How Autonomous Labs Will Transform Scientific Research: Ginkgo Bioworks’ Jason Kelly

All of the previous revolutions in tech internet right like social media like whatever have been totally meaningless to biotechnology and biopharma like yeah it's nice we communicate slightly better or whatever it's just I'm like back off as IT crap right like not this but like this this is is actually going to change the fundamentals of how we do science and or big science industries like biopharma are going to get disrupted I really believe of that and that's not been true for the last 30 years of tech we're thrilled to have Jason Kelly founder and CEO of Ginkgo Bioworks with us today thank you for joining us yeah thanks on you so you started Ginkgo Bioworks in 2008 with the goal of making biology programmable and programmable has taken on completely different meaning in the era of AI so I'm very excited for the conversation today maybe tell us about the journey so far I mean I'll do the Ginkgo journey in short right yeah we started in 2008 but we didn't actually raise any capital until 2014 so we bootstrapped for four or five years which like if you're not a bioperson this doesn't make sense but in biotech VC they really don't like like young people for example so we had to start the company like straight out of grad school it was 2008 we weren't trying to make a drug so we were like totally uninvestable were you full-time focused on the company for those for six years oh yeah okay oh yeah yeah we were basically like applying we did government grants and service business so it's like pretty brutal start and then summer 14 say I'm Altman now mr famous right this blog post because he just took over YC and he's like hey I think the Silicon Valley model can work for like deep tech you know nuclear vision biotech material science and so I wrote him an email I was like oh man like thank you for I mean we're like five years old like I've 15 people in a lab in Boston we don't make any sense for YC but this is like a oasis in the desert you know like nobody will invest in weird companies like this and he's like no you got to meet me so I flew out to San Francisco at him and he's like you should do YC I was like I should do YC so then we did YC so we kind of that was sort of when you know if you really want to mark Ginkgo from like having capital it was sort of in 2014 and that was the product change since 2014 oh well so the mission has changed but the product has gone all over the place different different roads so we've always wanted to make biology easier to engineer that was the idea and so if you're you know this is a very much I got I remember a make biology programmable has have the words always been make it easier to engineer because I feel like that's a slight I was always we were talking down why it was always like this the computer science rapper on make biology easier to engineer but yeah that was always our mission okay got it but the analogy solid right so you know DNA is code right it's ATC's and G's not zeros and ones it's really our only other like coded product other than computers it is really biotechnology and so the core idea of Ginkgo was well if you could design DNA code you can program cells to do things and cells are you know they're programmable like computers right but unlike computers which just move information around cells move atom around so if you can build whatever you want that's we think ultimately going to be a huge market huge opportunity but the challenges our ability to program cells today is really bad and so how could you fix that that was the core idea behind Ginkgo and how is the product itself changed over time yeah so the way we went we had to market it first was we're going to try to build what we called foundries which was sort of a centralized laboratory that would kind of automate the lab work associated with doing biotech and and the reason again if your computer scientists the way to think about this is if you want to compile in debug DNA code that's a physical process yep right so you're like ATC GGG like we have to do phosphoramidic chemistry got to build the piece of DNA you want and then put it into a cell grow the cell and test the cell and that's your kind of compiled debug cycle yeah that makes sense and so one half of what we worked on technologically was how do we make that cheaper right because if you want to get better at doing this you got to do more of it faster and for less expense and then the second thing we worked on was how do you get better at your programming in other words like that design you choose to test in the lab how do you improve the odds that it does what you want so sort of like get better at design the biology and make it cheaper to try things were like basically for the last 15 years the twin activities then in an era of the AI I see you see opportunities on both sides for that today and that's we shifted a little bit over the years in terms of how we do it but though does that make sense that you kind of have in roughly speaking yeah the design piece sounds like it's a bit more software the testing piece sounds like it's a bit more hardware controlled by software very much yeah that's exactly right and today the folks leading on the design side you might see companies like Chi bio for example like there's like these protein models bolts the folks at arc institute are really just came out today with a paper we'll call it about Evo 2 which is like a genomic model there's a whole community of people now trying to solve the problem of designing biology with AI the big change again over the last two years is I've kind of stopped working on that problem I'm like we had our own approach to solving it it's hard right like designing cells is tricky we're going to try to solve this half of it out of the problem which is how do you make it cheaper and faster to try things in the lab and how can you we can talk a little bit more recent a project with open AI how could you have AI models help you do that is anybody else focused on the back in so to speak I kind of think about design is the front in the process and testing is the back in when I say anybody else obviously people do this yeah is anybody else with a similar approach focused on that part of the market there you there's some new companies right so there's companies like Madra out here is one that's doing it with like robotic arms trying to accelerate it you have like the life science tools industry but I would say it does not have a Silicon Valley attitude about things in the sense that they're not really trying to change the fundamentals of how you do it they're sort of like just providing the next tool to people doing it the way they've always done it and so we've always been this kind of unique force trying to say hey is there a new platform right there's something like the jump to planar semiconductor manufacturing in electronics at the beginning of intel you know like there's some some way we should just do all this stuff differently that could make it way better in the future and that's always been the ginkgo like what I think is unique about what we've been doing the last 10 years what catalyzed the focus on this part of the business on the on this half of the house yeah I think this is an engineering problem and I think this is a science problem okay so and I went after both initially and I kind of took my licks for that and I think the the good thing about an engineering problem is you can ultimately like render it yeah to dust right a science problem I think is great if you hit it but it's much more unpredictable and so in this era of ginkgo I've got like the resources marshalled at this point to go after this and kind of see it through and so that's why you see me pointing in that direction and with the efforts that we see with the arc institute and others of that ilk what ending are we in so to speak like is the ecosystem around the science problem going to start producing meaningful results soon other than papers uh it's a good question I think the the hard part about designing biology biology is amazing by the way right like just as a substrate again right like if you think about