01:51:35 · Praveer Melwani, Jordi Hays, John Coogan
[01:51:35] S4: that he's [01:51:35] S2: doing. And now I saw for the first time this week somebody like say, global r and d spend is some number of trillions of dollars. Like AI can easily capture 2030% of that. And that was the same exact like logic of like looking at like white collar work which was like, oh [01:51:55] S1: Yeah. [01:51:56] S2: This much spending. So of course AI... Because AI can do some percentage of that. Mhmm. And the question I I keep coming back to that I'm curious to get your thoughts on is, say I am a drug... You know, I'm in the drug discovery business or I'm a pharma company and I come up, like, I can use the models to come up with like a thousand new ideas. Mhmm. There... We are so far, it seems, from being able to simulate the human body perfectly enough that you can like close the loop where it's like, I have a thousand new ideas for drugs and now I'm gonna simulate the trials for them and tomorrow I'll know if we have a a hit drug. Right? I assume that eventually we would be able to like simulate human biology well enough that we don't need to run a five year human trial to figure out if our idea is valid. Right? But like the problem is right now, if if I'm Eli Lilly and I'm using, you know, the latest models, great. I came up with a thousand new ideas. That's like potentially, that's like many many billions of dollars of new money that I need to then go spend on figuring out if they're valid. And so you're back to this idea of like, well, we can only take so many shots, you know, we can only actually run with so many of our ideas. Mhmm.
01:51:56 · Jordi Hays, D.A. Wallach
[01:51:56] S2: This much spending. So of course AI... Because AI can do some percentage of that. Mhmm. And the question I I keep coming back to that I'm curious to get your thoughts on is, say I am a drug... You know, I'm in the drug discovery business or I'm a pharma company and I come up, like, I can use the models to come up with like a thousand new ideas. Mhmm. There... We are so far, it seems, from being able to simulate the human body perfectly enough that you can like close the loop where it's like, I have a thousand new ideas for drugs and now I'm gonna simulate the trials for them and tomorrow I'll know if we have a a hit drug. Right? I assume that eventually we would be able to like simulate human biology well enough that we don't need to run a five year human trial to figure out if our idea is valid. Right? But like the problem is right now, if if I'm Eli Lilly and I'm using, you know, the latest models, great. I came up with a thousand new ideas. That's like potentially, that's like many many billions of dollars of new money that I need to then go spend on figuring out if they're valid. And so you're back to this idea of like, well, we can only take so many shots, you know, we can only actually run with so many of our ideas. Mhmm. [01:53:16] S0: Oh, that's right. I mean, that's that's the bottleneck is Yeah. The only way [01:53:20] S2: to find out if things are safe active is in a human. But like, was the bottleneck ideas? [01:53:25] S0: No. The bottleneck is not ideas. Yeah. The bottleneck is translation.
01:53:25 · D.A. Wallach, Jordi Hays, John Coogan
[01:53:25] S0: No. The bottleneck is not ideas. Yeah. The bottleneck is translation. [01:53:28] S2: But but like the potentially exciting thing is that as you see like things like Navier Stokes fall, maybe there's just things that humans have been missing and like we'll get a bunch more new ideas. And again, we're not bottlenecked on ideas, but maybe we'll just get some like better ideas. Is that... [01:53:46] S0: Well, we almost certainly will get better ideas. And as with all the other AI opportunities, it's both going to come from the AI generating ideas and from the AI giving leverage to really smart scientists who were trying to come up with ideas before but now are going to be supercharged in their effort to do that. I think the sort of challenge right now, even before AI, was that we have a super abundance of good ideas for new drugs. And you can go look at any university in America or now in China and Europe. There's great science everywhere. [01:54:20] S1: It happens every time a new blockbuster drug comes out. I do the research on where it came from. It's like, well this guy invented it in the eighties and then it was used for some really small thing in the nineties and then they finally got it to scale in the twenty ten's. [01:54:33] S2: Or they were using it for this other thing. Realized it was actually [01:54:36] S1: having [01:54:36] S0: Yeah. Another The GLPs are one of those stories. Yeah. And it's kind of an indictment of the startup ecosystem, the biotech world. That didn't come from a startup. [01:54:46] S1: Oh,
