we accelerate drug discovery?
- SPSpeaker
Mm-hmm.
- JJJoy Jiao
So given a disease kind of model, how scientists understand the mechanism, can we, once given a target, actually find, design a drug against that target? Can we even accelerate the FDA approval process? So I think there's a role for AI to play kind of at every, uh, step of this pipeline, and yeah, I think just... I, I think there's a lot of AI possible in, in everything.
- SPSpeaker
I've been to some of those cutting-edge labs, and on the outside you have this impression of it, then you walk in there and you literally see somebody with, you know, a row of Petri dishes, a row of samples, and just some grad student going click, click, click, and I'm like, "Oh, this is the pace of science. This is as fast as-"
- YWYunyun Wang
That was me. [laughs]
- SPSpeaker
But yeah, yeah, exactly. Like, enough of this, I gotta go speed this up. But we forget that's often the pace of science is-
- YWYunyun Wang
Mm-hmm
- SPSpeaker
... just how fast the human hands can move through that.
- 9:48 – 15:43
Biorisk, access, and safeguards
- SPSpeaker
And with a tool like that, it's, it's kind of exciting. When you start using these tools to maybe think about, uh, new pathways for treatments or just evaluate, you also introduce the idea that these could be used for things that maybe are less desirable. You know, bioweapons is something that comes up a lot, the fact that if, you know, an AI can figure out how to do a code exploit, might be able to figure out how to do a gene exploit. How are you addressing that?
- YWYunyun Wang
Yeah, that's a great question, and I think it is just probably one of the most severe risks that we're currently really tracking for, uh, rising AI capabilities. Our first approach to that was really thinking about, um, how do we assess for information hazards? At what point does a model now, um, maybe give, like, the, the final step in, like, the synthesis of a dangerous, um, pathogen? And what we found is that, like, the precursor steps to that really looks very benign, and it's really hard to distinguish between.
- SPSpeaker
Mm.
- YWYunyun Wang
So another way to put it is, like, the same steps that a beneficial, like, a legitimate actor might take is, looks very similar to the, the ones that a, um, dangerous, harmful actor-
- SPSpeaker
You start with E. coli.
- YWYunyun Wang
Yes.
- SPSpeaker
You start with something that's-
- YWYunyun Wang
Yes.
- SPSpeaker
Yeah.
- YWYunyun Wang
Exactly. So I still think that we, we made the right call for really taking a very, um, risk-averse approach to that. But now I'm really excited about, like, uh, differentiate access and, like, responsible deployment as really a core pillar of all of our safeguards work, and really understanding that, um, there are different user segments, and I almost feel like the future we're going towards is, um, something like models as, like, professions, similar to how they, um, models have different personalities and sometimes you wanna invoke the right one depending on, um, the, the type of, like, um, workflow you're looking at. So I think how this translates is, um, similar to how biologists working on, like, therape- therapeutics and their research, um, they require access to, um, data sets are often b- uh, very tightly controlled, or they require access to just expert level, like, they all have PhDs and have, like, expert level, like, biology, like, knowledge.How does that compare to-- how does that translate over to, to models? I think that's why we have to kind of similarly take the same training approach, but also the same security approach, and deploying that in, like, a way where we can have those very heightened, uh, enterprise-grade controls in place.
- SPSpeaker
So you just mentioned safeguards. Can you explain how that applies here, where you would need them, why you would need them?
- YWYunyun Wang
Yeah. So we very thoughtfully design and, uh, design new safeguards for pretty much all of our models across very different risk areas. But I think when it came to bio, this was, like, the first dual use risk that is both also a capability risk, so it very much correlates with how we, um, as capabilities improve-
- SPSpeaker
Mm-hmm
- YWYunyun Wang
... the risk correlates. And I think that's why wh- when our first approach, wh-when we really there was no precedent for a lot of this work, and we were the first to really activate these high safeguards, when we saw that, that significant, uh, reasoning jump in our model capabilities, uh, we really wanted to make sure that we did it right, and I think the best way to get it right is to incrementally deploy.
- JJJoy Jiao
Yeah. I think it's really a fine line between having a very capable model that's capable of accelerating benign science and beneficial science versus a model that could be used by a bad actor. And I think the safest model here would be a model that just had no capability, right? And you-
- SPSpeaker
That's not very good [laughs] .