what's happening inside of a cell it is producing you know intel or now in video TSMC level caliber atomic placement basically for free yeah right it's so it's able to do molecular assembly itself repairs itself replicates like as a physical substrate it's insane it is the product of four billion years of evolution so the complexity embedded in a cell is actually a lot bigger I think that people like give a credit for and so there's a march there now that said you know all of you know more than half of your drugs today are produced by biotechnology we cure cancer we do this we have huge value coming out of even with the tool the limited tools we have today so you don't have to solve the whole problem yeah over here to create a lot of value you just have to be better than how we do it today does that make sense and so I think they have a really good opportunities already in the near term with all the protein models you're seeing that right like I just did a big deal with lily like there's real opportunities there I think right in the near term yeah that makes sense so speaking of open AI and Mr. Altman yeah you recently announced a partnership research result with them can you say more about that yeah okay so this is this is pretty exciting I think for the that folks that are following AI it's pretty neat so basically what we did was we took our we call it an autonomous lab right and so I can talk more about this but the short answer is if you really want to drive efficiency on the lab side you need to get the human beings off of the lab bench right so the way we do and this is in true in biotechnology it's kind of true for science broadly the way we do science today 95% of science the stuff that's not theoretical so not you know is everybody's working on math like let's work on Terrence Tao let's get a Terrence Tao in a box or whatever right like you know like the reason is you can just simulate all that stuff on a computer let's play chess you know right like yeah no kidding right but if you look at the majority of like what we spend money on in the United States and and just generally across the world in science it's largely on experimental work yeah and the reason is if you want to learn something new about the world which is what science is fundamentally you have to go out usually and like poke it yeah probably you have an opinion of hypothesis but you got to go test it to actually figure it out so it's experimental science that moves the needle in my view and so the question was could a reasoning model do the work of experimental science if you gave it a robotic lab that was the question and the answer was yeah it's actually pretty damn good so we did basically the way the project work was we had there's a biochemistry problem called cell-free synthesis okay so you take a piece of DNA ATC GGG right if you were to put it in your cell right now remember like a central dogma in high school right like it's like DNA makes RNA makes a protein right and so you put that DNA into a cell and it'll be. make a protein. Well, you can do a thing called cell free where you pop a cell open, take the guts, put it in a test tube, and then add the DNA to that. And because the guts are still there, it makes the protein. So this is kind of like, it's like the world's smallest three printer or something, right? Okay. And so scientists use this. They try to optimize it's very expensive usually. And so there's a paper that came out of Stanford from Mike Jewett's lab in August that set the benchmark for like how cheap people had been able to do cell free protein synthesis. And so we said, all right, let's try to optimize that with the model. And so we gave the model, we did each round, we would do a hundred three, eighty four well plates. Okay. So each well and a plate is like a little kind of cup of liquid. And you can do an experiment in there. And so we gave it, you know, 30,000 experiments to run. And after it would run those experiments, it gets the data back and designs another set. So after four rounds of that, we beat state of the art. And after six rounds, we beat it by 40%. And so that was a, I think it's the most interesting sort of model doing experimental work result that's been shown to date by a lot. And the 40% was a function of what just faster cycle time or no more intelligent experiment design. Yeah. How did it be to the state of the art? So this is, this is my point, this is I think my, my larger point about science because I think we're going to do science differently in the future, in my view, based on what I'm starting to see here. So what does the sciences do when they're doing experimental work? They're coming up with an idea. And then they're trying to design an experiment to ask a question about that idea. Then they're going to run the experiment, take the data back, interpret it, and then poke again, based on what they learned. And they're going to go through that process a few times to like resolve something. Oh, this is how, you know, whatever. This cancer works. This is how this piece of materials, you know, works. This is this, this is that. And, and so that cycling is just logic. Yeah. Right. And so it doesn't require you to model biology or simulate anything. It's not that half of the house. It just requires you to be almost like a programmer. Like you need to be logical, run through set of things, do data analysis and draw conclusions. And so that, like, that's all it has to do. Does that make sense? Yeah. And so we didn't do anything other than that. Right. What really let it break through wasn't that it was so smart, it was that it could run experiments. And the question was just, could it design them like a scientist could the answer was, yeah, hell, yeah, could. And so now I think that opens a real interesting question about like how we do science in the United States. Well, and it's easy to imagine a version of the future in which the, you know, the scientific method, the design and the hypothesis testing and all that is done by reasoning models of some sort. And the actual testing is done by ECO, you know, autonomous labs. Yeah. What's wrong with that vision in the future? And if, and if that is the right vision, the future, how far out is it? I mean, I think this is how it's going to happen. I really do. Like, like, I mean, I'm probably more substantially more aggro on this than like the average scientist today. But like, it's, so I'll just explain like why, where I think if you had a heads up competition, which I want to do, right? Yeah. And then tell me how the average scientist would push back and say, like, no, no, it's not going to happen that way because. Right. So I think what the, what the scientist will push back on is like, this thing can just be as creative or as me or something, right? Which I actually, I, I'm sympathetic to that. Yeah. I'm not saying it's going to be more creative. I'm saying, it's going to be way more creative. I'm saying you do. Yeah. No, that's still a good value. I'm saying it, it can run a lab 24 hours a day. It, I'll give you another example. Like the way science works today is you would have a lab, you'd have a lab, I'd have a lab. We're all working on the same area. Let's say we're working on like all timers. You have a hypothesis. I have a hypothesis. We each kind of pursue it. We're collecting data over the course of a year or two. And then based on what I see at the end, I write a paper. And when it comes out in the public literature, you get to read it and you get to read it. And you're all doing the same thing. So we're kind of like exchanging information every year or two. And you're, I'm not getting to see every experiment you did, by the way, I'm getting like the distilled output of what you think you saw over two years. That makes sense. Yeah. All right. So let's contrast that to like what I think should start to happen now based on what I saw with this open AI project. What I think should happen now is you should have a robotic lab that has every piece of equipment that we all have in our labs. So it can run any experiment you want. We can talk in a minute. It's actually a pretty technically difficult problem, but let's just wave that