Welcome into the Inside the Ice House podcast, our second annual AI Biotech series. Today's guest is Jason Kelly. He is the CEO of Ginkgo Bioworks. Jason, thanks so much for joining us inside the Ice House. Happy to have you here.
Yeah, happy to be here, Lance.
So, at its core, when you and your co-founders decided to start Ginkgo Bioworks, and I should mention we're obviously talking here on the day of the AI Bio Summit at the NYSE. When you started, when you were thinking of starting Ginkgo Bioworks, what problems were you all trying to solve, and what is the ultimate vision that you're still working towards this day?
Yeah, so I started Ginkgo out of MIT. So, I did my PhD in bioengineering at MIT, and my co-founders did, too. We met there, there was five of us. I don't know if you have had a chance to work in a lab at all in your life, Lance?
I have not. I was a journalism guy in high school.
Okay. Well, you're lucky.
So, it turns out you want a PhD at MIT in bioengineering, it is five years of standing at a lab bench, very carefully moving clear liquids around, trying desperately not to make a mistake setting up your experiments.
Well, your precision must be just like phenomenal.
No, it's horrible, in fact.
You must have very steady hands.
Yeah, you know, they say that, though. Yeah, actually some people in the lab are known to have, quote, good hands. All right, that's like a thing.
Are you one of those people?
Yeah, and so that little bit leads to why we started Ginkgo, right? So, our view was, if you look at bioengineering, and I can talk a little bit about even what that means, but come back to it. The limiting reagent into it, like what slows it down, is the lab work.
All right, and we believe that you can take a lot of that work and automate it, so that scientists can spend their time thinking, and just order what experiments they want from the lab bench. All right, and so we could either then commercialize that by selling you the robots, or we could be like an AWS running it for you in the cloud, but the idea is you should yourself stop doing that manual work and free yourself of the bench.
And so you bring up this engineered biology.
Where do you think... Well, you can first get into it, what is engineered biology, and then once you do, where do you think sort of engineered biology has the greatest opportunity to create real-world impact?
Yeah. So, all right. I'm a product of Jurassic Park. So, that book came out when I was 13 years old. So, I'm sort of the second generation of biotech people, right? Like Michael Crichton writes that because Genentech is out and these first-generation biotech companies making these drugs and he sort of jumps forward and says, 'Well, what if we could really just design DNA?'
And I think the thing like to help put a point on it, if you think about the technologies we have in the world that run on code, there is exactly two.
Computers and cells. All right? In computers, it's zeros and ones, in cells, it's A's, T's, C's and G's. But everything else, like if you look around this room, this table, the chair, none of this other stuff is running on code. Okay? It's just like the phone in your pocket and you.
All right. And so, what's amazing about cells compared to computers? And by the way, like running on code is very powerful. So, if you download a new app on your phone, new code goes in and suddenly magically like you order a car or you can talk to it and it can make the same thing do lots of different stuff. That's the power of like a code-driven product.
The limitation of computers is they only move information around. So, if you look at what they've disrupted, right? Media, finance, telecom, things like that. What have they not disrupted? You know, I don't know, like hamburgers and building materials and fuels and, you know, all the atoms-based things.
Well, a cell is programmable like a computer, but it doesn't move information around. Biology moves atoms around. So, the reason I get excited about bioengineering in the long run is it's sort of our atom compiler in the way the computer is our information compiler. And we have absolutely, we're in the extreme infancy of using biotech. And it's so early that we basically equate it today to developing drugs because that's the first app that's come out on top of, in my opinion, like the matter compiler computer of a cell. Well, the only thing we've really been able to make money doing with it yet so far is disease treatment. But if you think of it in the long run it should be touching every physical goods industry if we actually got good at programming these things.
So you say we're in its infancy.
What does the process look like toward getting out of the infancy into, if you equate it to baseball, a middle innings?
Yeah, yeah, yeah. End of game eighth, ninth, you know, maybe even extras.
Let's get to the third inning. Yeah, that's right. Oh, yeah, okay. So this is a great question. So let's say you want to program a cell.
All right. So example would be like a yeast cell like you would use to make beer. Like a very easy thing to do. You learn to do this as an undergrad in school.
Super simple. Just pick up a pipette and get to work. So the way it works is that the cell has a genome.
All right. So like a bacterial cell for example has like a million to 3 million letter genome. You have like a 4 billion, 3 billion whatever letter genome, all right? And so you identify what you want to add to that cell. All right. So let's say it's like a, you know, a brewer's yeast. Let's say you want to add a gene. The very first drug product was human insulin. So you could look up the gene for human insulin, go on your computer, type the ATCGGG or whatever of the gene, hit print. The DNA gets made. So we can chemically make any DNA sequence you want now. It's called DNA synthesis is the technology. It gets made and then you heard of like CRISPR?