- JJJoy Jiao
Yeah. It w- it's not very good, but it's very safe. Um, and on the other hand, if you had a model that is basically an oracle of the physical world, it basically knows everything about every experiment, um, that model could fall into the wrong hands and do s- potentially very bad things, 'cause someone can go and say, "Hey, design a new novel pandemic potential pathogen," and the model can just go and do that autonomously. Um, so I think we need to kind of figure out where we draw the line in between the two and kind of think about who gets access to a potentially very capable model and who doesn't. And what we found in kind of, um, we call general access traffic, is that it's very difficult to figure out what a user's actual intentions are just from kind of reading a prompt. And I think as an example of this, let's say someone says, "Hey, help me clone a gene." The model might not even be given what the gene is, um, but it can come up with a protocol for it, and so this gene could just be something like green fluorescent protein-
- SPSpeaker
Mm-hmm
- JJJoy Jiao
... or it could be a toxin, and there's basically no way to figure that out from the context of the conversation. And so this becomes a very difficult problem in production. And basically, I think, like Yunyun said, we, um, decided to kind of err on the side of safety here, and basically say that, okay, if we think that there is a potential for misuse, we either s- have the model kind of self-refuse the user, in which case it, it tends to say things like, uh, "Sorry, I can't really help you with that, but I can give you a high-level overview of this protocol instead."
- SPSpeaker
Mm-hmm.
- JJJoy Jiao
And this, unfortunately, very, um, very much annoys our kind of professional scientists, understandably. Uh, and then we also kind of have multiple layers of mitigation on top of that. But I think really to unlock the full capabilities of our models, we need is this differentiated access. And what this means is we know who the user actually is. They are a professional working at a legitimate research institution or a, uh, pharma company. And w- because of, um, the regulations around these institutions, we know that, for example, all the reagents are being tracked, all the cell lines that they're using are being tracked. And so this gives us confidence that this is a legitimate user and not a random person in a basement doing who knows what. Um, and that ge- allows us to give them basically more capabilities than we are able to provide to the general access traffic.
- 15:43 – 17:51
What models can do in the lab
- SPSpeaker
What can you do right now if you're working with the models, you're working it within a laboratory, what would you say the capability is at this moment?
- JJJoy Jiao
So I think people use the models, um, for very different things.
- SPSpeaker
Mm-hmm.
- JJJoy Jiao
Um, I've talked to people in the Baker Lab recently on kind of, like, how they've been using our models on, like, codecs, and sometimes it's as simple as, "Hey, can you write a spreadsheet for me?"
- SPSpeaker
Mm-hmm.
- JJJoy Jiao
"I want to just minimize the number of pipetting steps that I have to make." And this hits me very hard, because I had done the same thing by hand in grad school. Um, so that, that's like a, a very simple just mathematical, um, software operation, and then there's much harder tasks, like, "Can you, um, design a enzyme for me with all these biological design tools?" So I think there's a huge range of sophistication.
- YWYunyun Wang
Yeah. And something I'm very excited about is how we can use our models to be a, uh, more powerful discriminator in, like, really testing and assessing, like, new novel ideas. And I think something that I've been noticing as a trend with a lot of our research partners and also, um, the users of our models, is that, uh, models for scientific research and tasks almost require a different, like, persona or a different prompting style.
- SPSpeaker
Mm-hmm.
- YWYunyun Wang
So we-- I often feel like, you know, like a model that is much more scrutinizing or skeptic at good ideas is it's very similar to how human scientists would go assess, like, originality and feasibility. It's really like, I think, uh, helping understand, like, out of all the ne- new papers and new publications out there that t- push the frontier of a lot of these hypotheses, what are the ones that are really feasible and valid for testing that's gonna help, you know, lead to new breakthroughs? So, and then translating this to something like, uh, disease target screening, selection, like, the potentials for these drug targets are endless, but it's really, like, narrowing down the aperture, and I feel like that's where, like, the assistance comes. Like, this is extremely difficult work to do at scale, and having a model that can, like, empower and accelerate that process, I think is kind of like one of the immediate impacts we're hoping to see by, by, uh, responsibly deploying this, this model to those,
- 17:51 – 20:14
Building scientific infrastructure
- YWYunyun Wang
those users.