away. Okay. Salt. All right. Great. So then I'm going to put a hundred AI scientists on top of this thing. Yeah. Each one is going to pursue a different hypothesis for Alzheimer's. All right? Great. And they're going to run their experiments just like you would in your lab that day. But at the end of the day, they're going to pass the data on those experiments. Like what experiment they ran and the raw data that came off it to the other hundred days daily every fucking day. Okay. And so they're going to learn from each other like you're even though your hypothesis is different, we're working in the same area. Yeah. So your failed result might like, for example, say your experiment went the wrong way from your hypothesis. That data might be relevant to my hypothesis. And I would never see that normally. Yep. And so that's all just shugging along. And every week it dumps a, you know, a lab notebook entry or like a mini paper, like a conclusion about what the hundred of them have figured out that week, that we can all read and see and use that we can direct, we say this, hey, cut that line of research or whatever. And so like that's number one, I think the information sharing and like the ability to handle like really broad context across a lot of projects for the eyes is just better than it's just socially different, even than how we do it today. That's unfair advantage number one. Okay. Unfair. Van number two, if you look at how we spend money in science, remember all this stuff like the NIH was like, what's up with the indirect rates at the academic universities and all this hullabaloo right? Well, what's an indirect rate? Well, it's basically paying for manual laboratories. That's what it pays for. Okay, you've got these labs and they're there 24 seven, but they're used five days a week. Okay, they have equipment, but every lab, all three of our labs, we have same copies of the same equipment. Yep. Because we all got to do the same work. We don't share each other's new, new, new, I like a door in my lab, only my lab gets to use it. Your lab uses yours. So we have all low utilization rate of our equipment. It just now works. Okay, right? And so you have a very inefficient, like if you look at the spending on research, and this is true, the 60 to 80 billion dollars a year that biopharmus spends or the 40 billion that NIH spends less than 5% is on the reagents. Everything is on overhead. It's basically overhead. The people, the regulatory, the lab space. If we were running it efficiently, you would budget a research program at the NIH, not on indirect and heads and everything else, but just on the reagents, because that's like the usage-based pricing of science. Yeah. Because to actually do experimental work, I have to consume some chemicals. I have to consume a piece of plastic plateware, like whatever the hell it is, like I'm actually doing atoms in the physical world. I got to burn some stuff up. That should be the dominant cost. It's the opposite right now. Yes. It's like less than 5%. So the other advantage those AIs will have is if they're able to run robotic labs, now they're running where 90% of the cost of a research project goes to the reagents. Yes. Oh my God. Right. So that's like a 10X increase in the amount of data per dollar that you're getting compared to how we do it today. So I think you combine those two things. But without the AIs even being smarter. They can even be dumber than the scientists. I think they win. I really think they win. And so I think we got to reevaluate like how we fund what we fund with the NIH. I think every biopharma head of R&D needs to care about this. And like, and I think there's a blind spot. By the way, like we did YC. I know all the tech people and you know, I've always been adjacent to this stuff, right? All of the previous revolutions in tech, internet, right, like social media, like whatever have been totally meaningless to biotechnology and biopharma. Like, yeah, it's nice. We communicate slightly better or whatever. It's just I'm like back off as IT crap, right? Like, not this. Like, like this, this is actually going to change the fundamentals of how we do science and and our big science industries like biopharma are going to get disrupted. I really believe that. And that's not been true for the last 30 years attack for our partner. Constantine has a good framework for that. He talks about how there are revolutions and computation and revolutions and communication. Yes, communication is about the distribution of information. Computation is about the processing of information. Yes. And what you're talking about here is just a different way to process the information. I got it. And so the last several revolutions have been about the distribution side of the equation, which doesn't get to the core of what it is you're doing. Completely agree. And that's, I think, fundamentally true. And so again, I think the leaders, biopharma companies and also the leaders of research universities and these people that are in the business of doing science to produce either products or for the government. Yeah. Cannot ignore this. They cannot ignore AI. It is, it is just different. Yeah. I am, I'm telling you, I'm a person who has been adjacent to this crap for 15, 20 years now. And this is the first time I've like, the tech guy's finally did something cool. Yeah. And just saying, like, so you think AI is the chalising force behind, you know, cloud labs should be a thing. Nobody really ever moved to them. But like, AI will be the reason they move. Yeah. Well, I'll be talking about cloud labs. Yeah. So let's talk about autonomous labs. And then I'll explain the cloud. All right. So why has it been hard? Right? Like, like, the average tech person's look at how science is done where you have literally PhD trained people. These are brilliant people. Yeah. Paid decent amount of money, standing, I did a PhD at MIT and bioengineering. It's five years of moving liquids around the lab bench by hand. I swear to God. Like, like, it is like, like, like, my friends from undergrad just were like, they would never write it out. Like, it's ridiculous. You would do like manual labor, right? Like absurdity, right? And our group, but like, that's what you have to, you have to do that. If you want to play at the edge of science, you got to do physical work. And so that's what you learn to do. Okay. And so it's like, well, everyone And so I was like, well, just automated, bro. Like, you know, like, why don't we just do that? Okay, so why is it, why is it hard? All right. And the reason it's hard is it's like the technical like automation need term is like high mix low volume work. Yeah, okay. And this is true at like places like Hadrian today, for example, that are working on this on the manufacturing side industrial. High mix low volume is hard to automate historically. All right. And my like transportation analogy that I've been giving to people in bio is like, okay. So imagine on the y axis, you have like level of automation. Yep. And then on the x axis, okay, you have like flexibility, like that variability of the mix in what you're asking it to do. So in transportation, low mix, high automation, that's like a subway. Sit down, takes you away, right? Like, you know, maybe you like don't have to do anything, but you got to want to go to one of the stops on that subway line. Yep. Low automation, high variability, that's car. Yep. So on the wheel, foot on the pedal, take your right to your house or the grocery store. And that's what the transportation system looked like for the last 100 years. Until thank you very much. You know, Google, we got Waymo up here in the corner where you get the flexibility, you know, the automation of a subway, but the flexibility of a car, okay? And it's so surprising that we don't even call it automation anymore. We make up a new word. We call it autonomous, autonomous car. Because the way I look at it is since the industrial revolution, we've basically been automating everything that is low mix. That's low variable from the lume on. Okay, right? And we just hit a wall. And AI, we hit a wall and flexibility of what you can do. And AI pushes us past it. Yep. That