01:54:46 · John Coogan, D.A. Wallach, Jordi Hays
[01:54:46] S1: Oh, [01:54:46] S0: yeah. Right? I mean, the story in big pharma has been all the new drugs come from M and A. [01:54:50] S1: Yep. Totally. [01:54:51] S0: But the GLPs are the biggest and They they were like a twenty year hard won project inside of big pharma. [01:54:59] S2: It came from the Gila monster. That's right. Exactly. [01:55:05] S0: I mean, look, there are a ton of good ideas out The challenge is how do you prove whether they're safe and effective in humans, which is, to your point, it's very time consuming, it's very costly. And AI, from the standpoint of a big pharma company, is going to chip away in incremental ways at a million different things [01:55:27] S1: Speed up data analysis, speed up capture of notes from the trial. [01:55:32] S0: It's everything. FDA filings, It's legal going to be incremental and you're going to add up a lot of tiny efficiencies to accumulate and it's going to look more like, oh, maybe a drug's development time goes from twelve years to eleven years and then maybe to ten years. Now, the average blockbuster drug is selling, I think, 3,000,000,000 a year. So more or less every day you can accelerate its time to market is worth millions of dollars Sure. For a sponsor. [01:56:03] S1: Yeah, it's [01:56:03] S0: valuable. So anything you can do with AI is going to be a great thing for the industry. But is it going to cure cancer overnight? [01:56:11] S1: Yeah. [01:56:11] S0: Is it going to solve aging overnight? Yeah. Those are sales pitches.
01:56:11 · D.A. Wallach, John Coogan, Jordi Hays, Keller Cliffton
[01:56:11] S0: Is it going to solve aging overnight? Yeah. Those are sales pitches. [01:56:14] S1: How are you set up to invest? Do you you trade in the public markets? Do you have a fund? Do you have restrictions? Like where where do you see opportunity because the GLP one trade again didn't really happen in the private markets. Of course, were a lot of private market beneficiaries. But how are you investing structurally? [01:56:34] S0: We do both. So we have a private fund and then we've got a hedge fund. [01:56:36] S1: Okay. [01:56:37] S0: And you know, the the... Where the risk return is more attractive Yeah. Changes over time. [01:56:42] S1: I can imagine. [01:56:43] S0: Again, this industry is very sensitive to rates. [01:56:45] S1: Yeah. [01:56:46] S0: So when you've... You know, when you're in a secularly rising rate environment in Yep. General it's tough for biotech. Yep. You think about it... [01:56:54] S1: Because the earnings are falling So we [01:56:55] S2: need to go back to ZERP for humanity. [01:56:57] S0: We need to go back to ZERP. For humanity. The bottom line. And I'm buying all the long dated treasuries I can but it's not moving the... [01:57:05] S3: Yeah. I feel like [01:57:07] S1: a Fed chair who came out and was like, we need to lower rates to save lives because we're going to cure cancer if rates are low. That's a really effective message. [01:57:15] S0: It's a good pitch. I think it's better than like, oh, need AI to add 2% real GDP growth a year, which to me is tall order. [01:57:23] S1: Was I was... Tall order.
01:57:23 · John Coogan, Jordi Hays, D.A. Wallach
[01:57:23] S1: Was I was... Tall order. [01:57:26] S2: What's going on in the like biology and pharma data markets? Like is there is there That's a good question. Is there old companies that were dead and long... Effectively long gone that are being like, you know, that are suddenly worth something because they they had a bunch of, you know, they did a bunch of trials on something and that data is valuable and then I have a follow-up question. [01:57:48] S0: Okay. So when DNA sequencing first kind of hit the scene several years ago, there was this first wave of data is valuable as a market story. And I would say... [01:58:01] S2: It's new oil. [01:58:02] S0: Yeah. It's like the new... If we do a start up and all we do is we generate a boatload of data, maybe some big pharma's going to come and buy us for that data because they're going to mine new drugs out of it. There's really, to my knowledge, only one or two instances of that working. Amgen bought something called Decode, which is really interesting. Decode was a sequencing project for basically everyone in Iceland. Oh, [01:58:26] S1: I've heard of this, yeah. [01:58:27] S0: Interesting, quite genetically homogeneous population. Sure. That was good in some ways in terms of what insights you could derive from [01:58:34] S1: it. Yep.