- SPSpeaker
It seems like it's a very interesting trajectory. You went from there was, you had GPT-3 on the API and GPT 3.5, then you get ChatGPT, and now we have ChatGPT Apps, and now we have Codex, and it sounds like these things just, the number of things you can do with this continues to grow.How would you see this building? You know, do you see this as basically just becoming a complete infrastructure for kind of every kind of inquiry you might wanna pursue?
- JJJoy Jiao
Yeah, I think the dream is to have a lot of, uh, the basic foundations of scientific workflows happen on Codex. And I think that the goal is to have Codex to pretty much be able to do everything that is possible, uh, to do on the computer. Um, of course, we also want to extend beyond that with kind of hooking it up to robotics and so forth. But I think right now, um, we already do things, for example, if we have a bunch of different dev boxes on our remote, on our laptop, we can actually say, "Hey Codex, go and run this code on all of these different, uh, dev boxes that are all remote," and then, uh, Codex can do that, or can say, "Monitor this for me," and it kind of like go away and do something else, and the Codex is like there watching all the logs for you. Um, it can build a lot of just kind of fit for purpose software, uh, for analyzing specific pieces of data, for visualizing data. So, for example, if we have experimental biology data that we're sending each other on the team, what I've noticed recently is instead of sending the raw data, we've started sending HTML files-
- SPSpeaker
Mm.
- JJJoy Jiao
-or just these kind of like beautiful UIs that Codex has built with kind of like spinning proteins and it's actually just a really... It, it kind of changes the way that we share with each other and collaborate.
- YWYunyun Wang
Yeah, when we first started mapping out how, uh, users and organizations might adopt this, I think we envisioned that each, uh, scientist would get their personal assistant or their coworker, and this is a way that they can help scale their, their, um, their collective output. And then the next paradigm of that would be scaling up whole research institutions where a whole program team can actually deploy a, um, a, like, workforce of various agents, and they can all do, like, parallel task delegation, kind of mimicking a lot of these existing patterns, and we, we can figure out the pieces of like, um, how they can all collectively, like, work together to solve, like, larger, larger
- 20:14 – 24:54
Why compute matters for science
- YWYunyun Wang
tasks.
- SPSpeaker
It's interesting because OpenAI has talked about the need for compute, and I think that sometimes we just sort of think like, "Okay, so I can have more conversations and stuff." But when you're talking about the idea of building these tools to become entire platforms for scientific exploration, it sounds like the compute advantage is really critical.
- JJJoy Jiao
Yeah, I think there's two different axes we can think about how we are scaling compute. Um, the one that I think everyone's familiar with is just getting bigger models, and I think as we went from, you know, GPT II to III, there was a huge size increase, and there were just these amazing emergent properties from the model. I mean, thinking about, you know, when GPT II was released, we were all kind of collectively amazed that it was able to write a coherent article about unicorns.
- SPSpeaker
Uh-huh.
- JJJoy Jiao
And now we're in a completely different world, right? And a lot of that is driven by model architecture, yes, but also just the number of parameters in the model just allows it to achieve this incredible intelligence that we never thought was possible before. And then on the other axis, we have what we call test time compute scaling.
- SPSpeaker
Mm.
- JJJoy Jiao
And this is when you are inferencing a model, when it's kind of spitting out tokens. Um, and this is a, a thing that happened fairly recently when we call these reasoning models, is that it can think for a scalable amount of time, and this is variable depending on how difficult it think- it thinks a problem is. But we can have the model think for days, or really just kind of ways to just kind of have it think forever about a problem. And this allows it to kind of reach new levels of difficulty and discovery that we didn't think was even possible before.
- SPSpeaker
When we think about data centers, we often just sort of think about it as generating cat pictures or doing text conversations, but I think that's really the helpful framework to look at, is that these are going to be systems for doing extremely long-term, big, complex processes of thinking about this. And it's, it's... To me, it just makes a lot more sense when you, you know, projects like Stargate saying we're gonna be building a lot of compute. It's not just for what we're doing now, but it's gonna be for things like that.
- JJJoy Jiao
When we had first announced the team's formation-
- YWYunyun Wang
Mm
- JJJoy Jiao
... on Slack, I think one of the taglines was to scale test time compute to cure all disease.