is like every part of the physical, like our physical infrastructure, post-industrial revolution, everything has to get looked at again with that lens as we move up the variability. And so that's what we're, so like that's the tricky bit in like lab land. We actually have automation, but it's like subways. Yes. It's just repeat the same experiment at a diagnostic company like Quest. They would have automation if you're a high throughput screening and farm others automation, but it can't do the variability. So 99% of the work, just like 99% of miles traveled is in cars. 99% of the lab work is still at the fricking bench. Yeah. And that's what you got to fix. And so the waymo analogy is an interesting one because it's now such a magical thing that so many people have gotten to experience. Yes. You know, you kind of had your sensor suite, you know, your radar and your LiDAR and your cameras. Yeah. Then you had your software suite with the perception and the planning and the actuation, which then had to tie back into whatever vehicle manufacturer you're working with. And then there are gazillion corner cases that you have to simulate because you can't get enough of them in the real world. Like, what would that set of words be for your world? Yeah. Great question. What are all of these specific things that are hard to get right? 100%. And I think if you're like trying to generalize the problem of bringing automation to autonomy, it's going to be different in every domain. Yeah. So for cars, the hard part is the physical world is changing. Yeah. Every mile you drive the world, I got a bit of new place, right? Like, this one has a cone, it's raining, like whatever, right? Like, that's not the problem in the lab at all. Yeah. Lab, I can make it. It's my lab every day. It's the same fucking room that robots are. Nothing is changing. Okay. The physical environment's not changing at all. Yeah. So that like, it is not at all the same stack that brought you autonomous cars that will bring you autonomous labs. That's not my problem. Okay. So in my world, it's the variability in like what the scientist is asking for that makes it hard, right? So they're like, I want to use this piece of equipment. I want to use that piece of equipment. This is my combination of things, right? And so one of your big problems is getting a thousand long tail pieces of third party, like you can't believe the software on these things, a bench top lab equipment integrated into one big system. Okay. So that they all can be controlled by your software. And problem number one, it's like integration of bench top equipment problem number two. What are we doing with our hands when we do science largely in bio anyway? It's liquid handling. So you pick up a pipette, which is like, if you didn't do this in high school, it's like the world's fancy a straw and you like suck up a little bit of liquid and you squirt it out in the right spot. And like, but the thing is, if the liquid is viscous, like, it's like syrup, then if you do naturally with your thumb, like, adjust the pressure of the straw because you can see with your eye if it's working or not. So it's actually liquid handling turns out to be a trickier like problem than you think and you are doing some work as the human to like manage that. Yep. So you have two big buckets. One, solve liquid handling to send samples to a thousand different pieces of equipment. Okay. Yep. If you nail those two, you're done. Well, that's also, yeah. Tractable. It is. Tractable. Yeah. I agree. Yeah. It's totally tractable. And so it's just a lot of work. I agree with that. I think we basically have it working. Yeah. That's the honest truth. I mean, we basically got to pay-- What's the last, like, major hurdle? One reason it's hard for people to do this technically is they want to build the hardware. Yeah. But they don't do research. So they kind of have to go to a customer and be like, hey, you want to use my robot? That's where it's like, no. Right? Like, they're at the bench. They're like, no, you know, I don't like to try. You're like, so there's like an adoption issue that I think has made it really hard. And so we have this advantage that because we have a research, we sell like our original business, which was research partnerships, which you still do, means I have a bunch of scientists employed at Ginkgo. These scientists are basically, like, remember the Google engineers who had, like, sit with their hands like this next to the wheel five, seven years ago in Palo Alto, like grabbing it like the Waymo Droven to a mailbox or something, right? Yeah. That's my scientist today. So they are dog fooding on our-- we have, like, 50 robots. We're going to 100 in our big lab in Boston. And so they're the ones, like, trying out and breaking it, things that have broken, running a bunch of work in parallel across that system is a scheduling challenge. So you have to be able to manage all that. So just handling the scheduling with tight timing on experiments is, like, algorithmically tricky. And we've had to figure out a bunch of stuff. Getting the equipment to work all-- like, when it works all day long, making that reliable compared to it's being barely used at the lab bench, that's tricky, right? So there's just these things that we have to keep knocking down. But they're engineering at this point. And we saw, like, pipetting and good handling. Yeah, the good thing is there is a whole industry that's worked on that problem, like, liquid handling robotics. It's just a matter of having all the different liquid handlers. And if you have them all, you can kind of-- and you know your liquid class you're dealing with, you can manage it. Oh, one other one, big one. Scientists don't code. OK. So oh, cool. Use my robots. This is what everybody's done. They've made, like, visual programming languages. Like, if you're a scientist, there's a thing called LabView. It's, like, complete trash. But like, it's-- you know, make a flow chart, right? Because you can't write Python or whatever. Horrible. OK, like, even that, they hate. OK, right? And so no one will program chat. And so we ran to this issue where we now have all these scientists using the automation directly. So like, I don't know, three weeks ago or something. We had two instances where we sent a plate. So like, again, this is like kind of bunch of little wells with liquids in them. And we seal the plate when we put it in storage. So it doesn't evaporate. So they sent a sealed plate to the pipetting robot. And pipet comes down and it gets stuck in the seal. And people on our slack are like, hey, what the hell? Like, do you seal the plate before you send it to the liquid atom or dumb dumb, right? Like, you know, and it's like-- and this is horrible for scientists who has basically expert at liquid handling at the bench. And now they're like making like, you know, basic mistakes here in the deal horrible. That's a bad UI. OK. And I was just like, this is nonsense. We're from now on only what the way we're going to interact with writing the code is through Cloud Code or Codex. Like, you will now submit a written protocol what you want and the model will figure it out. And if the model sends a plate sealed, we will update the skills file and it will never do it again. And we will get through this. OK. And so that is a big win for usability of robotics for scientists is what's happening with Cloud Code and Codex. That does make sense. So like, thank you. You know, right? So these are the things that that had to get knocked down too. And so all this is in flight. But like, right now in Boston is like a very unique experiment happening where we have like 50 scientists submitting jobs into one big robotic setup. That exists nowhere else on the planet right now. And so it's pretty neat to watch it. Do you see a future for humanoids? No. Really? In the lab? Because the best argument I've heard for humanoids is like the world, the physical world is designed for them. And like, I would think that existing labs are made for like humans walking around by petting things, walking between different machines. Yeah. And so you could try to create a robotic arm that's able to like, you