Okay. So CRISPR is, it's actually pretty simple. It's basically find location in genome,
cut, open. All right, so now you have gone to a place in a genome and there's other tools that do this too, whatever. If CRISPR's only used for certain things, but you open the genome and then you stick in your human insulin gene. Plus, you've added a little thing on the front called a promoter that turns it on. And then you close it all back up and put the cell in a tank in South San Francisco in 1978. You grow it up and you start making human insulin for the first time where diabetics used to get it from pigs. And so that's like the simplest thing. I've just taken a cell and I've made one little change, make this protein for me. Does that make sense?
All right, here's the problem. Why are we in the first inning and not the third? Humans did not design cells. All right, this is the key distinction between computers and basically everything else, cars, planes, the whole shebang, and bioengineering is it is the one discipline of engineering where we are working with basically alien technology.
Okay, right? Like not invented by humans. Does that make sense?
And so it can run our code, we can put code in, but you don't know what half the parts do. It's like you had this computer landed from Mars or something on a meteor and you could put a disc in it, but you had no understanding of the operating system or anything. You couldn't read the code. It was a bunch of zeros and ones. It's like, 'Would you like to write some zeros and ones and put them in?' And so we've been essentially trying to figure out how cells work for the last, you know, whatever, 100 years and then since, you know, the late '70s, early '80s, we've been able to insert our own code and be like, 'How about this? How about this?' But most of the time it doesn't work because we didn't design them. Does that make sense?
So is that a lack of knowledge essentially?
Well, now let's talk about... Yeah, so that is part of it.
And that's half of it. Okay, so there's basically two ways we get to the third inning. One is we greatly improve our knowledge of the cell.
And a lot of what you're hearing at the conference we're doing today near the Stock Exchange is, could I have an AI model that understands a cell, right? Like can I put the whole thing inside a model and then now I can simulate it. Like I can simulate an airplane or like I can simulate a semiconductor chip. I think that's great. I think people should try. That's like one approach. That's what you just suggested. Improve our knowledge.
The other one is, well, we can try today. Like we work in the lab and we stick some DNA in and we see what happens. It's trial and error.
By the way, this is kind of how we write software.
Right? We don't simulate Microsoft Windows or something. We let a bunch of people get in there and go at it. Right? It's like it's high throughput trial and error. That's how we write software. Maybe it's like that. That's how our other coded thing works. So, could we reduce the cost? Because now this is why we're not in the third inning. I just said, 'Hey, make that DNA.' We pay by the letter.
So, imagine if you were writing a new piece of software and you paid for every zero and one.
You wasted all that money.
Yeah, it's expensive. And so what we are doing, like our bet in this whole thing, is if you reduce the cost to compile and debug, essentially to run through cycles in the lab, we can also get to stuff that works through like more rapid trial and error. We also want to be smart about it, but let's just be honest. Software benefited from the fact that you could have like monkeys at keyboards just banging away and it would go quick. Right? And you could just try. And if it didn't work, it didn't work and you had debuggers, you had all this stuff. We've got to get there with bio, but that's the Ginkgo bet. Is we can solve it on the lab side. A lot of other people are also trying to solve it on the knowledge side.
And so where does in all of that autonomous labs fit in? Because I know that's a focus of Ginkgo. Where do those fit in in the greater journey of getting to that, you know, third inning that we were talking about?
Yeah. So, then if you now start to say like, 'Well, why is it so expensive to compile and debug genetic code?' Is it that the chemicals are expensive? Like the A's and T's and C's and G's, are they expensive? Actually, no. In fact, we get them like from salmon sperm or something. Insane. Like you go collect this stuff.
It's got DNA in the environment.
So you can get these chemicals pretty cheaply. The reagents at low volumes for everything you need to do aren't that expensive. The equipment's expensive, but actually you can do a lot of throughput on a piece of equipment. It turns out when you really dig in, the expensive part of the compiling debugging are two things. One, the technicians, like the people doing the work,
and two, the labs themselves. Like you have regulated space.
Right? Like you have a BSL-2 laboratory. People are wearing lab coats. You need air changes. Like it's got to be safe for the humans in the room. And so those are your two big costs, it turns out. Which is crazy.
But when you really get down and you look at it like an engineer, that's the dominant cost. And so we believe with an autonomous lab, and the idea behind autonomous labs is like an autonomous car. All right? You sit in the back of that Waymo, you're like, 'Hey Waymo,'
'take me to my house.' And it does it with no human driving.