- SPSpeaker
Yes.
- JJJoy Jiao
So that is like our team tagline. [chuckles]
- YWYunyun Wang
It's our team motto. [chuckles]
- SPSpeaker
That's ambitious.
- YWYunyun Wang
Yeah.
- SPSpeaker
I had a friend whose, uh, child was born with one of those orphan diseases, and she would do fundraisers, do everything she could to try to support, uh, some researchers who were trying to find a cure for this, but there just not enough time, not enough people. And, you know, I'm hopeful that we're kind of in an age now where these kinds of tools are going to make that maybe a thing of the past.
- JJJoy Jiao
Yeah. I think we're already seeing the model help a lot in these cases. Um, I think from things like drug repurposing.
- SPSpeaker
Mm.
- JJJoy Jiao
So, for example, a drug that's already been cleared by the FDA for use in one different indication, but it... for kind of like me- from mechanistic understandings of how that, uh, drug works, the model has suggested in many different cases for maybe you can use this drug to temporarily ameliorate symptoms.
- SPSpeaker
Mm.
- JJJoy Jiao
Um, we're also seeing a lot of advances in personalized medicine. So, for example, the design of ASO's or other RNA-based treatments-
- SPSpeaker
Mm
- JJJoy Jiao
... is very common. Um, and I think, yeah, we are actually very, very close to being able to scale this up, um, in, in a really vast way with AI. I think just in the next year or two, I think we'll see very big changes here.
- SPSpeaker
Every, every researcher I know, when you ask them what they could use in their lab, they always say more hands, more people, more people doing this kind of work. And you hear some people talk about, "Well, is AI going to displace that?" And I think, no, it sounds like it's just a big accelerator for all the things that could be done.
- JJJoy Jiao
Yeah, I completely agree. Um, I feel like when you think about lab automation, for example, a lot of the, um, bottleneck comes from actually being able to translate a protocol into something that can be run on the platform. And we've had partners tell us about how Codex has been helping them do this, and this is kind of fundamentally a half coding problem, half understanding how wet lab works. And then I think-Thinking about the data analysis piece, I feel like having our models kind of walk through a user who maybe doesn't underst- have the deepest understanding of statistics, um, they can still rigorously analyze the data that's coming in. The model can kind of, uh, help them probe different hypotheses, or it can suggest different statistical tests, it can point out potential issues and biases in the data. I think these are all ways of kind of uplifting individual scientists and helping them do better science. But I don't think we can ever fully replace the scientist in the loop.
- 24:54 – 29:51
Where are we in 6-12 months?
- SPSpeaker
So you've been putting it into the lab, you've figured out how to help with automation. Where do you think we're gonna be six months from now, 12 months from now?
- JJJoy Jiao
Well, I would really love to get to the point where we can say that AI has designed a new drug or cured-
- SPSpeaker
Mm-hmm
- JJJoy Jiao
... a disease. I don't know if that can happen in six months, um, but I would hope in the next few years that's going to happen. I think we're seeing signs of this happening kind of all over different stages of the pipeline. I think obviously earlier, um, in the drug discovery process, where you're kind of looking at literature synthesis or, um, the model is kind of discovering new biology, for that to become a new drug on the market is going to be a very long process-
- SPSpeaker
Mm
- JJJoy Jiao
... or possibly like a decade. But I think there's ways that we can really speed up this process by starting at maybe the clinical trial stage or starting a little bit before then, in, uh, the safety reviews or in the drug design phase. So I think, yeah, basically that's, that's what I'm the most excited about, uh, coming up in the next few years.
- YWYunyun Wang
Yeah, for me, I think I'm most excited about all the possibilities that our users, our scientists can, can do on our, um, on our platforms.
- SPSpeaker
Mm-hmm.
- YWYunyun Wang
So for one, I think a huge win would be if a researcher can patent a new finding or a new discovery on our platform, and using our models, and that's why we really focus on, like, early discovery, and starting with building, like, teaching the models, like, the mechanistic understanding. So this is, again, like, trying to provide the most powerful tools through our life sciences models to these scientists, so they can really accelerate the speed of their research.