know, orchestrate all this, or you could just have a humanoid do with a human lab scientists does. Yeah. So again, like the primary thing that the human is doing is moving samples around the environment. Yeah. There are much better ways to do that than like walk them by petally among things. Like you just put them on a track. Like our system has like a nice little track. And the plates move with extremely high liability. They get delivered with micron specificity to where they are. The armpits them up. It's like, you know, that problem just disappears. Okay. And then the other reason is in the long run, the humans are the limitation. Yeah. Right. It's not like, are humanoid robots going to disrupt TSMC? Are they going to go in there and etch the fucking chips? Like, obviously not. You know, right? Like, you're like, no, it's a microscopic discipline. Like biology's a microscopic discipline. Like these things are, no. Yeah. Makes no sense. Yeah. How does it change the unit of skill for a lab? Like I'm imagining that labs in the future are going to be these enormous things the way the data centers have become these enormous things. So it's actually going to make them smaller. Really? Okay. Yeah. Because, again, you've probably not seen this. But like, if you were to walk through Merck's campus, you'll see a million square feet of laboratory benches. Yeah. Across a bunch of different buildings everywhere, whatever, right? And they're set up for basically humans to be able to walk in and find a piece of equipment. Again, underutilized but basically available whenever they need it to run whatever experiment they've come up with by thinking over the last two weeks and not even working in the lab. And so like that kind of like cycling is kind of how it operates. And you need it local. Like if you have a team now in this new place because you bought this company, they need to lab. So you replicate another lab, right? The labs have to go wherever your scientists are. Well, let's now instead imagine the scientists are ordering all their experimental work through computers. And it's going to some centralized autonomous lab, you can think of that like a local cloud if you want. Okay. Well, now you don't need a lab where the scientists are. So you get a rate of all the just duplication that you have because of physical people, you also get wildly better utilization of the bench top equipment like we're talking going from like sub 20% utilization at the bench to like 70%. So now you need less equipment. And then assuming you didn't decide to have humanoids, like you can just jam it all in around the track system. So it's actually a lot tighter. Yeah. Yeah. Yeah. So we have like a major space reduction at the moment. That's one of the big savings. Like when we talked to it, like we just sold 97 robots to the Department of Energy for this like Genesis mission. This is like the AI for science thing that Trump's doing. And like that is basically going to be ultimately much more dense than the equivalent set of labs would have been that would have like housed that otherwise. And that's part of the sales pitch. It's like you could have less spending. Remember, I told you earlier, like the spending is not on the reagents. It's basically like roof space. Yeah. Like laboratory space and people. What is the unit of work? You said you sold 97 robots. Is that 97 boxes? Yeah. So our particular device, it would be called a rack, it's like a reconfigurable automation cart, it's basically like a box that has a piece of bench top equipment in it, a six-axis robotic arm. This is like nothing special for labs. This is like tradition. This is like coming out of manufacturing tech. And then a piece of mag lab track. And what you do is you Lego block the carts. So we got like 50 of them all together in our lab in Boston. And then a sample can move on the track. And in front of every piece of equipment is an arm and they aren't picks it up and puts it on the equipment. Okay. Yeah. Yeah. And so our unit is we sell the box. We sell like a subscription, basically like service fee plus software subscription for per box. Yep. And then eventually what I want to sell is like automation friendly reagents because that's kind of like the usage pricing. So that would be my, that's like one half of my business now is like I'll build you an autonomous lab, you know, specific Northwest national lab DOE or Merck or whoever. And then the other half is what you said, I'll run my lab in Boston as a cloud and you could just order. Yeah. Totally. Can we talk about training? Like a lot of what we've talking about to me seems like inference. Like it's a use case of the reasoning. Seems to me that you have you are generating an enormously helpful data set here that should be used like back prop into the weights themselves. Yeah. How's that going to play out? That's a good question. I don't know yet. I think I think there'll be one training so there's two different levels of challenges. One was that thing I mentioned earlier, like you're submitting a piece of physical work. And again, I think this applies to labs. This will also apply in like light manufacturing, right? Like you want to do us like, oh, you're like a prototyping shop or whatever. Any place where there's like variability. Okay. And so I see a lot of variable requests from scientists. I see all the edge cases of how it breaks the physical equipment. Yeah. That you can't like compile out. Yep. Make a digital twin. No. Like, you know, right? Like you actually do this stuff. Like I do not think like I don't buy it. Right? Like a lot of it's edge case. You know, liquid classes. You can't really pick it up on a camera. There's just like all this stuff that are like the edge cases that Waymo had to see driving around are edge cases we see by doing a lot of variable work on the same system. That is one type of training. I don't know that it's fully like model training as much as it probably is just like a giant file or something. But like that's one. Yep. The bigger one is sort of the model's ability to take your intent and turn it into an experimental plan. And that's interesting. That's back to like, because these things just like blow science out of the water. And that I think you could have a really cool loop that has more to do with the results of every experiment. Yeah. Like it's open AI project. Like, as we saw what experiments worked and didn't, you could theoretically then teach the, the model to be a better scientist. It's like Newton Einstein. Like, you know, like these are the people that moved the species forward at the end of the day, right? Like everything else is kind of noise. Like we're just going around circles, you know, like the Romans did it, you know, right? Yeah. Like we're just doing the same stuff. Yeah. Right? That's the Greeks. But, but except for science. Yeah. Right? And so I do think this is like, if you crack that nut, like if you can, if models plus, again, I think you've got to have the experimental work. Like that really 10Xs or 100Xs speed we do scientific discovery, like that 10X or 100Xs is the progress of the species. Am I? Yeah. It just seems to me that the results of what you're generating in the lab need to feed back into updating the model weights somehow. Otherwise, we're not going to get to that point. I agree. But I think that's totally doable. Like I think that loop and I think this is something that the frontier models are starting to care about. Like, like getting better at that, I do think is, again, in my view. Some of the most important part of human intelligence is our ability to push the frontier of knowledge. I mean, spreadsheets are cool too, you know, right? Like doing back office for a dentist, also fine. You know, right? Like, you can make money with that. But like, if we really want them to be smart, yeah, this is, this is where it matters the most. I agree. You mentioned Project Genesis earlier. Yeah. What is it? Why does it matter? Yeah. So, this is came out of White House and OSTP. So, the Office of Science and Technology Policy, like Micraseo's there. And so, and it's run by the Department of Energy. And that's in part because like the Department