- SPSpeaker
Do you think we'll get to a point where the models are gonna get really good at basically predicting the cell or predicting the outcome?
- JJJoy Jiao
I think definitely yes. I think it, it depends on the complexity of a system. So for example, one thing our models are already very good at is, uh, predicting the outcome of a chemical reaction.
- SPSpeaker
Mm-hmm.
- JJJoy Jiao
And I think as you increase in biochemical and biological complexity, some of the h- uh, hardest things to predict is, given a drug, will this be toxic to a specific person or to a specific system? And I want to slowly work our way up to that, but that is definitely on our roadmap. That's something we want to do eventually.
- SPSpeaker
When we're looking at models that do things like language or math, it's pretty easy to put together evals for it. Did it get the problem right or get it wrong? What do evals look like for models that are doing biology?
- JJJoy Jiao
Yeah, we have, uh, various different ways of evaluating model performance. A really, uh, nice way to do this is kind of with experimental data. So someone has already done the experiment, and then you ask the model, can it kind of predict the outcome of this experiment? So a lot of the kind of virtual cell work, basically it looks like this, right? So someone has done single cell RNA-seq on millions of different cells, and then you feed this to a model, and then you try to get it to pre- uh, predict a unseen perturbation. We can also do a lot with synthetic data, and this means that maybe you have generated a set of data, and you put very specific, uh, characteristics in this data that could be kind of like footnotes for the model. Um, and these are things that maybe a typical computational biologist might encounter day to day, so this could be some weird bias in the data, it could be some QC thing that you have to do, or statistical correction. And because we generated the data ourselves, then we can actually go and test the model's capability as a computational biologist, like, does it catch all of these different mistakes? So there's a lot of different ways we can be creative with evaluation. But that being said, I think wet lab is still kind of the, the final, like, real evaluation of the model, right? And I also like to say nothing in biology is really real until you can prove it in the real world. And so we do have a lot of research collaborations where we try to do just that.
- YWYunyun Wang
Yeah, evals have really become more complex and sophisticated over time, and I think that's especially true for designing evals that can really capture, uh, both value creation, but solving complex problems for life sciences. So I think we really try to focus on examples that are not like toy problems, but really capture that, like, for example, like the, the messiness of, like, preprocessed, like, data. And when we design these new, um, evaluations, a starting point is often just trying to recreate an existing experiment, so something that has already a baseline so we already know, uh, what the, um, either current state-of-the-art looks like or the current ground truth looks like. So a evaluation I'm really excited about is looking at if our models can, um, assess, like, the antibody binding pr- predictions and looking at how that's been done for an existing virus, uh, variant. And then once we have already done that baseline, we can push forward and say, "Can we do this with something that hasn't been done before?" And I think that is, like, some of the precursor steps to de novo antibody design, um, maybe expanding the, the neutralization, uh, for new viral variants, and that's also on the path to new treatments and potentially developing new vaccines.
- 29:51 – 33:17
Scientific adoption and skepticism
- SPSpeaker
What has been the reception in the life sciences, particularly at conferences, in the community, people you know? Have you seen a lot of, uh, willingness to embrace this or skepticism or people who just don't think it's helpful?
- JJJoy Jiao
I think it probably depends on what part of the country you're in.
- SPSpeaker
Hmm.
- JJJoy Jiao
Um, I feel like kind of being on the West Coast, everyone is pretty AI-pulled, and so they really embrace this AI scientists, AI agentic workflows, um, and they really kind of see the future for AI.When I'm at a conference on the East Coast, this changes a lot. I think people are generally a bit more skeptical. Maybe there's a, a little bit more, uh, doubt around the AI capabilities, and yeah, I think it's just maybe, like, a cultural difference. I think most of the big AI labs are here, and so we-
- SPSpeaker
Mm-hmm
- JJJoy Jiao
... kind of have a first-hand experience of what the models are capable of, and this kind of changes our perspective a little bit.
- SPSpeaker
How do you bridge that gap? How do you get more scientists to understand? 'Cause it sounds like the more people contributing, the better, because there are weaknesses or areas need to be improved upon, and the more you get people who are maybe skeptical about this to sort of figure out how to participate.
- JJJoy Jiao
Yeah, I think there's a few different ways. Um, the easiest way is by launching our models through different platforms, like Chat or Codex, and I think just by kind of showing individual scientists how useful this could be, maybe just making a serial dilution spreadsheet for someone who's-
- SPSpeaker
Mm
- JJJoy Jiao
... pipetting. Um, but that has real value, right? And I think you can kind of build up from there. I think coming from the other end, we do have these more deep research collaborations with labs, for, for example, like antibody design or enzyme design, and these sort of things are kind of more, you know, they result in publications, and then people will read and say, "Okay, you know, a AI system did a lot of work. It has biological novelty. It's been proven out in the wet lab." And so I think that also lends credibility to the system.
- YWYunyun Wang
Yeah, I think the simple answer is you show by doing-
- SPSpeaker
Mm
- YWYunyun Wang
... and you show by publishing and engaging with the scientific community. And I think the skepticism is really healthy and should be welcomed. I think it's, it's just really great to see people get really excited because, and also trying to, like, disprove maybe because the potential for this technology is so great if we get it right and if we can actually really leverage its full capability. So I feel like the, um, carefulness about how do we actually make this work for real problems is, like, very much warranted. But yeah, I think when we publish, um, and I think that just also shows a need for more, uh, rigorous evaluations that represent, like, these life science, um, workflows or research problems, so people can look at and eval and say, "Yes, like, I feel like now I have, like, 100 different ideas for how I can implement this into my, uh, my lab and solve some of the current bottlenecks I'm, uh, facing."
- JJJoy Jiao
I actually think there's a certain amount of, um, stress I've encountered from people who are worried that, you know, AI's really powerful, but they don't know how to use it the right way, and so there is this general feeling of, like, I need more AI in my workflow-
- SPSpeaker
Mm
- JJJoy Jiao
... in my life, but they don't know where AI should come in. And I think part of the product vision is to just make it so simple that it just works.
- SPSpeaker
Mm.
- JJJoy Jiao
So you can just go to something like Codex and say, "Hey, I want to do whatever I'm doing today," and then Codex can figure out all the different pieces, the multi-agent workflows, the tool calling, all of that. And so y- yeah, basically you don't have to stress about how to get uplift from AI, and it just happens naturally.
- 33:17 – 40:27
Advice for students and researchers
- SPSpeaker
We, we do see those step changes every time these models become smarter and understand users better. You get more utility because some people go, "I don't have to spend a lot of time trying to prompt it or figure out all the tricks to it." If you were talking to somebody who was considering getting into the life sciences, maybe a high school student right now, what advice would you give them?
- JJJoy Jiao
I feel like when I was in high school, so I did the, um, USA Biology Olympiad back when I was a high school student. And I think out of all the different Olympiads, I think biology was seen as kind of the most, like, memorization heavy one-
- SPSpeaker
Mm
- JJJoy Jiao
... um, versus, like, math, right, where it's kind of more, I don't know, test time compute scaling-
- SPSpeaker
Mm
- JJJoy Jiao
... whereas biology is more kind of memory and retrieval. I think my hope is that with AI having kind of, like, learned all the relationships between, uh, all the different research pieces, is that it can really uplift human creativity and just make the process less memorization and more kind of helping people connect different fields of research together and just kind of, um, I guess furthering the frontiers of what people are able to-
- SPSpeaker
Mm
- JJJoy Jiao
... explore in biology. So yeah, I, I feel like my advice to, I guess, a high school student, um, would be that maybe you, you don't have to kind of go and memorize all the biology books.
- SPSpeaker
Mm.
- JJJoy Jiao
You should just do more exploration with AI. I think you can definitely read papers and just ask questions. Um, and I think you can do, uh, both deeper dives and broader overviews this way, and I think just, just the, the way of learning really changes.
- YWYunyun Wang
I found that when I was in the lab, there wasn't a real, like, solo, like, individual aspect of doing biology research compared to, like, for example, when I went to my first, like, CS hackathon, there was some excitement about just, like, the collaborative nature when we first, like, built our, like, app together. So I feel like that's really a, the future I hope to see for s- uh, early adopters and students using our models and maybe using it in, like, the Codex runtime because there is, like, a collaborative nature to it too. Um, I think, like, for example, sending your scripts or sending your conversations, or maybe one day we have, like, we all have, like, our own, like, co-scientist or agent, and we can, like, deploy our agent to now work with a teammate in that way. I think that there's just, like, new, like, interactions and new modalities for us. So I would just encourage students, uh, to adopt early and just to, like, also pioneer their own path for how they would like to use it. Uh, for me personally, I always actually kind of felt like I got into the wet lab a little bit too early, and like we mentioned earlier, I did not enjoy pipetting. [laughs]
- SPSpeaker
[laughs] That's a theme here.