of Energy is, if you're a science nerd, that's where we do like our big science projects. Okay. So, like starting with the Manhattan Project, but also like the human genome project actually was like a Department of Energy project. Okay. So, when it's like project based, it kind of tends to live in the Department of Energy and like more open-ended sciences like the National Science Foundation. Okay. So, DOE is running it. So, it's a project. And so, they're like, all right, great. We're going to have a list. And they actually put out a list of like, these are areas where we would like to see breakthroughs that are relevant for the American public. Yep. Okay. And one of them's, you know, a couple of them are bio-related, but there's other stuff too. Material science, new energy, all these different things, right? And then what we want to do is bring AI models into the national labs, which is where we do a lot of our big science in the country, and accelerate them. And they're targeted to double the acceleration of science in the next few years. All right. So, that's the idea. And so, one way you're going to do it is they want to take the existing data that is at the national labs, and like, basically feed it in the models and see if you could find new stuff from data we've already collected. But then the other way they want to do it is to have autonomous labs that can generate new data in the direction of models. And so, when we did that deal with part of energy, the part of energy is secretary right. And I, like, ribbon cut the first 18 robots up in Washington, like, signed it, it was really cool. And so, like, I think that's a, to me, again, as a science nerd, like, that's a good push for us. And I think you want to see some good results soon, because ultimately they got a, you know, Congress has to get excited about this. You'll need, like, a bigger bolus of money in the future to really make it a big deal. But I like what they're doing. Like, I think it's the right thing to do. What is the dream scenario? What does that do for America? So I chaired this National Security Commission on Emerging Biotech in DC for, like, two years and Senator Young in Indiana chairs it now. And there's a similar one on AI that Eric Schmidt chaired, like, seven years ago. So it's super fascinating to see, like, like, learn more about DC, number one, Congress. And, but also, like, the point is, how does the U.S. stay competitive in technology areas that are strategic? Yep. And so there's one for cyber, like, 15 years ago or something, and there was AI, and then this was on, on bioengineering, bio tech. We have had an unfair advantage, like, since the Soviet Union fell, basically, where, like, we were automatically at the lead in science, like, there was no, one else that had the money to spend on science, basically. Because science is, like, it's not just spending on the science, you also need scientists. So you have to have research universities to train these people, like, it is esoteric. So that was true. That was true. And now, with China, it's rise. It's not true. Right. Like, if you look at the number of, like, scientific papers published, it's more from China now. If you look at, in my world, like, biotech drugs, not manufacturing drugs, like, making new drugs, the way it works in our industry is, like, a startup discovers a drug, and then they try to sell it to a market advisor, and they're, they're, they're like, kind of, go to market channel, and the market advisor buys it for $2 billion or $5 billion or $10 billion, depending on where it is in, like, clinical trials. Yep. Great. Three years ago, less than 5% from China. Last quarter, 40%, plus. Wow. Okay. Yeah. So we, and that's innovation. That's, like, discovery, right? Why is it, why are, why are there, why is that growing so fast in China? They have just as many scientists as us. They're just as smart as our scientists. They get paid less. And remember, it's like a hands in the lab. You got to do, science is driven by experimental work. So if you now have more experimentalists in China, and you get more research per dollar, I don't see why they don't win in research. Yeah. And so, so from my standpoint, we need to, we need to make this change, both in how we do experimental work and bring in like the AIs to just increase our amount of like intellectual horsepower. If we're going to keep up in science, and you don't want to be surprised, right? Like, if you know, like DARPA, like, you know, like, the internet DARPA, right? Like the, the founding point of DARPA was like, after Sputnik, when the Russians, like, were the first ones to put a satellite up, and it was created to say, like, we will not be technologically surprised again. It's behind the scenes thing, but it is very important, right? It's really scary if you get technologically surprised. And so I think that, that is why it's, I think, important from a natural security standpoint, important for the country, important for the species is to, is the rate at which we get scientific discovery. That makes sense. other thing I'm curious I want to ask you guys about I mean God bless Sam and this freaking open and I think but like you know when they started that it was like a pie-in-the-sky research project yeah yeah right this is like almost like Bell Labs he kind of you know like whatever like go for it everyone's like it's bullshit oh what is this nonprofit okay but here we go it's now worth what what's the last round done that happened really great 830 all right so so I to me it signals in fundamental research in industry can be valuable yep and I think that's also a thing that we've kind of forgot about over the last 30 40 years because so much of like where the money was was just like engineering engineering engineering engineering and it wasn't really about trying something that was just like just probably not gonna work but we should be able to try farm has been like that but lots of the rest of the economy has not so that makes sense yep and I wonder if that's gonna change like I'm curious if you think like I don't know every big industry like you know the chemical industry like should everyone just get that be like well based on these models and like some acceleration in science like actually the most valuable thing Dow chemical would do would be some fucking crazy breakthrough not you know let's do another chemical plan or run the numbers on like putting something in Louisiana but like you know like like actually like no no we're gonna like go for it you see that at all like you know like like you like you know I don't know but but that would be like one of my naive hopes is the industrial side of the house wakes back up yeah on doing research yeah so yeah I know you have a point of view on this I would say we've seen a few of these I mean you mentioned Chai earlier yeah I think it's likely to come from researchers on the research of the house that are you know fundamentally rethinking for example the protein design process I think it's my my gut instinct is more likely to come from folks like that who are just like taking really big swings and the tricky thing is just the the song and dance of how they get funded especially though we're still in biotech winter right and like how do they even prove out to the world that like hey we have better discovery engines we have we have better candidates because I think if there's like a real problem in the biotech world is asymmetric information and you just like you can't tell yeah feels Google with isomorphic feels like the closest because it's actually got deep pockets behind it yeah to actually prove that story out but yeah my bet would be a vertically integrated research team taking big swings yeah yeah the challenges like funding these companies all the way through well I think to the point on should there be more breakthroughs more fundamental research I think the answer is unequivocally yes like that is one of the wonderful dividends so to speak that's going to come out of this whole AI wave yeah I think phase one is sort of becoming human level intelligent across a bunch of different categories and I think that will largely go to just doing the things that we do today better faster cheaper I think phase two is going to be becoming super intelligent across specific categories one at a time yeah and that's where all the breakthroughs are going to come from yeah and I feel like we're kind of in this transition phase where we're getting to the point of human level of intelligence across a bunch of different things yeah we're about to start being super intelligent in a bunch of different things yeah and we're about to start seeing a bunch of breakthroughs yeah and so I I think it's like we're within months or years of it becoming really interesting we can example we just back this team that did the alpha chip project at Google so they just like they're using AI systems to actually just design chips that perform better than like human chip those designers can do and so like I think we're gonna see these pop it pockets of super intelligence in different corners of industry and what was the alpha go move 38 or whatever 37 37 I always forget the number I know I was like 83 37 great yes like move 37 right yes yeah yeah I mean if that's true right like and I think there's two ways to get out of by the way my secret on this would be one is the thing you're saying like it's gonna be super intelligence right it'll intuit something that a person wouldn't have and then like I really think if you can solve the problem of greatly accelerating the experimental work it's almost the same yeah the common and torque approach it's you know like that you can't tell me a lot of the the physical world stuff is not simulatable yeah right and so the it's just can you run that machine faster and then importantly like you're saying like can you feed it back and make it smarter based on what it's actually seeing in the world and that okay like and then I think if you're if you start to believe you can get those breakthroughs I do think you have to ask what's the how do you commercial like what's venture look like and that's an area right because you're like oh crazy I got this insane break you know we got a room temperature semiconductor right or you know right like now what right and is it like you go to you go to the big guys you do yourself right like there's a whole thing there I think it's fascinating yeah we have this conversation the other day on will the venture capital industry shrink because the cost of producing everything gets cheaper because it becomes so much more efficient yeah and if we analogize back to the cloud transition when all of a sudden you didn't have to build your own data centers you could just spin up a cloud service yeah you might have thought the same thing would happen but actually the opposite happened there was this explosion in creation yeah which made distribution that much more competitive yeah and so there are a million different companies but then all of them had to fight so hard to like breakthrough the noise yeah and I would I would I would guess the same thing happens which is the cost of creation goes down and down and down yeah the cost of distribution goes up because there are just so many different things out there yeah that could be right yeah do you think language based foundation models are the right substrates or do you think something is going to train like a you know ACG native model yeah so so that is like the arcs evil is an ACTG native model right so it's trained on a trillion bases of DNA and I think that's awesome right I think it's super exciting I think it will apply it'll be like a tool available to the reasoning model to do its job yeah that's already the case yeah right like the reasoning model working on say our even our own eye project could go access alpha-fold design a protein get it synthesized as that as a reagent into the project like that's allowable yeah right and so I think those will end up being powerful tools but I still think I think the reasoning models are really they can do the job of an experimental scientist and so now we have like a thousand experimental scientists in a box yeah that that's already true I don't need like a miracle right like yeah yeah so you're the question yeah alpha-fold all these all these papers have come out on how AI is changing drug discovery yeah do you think the pace of drug discovery is actually accelerated or not okay yes let's nerd out now on like what the hell you can actually do with bio which is like a pile of trash all the time right okay so so like I was right like like I think the down so I got blessed like I basically you know Dressy Park came out when I was 13 all I want to do in life is like make Dress Park engineering is awesome right like like that's why I'm doing this I think I brought back the woolly mammoth yeah yeah I know them well yeah yeah over there colossal yeah that's great I love that stuff right I haven't brought back the boy but yeah but they brought back a dire wolf so the yes okay right so but really what I'm excited about is like the ability for you know kids someday to design biology like they program computers right like that is what I want like that's the world I want to exist and so the question is like how the hell do you get there and one of the issues we have in programming biology designed in a genetic engineering what everyone call it is that the only working app ecosystem is therapeutics there is like a 5% of the market for biotechnology is therapeutics and then there's like 10% is ag remember months on to right like like that's that's engineering in plants and then there's like 5% that's like industrial like you and you a cold water laundry detergent that's a product of biotechnology there's enzymes in there that break up dirt without you having to make your laundry hot and so that was the that's 5% okay that is the totality of apps we've come up with so far for programmable matter compilers i.e. cells it's embarrassing okay right like like the Kate the the fundamentals of this but if you but again like let's imagine computers the only application for computers was drug discovery we'd all be like well there's so you know man such a pain in the ass of these computers right like you know and so so there is like a distinction between like what we're really working on at Ganko and other places like us which is like how do you really make it easier and faster and cheaper to just design biology make it do new things yeah and then the fact of the matter being that the only apps that really like have RLI are drugs yep okay and drugs have like annoying features the biggest one being like the time it takes from like inception of the drug to making money is a killer and it has to do with regulatory yeah like it's like well we don't like to stick things in people yeah be careful like all that stuff is fine and then we can get faster on that like China's also eating our lunch yeah like you might have seen this but like they can do a trial in six months it takes us like two and a few years like like for like phase one crazy Australia is actually eating our lunch I think our FDA will just match Australia's which is great so so we will get faster but it's still not like launching a phone app okay and so that does that make sense yeah so that does remain like like a problem I would say we try there's been a bunch of other attempts at other things you know like animal-free me like it like it is ever quite been good enough to like disrupt another industry yet yeah I would love to see that happen that would be a big accelerant not just for whatever industry but ultimately for genetic engineering And then if you accelerate genetic engineering you you try to create that flywheel that we got in computers We're keeps going going now that said It to pick our app of drugs it has gotten more expensive to develop drugs not less year over year for the last 25 years So that's not great. That's the opposite of what should be happening, right? And so and why is that? Because we do it manually That's my opinion. We have we have not like and it's like bomb all's cost disease or whatever like they scientists are getting more expensive The rent is getting more expensive. That's it. They're actually more productive like we give them new tools Like we are getting like