- YWYunyun Wang
So-
- SPSpeaker
Nobody likes pipetting.
- YWYunyun Wang
Yeah, there's a lot of, like, very intense manual tasks involved, and so I, I hope that, like, you know, when our, uh, AI models can connect with physical devices-
- SPSpeaker
Mm
- YWYunyun Wang
... that, yeah, we can just, like, make a lot of, like, the, the learning curve more fun for, for, um, students so that they can kind of like, uh, learn with the models and then kind of like maximize their time with, like, the really, um, interesting, uh, uh, interactions.
- JJJoy Jiao
Uh, spaces.
- SPSpeaker
So I've been working with a student. I like to help students come up with projects, and one of them is we've taken Codex, and he's connected it to a greenhouse and basically using it to get photos back and to look at it and to evaluate it. And I think it's been fun to see how he's been taking both, you know, AI technology and then something traditional like a greenhouse and combining them two, and basically building up the skill set of learning how to use the two of them. When you talk to your peers, you talk to people who are running labs or running experiments or researchers, what advice do you have for them? Because the problem I see is that a lot of them go, "That's great. I just don't have the time." But ultimately what we're trying to do is save them time. So do you have any kind of quick advice that you give them or any ways you try to maybe inspire them?
- JJJoy Jiao
Most people that I know, um, I think in academia use AI in, I think, two main ways that I've seen. One is to kind of talk to AI about an existing piece of, uh, research paper or something and just kind of make sure that you're understanding things the right way or kind of fact-checking. And this is personally what I really like to use AI for, 'cause you can ask really dumb questions, and you don't feel any judgment. It's actually just-
- SPSpeaker
Mm-hmm
- JJJoy Jiao
... really wonderful for learning. Um, and then I think people use it a lot for analyzing experimental results. And I think this comes back to the statistics, uh, piece that I learned where, where, where I mentioned before, um, where sometimes you don't know what the right way to analyze your data is, or there's, there's kind of so many different interdisciplinary fields that you are, uh, your data might touch on something in chemistry or something in, like, a random niche field of, like, protein biology. And the really nice thing is that a model can kind of, like, pull those different ways of data analysis in for you and kind of explore all of these different paths. Um, I feel like both of those are pretty, uh, low lift ways-
- SPSpeaker
Mm
- JJJoy Jiao
... to try things out, so you could just kind of, like, throw a PDF file at AI and just be like, "Hey, help, help me understand this paper," and just have a natural conversation. Or you can, you know, boot up Codex and do some data analysis directly on your laptop.
- YWYunyun Wang
Yeah. I, I would say that you'd have to start with making sure it doesn't feel like work right away.
- SPSpeaker
Hmm.
- YWYunyun Wang
So maybe it'll be easier when you're focusing on AI adoption to just, like, work on, like, a hobby project or a passion project. Uh, for, for me, for example, I actually started, um, working on, like, more like literature synthesis, um, tasks when I was doing creative writing projects, which are kind of like just something that was not at all related to, um, like our day-to-day, even though the-
- SPSpeaker
Mm
- YWYunyun Wang
... uh, biology is a very creative space. [chuckles] I was just exploring that through, like, a different, different medium. And I think that's actually when I started unlocking a, like, a lot of different ways to either prompt the model or to actually access different, uh, data sources. So I think that kind of just gave me a lot of, like, um, pattern matching abilities for when I was trying to apply it, because we're not gonna get it right in the first try, and it is really hard, and I feel like the, the progress and pace of this field moves so fast that every like, week or month there is, like, a new, like, pretty exciting development that might change how we, um, engage with models or AI systems. So I think it's just important to get started somewhere, and I think another theme is the collaboration element. I feel like it's more powerful when you have a recommendation from either somebody on your direct team who is doing the same day-to-day tasks as you. Um, that happens a lot on our team as well, where somebody will say, "Oh, I got Codex to, like, now touch these three different, like, internal, like, databases that we weren't able to connect before." And I don't even-- like the, the latent space, the latent capabilities are just so vast that there's a lot that we just don't know until a- again, we can do it. So I think just having conversations with your friends, your lab mates, your teammates will, I think, spark a lot of those conversat- uh, a lot of those cre- creative juices and then help you, help you with your own adoption.