slightly better, but the majority of the cost is manual work. Yeah That shit does not get cheaper and so so that is what I think is the root of it actually and that's why you see me 15 years in of this being like retrenching to solve that problem first because I don't believe we really get out of the mud Until we've got the people out of the lab Then from that base we can start to climb out. Yeah, right now it's fully automated You can do all kinds of crazy stuff and eventually it looks like chips someday, right? It's like alien technology. You know, you're doing it in some way that humans never remember before chips was vacuum tubes It was like human scale electronics Yeah, yeah, and then we we were like okay cool like let's get you know We we saw the curve that will happen for lab work and genetic engineering I promise you but the first step is like put down the vacuum tubes, right? Like get onto some system that does not need people in the middle. How does the application space change? I don't know that's very unpredictable Well, I'll tell you something because I think I'm excited about the near-term You're familiar with like the GLP one drugs. Yeah, literally where it's closer to trillion dollars. That's great like that That is in my opinion like a consumer product. Mm-hmm. I'm going to I'm on the clips. It's awesome It is like the best things. It's the iPhone you don't think about food It's like you can like you get to spend your day thinking about work or kids. I don't know what you want to do You don't have to like think about gonna intermittent fast through lunch or I'm gonna be obese, right? Like you can like you can just get your will power back. It's awesome. Okay, right awesome But that's to me a the reason it's worth so much money is because it's not treating a disease Right like the the biotech industry thought the therapeutics industry today is really the disease industry. Yeah Right and how much of your life and again depends on the person, but how much of your life do you have a disease? It's like your small amount of the time How much of your life do you want to like weigh 15 pounds less? How much of your life do you want to sleep better? How much of your life do you want to have more muscles? How much of your life do you want to feel better? Like the applications in the consumer space. Yeah for biotech are Bananas. Oh, it adds two years to your lifespan. What's that worth? Like what is the value Of a biotech product that adds five years to lifespan This is Sequoia capital throw a number at it Depends on your customer 50 trillion like it's it's infinity. There's no limit on the value of something that would like stack Extra of years of healthy life on the people's like like it's nuts. Yeah, right? Like you know like we it's that's effectively what our health Here says it was trying to do and it's like they give like the total consumption cost of that. Yeah, right? So if you could have that in a pill and a shot So so that but right now today we don't even have a good pathway to get something like that approved Right because all of the regulatory the FDA and everything is is oriented around treating disease. Yep And this is actually where like all the people like oh maha Like like I think like that that line of the maha thing which is like hey actually It's about not just about disease, but about being healthy. Yeah when you don't have disease I think it's really good. Yeah, like I think that's a really good thing for the industry And so I do think you'll see that That set of things um happen and so that's one half. It's like new drugs there and then the other one What our first investor out of well? I see you know who was our angel Mr. Brian Johnson, okay back You know, yeah, there's like Pudgy VC Brian. No, yeah, of course. Yeah, uh-huh, not like you know longevity like you know Yeah, yeah, he was like like a normal person right like yeah, I got like jack. He's awesome now right like they yeah, and so but like What he's done What's interesting about what Brian's done is he has normalized the monitoring like it's like how are you? You know like these are like all these interventions like you're like well, you know like like Brian's got a good life You know like would you be oh you're taking some random thing and trying it out like is that scary and he's like wow I'm monitoring all the time Right, so like every week. He's like taking all these tests and everything else is the most measured person two million dollars a year of diagnostic stuff right like That is the other area. So like oh, we all love our ore rings and everything. This is pathetic. Okay, right Like it's great. Get heart rate like it's like telling you nothing all right like like the real and I love or by the way I've had it for 10 years, but like the Real meat of what's going on inside your body is molecular. Yeah So what we really should be doing is like taking a blood sample every week and giving you like a whole readout about ton of stuff Like longitudinally over time so that you can try different interventions for you and see you how it affects you Luckily, because that's what actually matters like molecular like aging is molecular, right? It's not you're fucking whatever That the uh, and so so like that whole world These stuff like functions getting going doing quest tests like I mean, I got it in it over Christmas But it's like 10 miles. It's like the worst experience in the entire world, right? Like like that There's the at-home stuff now too, right? Yeah, but it's made it so early, right? That's my point Like so I think that line is another place that could be a big if you're asking about near-end apps That one. Yeah, is the other one, right? And so I think I think you could see that I think you could see Other things like the glips Those are ones I'm excited about awesome, and then I'm always hopeful Jurassic Park something like that. Yes, you know right like like uh, that there'll just be some other weird thing And we did just launch a cloud lab service that can go where you can um Like we have experiments as cheap as $39 that you can just run and we don't send you anything like we won't send you a sample But we will send you back data So it's like you do the experiment. We run the experiment for you. You got the data back, right? And so my my last point on this one is like I think if you like science is thought of as this very like precious Genius thing, but really what it is is like formalized human curiosity It's like a process by which humans of which all of us are curious about things Like do curiosity right like really try to answer our curiosity. That makes sense Yeah, but I and I think everybody's curious and so I believe that if you drop the cost of like like like a lot of what blocks people from science Is actually not like the esotericness of it. It's in my view the lab. Yeah It is that that is brutal, right? Like you don't have acts a it is a total gate kept like you cannot get access to one It like almost legally you can't get access to one, right? Like and and so there's just this whole thing and I'm like, well, what if that went away? What if people every day of people could go to order an experiment What if the model would help them design the experiment to ask a question that they have about the world? Would they suddenly ask questions and do these experiments? Would they be what everybody would millions of people want to be scientists? Yeah, and I know that sounds like that's nuts But like if you were one in the clock god bless the looking Valley and the computer industry To the 1960s yep when it was IBM and it was main frames and you told people That kids Would program computers. Yep. They would say you're fucking insane And so I believe if you do manage to drop the cost of all this stuff You may have kids and everybody else wanting to do ask original scientific questions and being able to do it And that would be a cool market, right? Like and so anyway that that All this stuff I feel is on the other side of getting to say I for science stuff working but I'm excited about it Extremely cool vision for the future Great note to end them. Thank you. Yeah very inspiring. Thanks for having me on. You