- 40:27 – 44:25
Where are we in 10 years?
- SPSpeaker
What does science look like 10 years from now?
- YWYunyun Wang
I think when we started this team, we do have, like, really just ambitious targets, and one of those is, like, I think we do want to make meaningful strides towards, or even if, like, assist with, like, um, curing a disease. And I think there's just so many, uh, rare, like, orphan diseases that doesn't really have the attention, uh, um, and the resources that it warrants because it's just such a, um, a difficult field to actually, like, for example, like, clinical research is so difficult to actually bring that to patients and to market. Uh, so while 10 years I feel like is just really, uh, a really long timeline, I'm, I'm really excited about, like, some of the progress that we can make, and I think it's good to, like, be carefully optimistic that, like, we're gonna see some of those breakthroughs pretty soon.
- JJJoy Jiao
Yeah. I think maybe this is a bit of a sci-fi vision that I have of the world that I, I really hope becomes reality, which is that you have these autonomous labs that are just mostly robots, and you have them all hooked up to AI, and you just have autonomous research institutes that are constantly running and curing human disease. It's maybe making new materials, uh, making new drugs. It's maybe solving personalized medicine. There's a lot of N of 1 or just ultra rare diseases where people without vast monetary and research scientific resources can't even begin to think about solving, but we can solve that with AI, and I think we can kind of almost break through the financial and regulatory and monetary constraints with the system. So I think that that's, like, kind of the dream. And I think also even separately thinking kind of more about the biosecurity side of things, um, these systems can be kind of constantly sampling our environment, right? It can be sampling wastewater. It can be sampling the air and constantly detecting potential threats or even just, you know, better predictions for the flu and getting better flu vaccines. But just generally, these different m- medical countermeasures, I think should be happening autonomously in 10 years, and I think that that's basically something, yeah, I'm really excited about.
- SPSpeaker
The, the AI lab is exciting because I think if people really understand what it means is it's not, there aren't scientists, it's they're more scientists, but they sit at home, and they go into Codex and say, "Can you go run this experiment for me?" And like, you have a data center, you have a science center doing that.
- JJJoy Jiao
Right. Exactly. Yeah, and I think I, I didn't, um, talk about the scientist in this, in the vision I was just, uh, describing, but obviously there are people involved in here, and I think it's really kind of high level direction setting from the humans. We're saying, "Here's a patient with this disease. Here are some potential solutions or things that maybe you can look at." And I think the AI can then go off and explore different ideas. It can design experiments and then come back to the humans and say, "Here's what I found. What do you think, um, we should do next?" And this can be kind of a, um, academic discussion. It's a little bit s- uh, similar to kind of the way that people interact with Codex today, where you say, "Here, go write a function," or, "Go write a piece of code," and it writes it and say, "Here, here's the code." And then the person tells it the next thing to do. So I think it's a little bit similar to that kind of interaction, but on a much grander scale and on a much longer time horizon.
- YWYunyun Wang
I think it's really, like, the democratizing science aspect and putting, like, really capable expert level, um, knowledge in the hands of a greater amount of people, and I think what that can mean for personalized medicine, for bolstering our societal defenses. Um, there's just, like, so many naturally occurring, um, new, like, variants every year, new, like, uh, influenza strains. So I think it's really just, like, securing defenses and feeling like we actually have more agency to counter all of that, and I think I'm really excited about a lot of, like, the, uh, medical countermeasure acceleration work as well.
- SPSpeaker
Well, it's very exciting. Thank you for sharing this with us.
- YWYunyun Wang
Thank you for having us.
- SPSpeaker
Yeah.
- YWYunyun Wang
Yeah. Thank you so much.
Episode duration: 44:25
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Transcript of episode UZyH0nx5zgI
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome