Latest / CareTalk: Healthcare. Unfiltered. / Anthropic’s Big Bet on the Future of Biotech

Transcript

Anthropic’s Big Bet on the Future of Biotech · CareTalk: Healthcare. Unfiltered.

• 0:00
  Anthropic just purchased a one-person biotech running an AI agent that coordinates 11 outside labs.
• 0:07
  Is this the blueprint for the one-person unicorn biotech, or is it just a snazzy orchestration tool that's gonna run into the usual bottlenecks?
• 0:16
  And while we're at it, how does Anthropic's entry affect its deals with pharma customers and the competition between the US and China for drug industry supremacy?
• 0:33
  Welcome to Care Talk, America's home for incisive debate about healthcare business and policy.
• 0:37
  I'm David Williams, president of Health Business Group.
• 0:41
  And I'm John Driscoll, the chairman of the UConn Health System.
• 0:45
  David, what exactly is anthropic?
• 0:47
  Is that a class- … at Wesleyan where your, your esteemed university… Maybe give people some context here for the AI uninitiated.
• 0:56
  Yeah.
• 0:56
  Well, I don't know who those are, uh, John.
• 0:58
  It wasn't the class at Wesleyan when I attended some years ago, but maybe they were looking into the future.
• 1:02
  I mean, anthropic- Oh, that was before the invest- inven- invention of computers, right?
• 1:08
  They had a mini computer, John.
• 1:09
  It was a, uh, DEC, um, it was a DEC machine.
• 1:13
  I'll try to remember what it was called then.
• 1:14
  It was, it was the one that, that was after the PDP-11.
• 1:17
  Uh, it was the next one that was the super mini.
• 1:19
  In any case, John, Anthropic, as, uh, you'd have to be under a rock, and even then you might, uh, hear about it, is the hottest- I've got rocks around me the hottest company, uh, on the planet.
• 1:29
  You know, it's a big, big AI company, and it's recently filed to go public, and it's claiming these just unbelievably gigantic market shares.
• 1:35
  So Anthropic is just huge.
• 1:37
  And John, whenever a company's going and, and looking at, um, a market size that's as big as they are, into the trillions of dollars, it kinda means either brand-new things have to be invented, but also they kinda have to take over existing markets.
• 1:50
  And in this case, one of those markets that we're gonna talk about today is the life sciences market, and specifically pharmaceuticals.
• 1:57
  If you'll permit me, David, I think I, the, the, the description of Anthropic was somewhat incomplete.
• 2:03
  Anthropic is one of the frontier artificial intelligence labs.
• 2:08
  Labs meaning they are creating the big, massive models that can appear to answer any question and start, and are starting to show signs of reasoning, that have been gobbling up capital, that rely on data centers and rely on advanced chips to basically create thinking and reasoning engines.
• 2:31
  And in order, because of they're incredibly expensive to build, but also incredibly insightful because the, the, the, they're the power behind open evidence f- that doctors are using or ChatGPT, which is the OpenAI labs, or Claude, which is the primary tool that Anthropic has deployed.
• 2:51
  They need to… They're, they're, they're, they're creating these massive thinking platforms that we're all relying on, whether we pay for them or not, and increasingly are part of workflow for many people, including in healthcare.
• 3:06
  Um, and now they're taking on new industries.
• 3:09
  To your point, they're, uh, the Anthropic is going public, I think, like soon.
• 3:12
  I'm not quite sure when.
• 3:14
  In order to justify a big valuation, a trillion dollars, I think it's north of that, they're looking to be clear about what industries they're going to use those powerful models to then not just deploy, but to build new businesses, and that's what we're gonna talk about today.
• 3:31
  Can they transform, and how fast, biotechnology, biopharma, to create, uh, medicines that will be transformational and can potentially think, cure things like cancer?
• 3:43
  Well, John, they went ahead and bought a company.
• 3:45
  Now, not a big company.
• 3:47
  It's a company with one employee, and I think it's called Mythril Medicine.
• 3:51
  And there's a- Mythril?
• 3:52
  … Mythril, I don't know.
• 3:53
  It's, uh, not a word that I had heard of.
• 3:54
  It's one of those things if you had a gap between your teeth, you know, you'd have to, uh, to, to try to, try to pronounce it differently.
• 4:00
  But there's a founder who goes by Professor Oak, and he used Claude to orchestrate 11 different CROs, contract research organizations, and got to some decent preclinical data in under 18 months, which is pretty fast.
• 4:12
  They also im- Massive.
• 4:14
  Yeah.
• 4:14
  Massive win.
• 4:16
  Anthropic also bought another company, um, uh, Coefficient Bio.
• 4:20
  They're opening, uh, a wet lab, uh, in the Bay Area, and they've got partnerships with six of the major pharma players, including Bristol Myers Squibb.
• 4:29
  And venture capitalists are starting to invest with the idea of, "Hey, I can sell my company, not to Merck or Pfizer, but maybe to Anthropic." So that's kind of what's going on.
• 4:38
  Um, I don't know.
• 4:38
  Does it mean anything?
• 4:40
  Well, I think it means, it means that the… Well, there's two key implications.
• 4:44
  One, that these models c- might be the AI platforms might be directly relevant to transforming the way we assess, invent, and then deliver medicine in a much faster, more personalized, more scalable ways.
• 5:00
  I mean, the fact that one dude could, you know, it used to be it's like a magic trick that some guy, a m- a man could, it's like sort of lo- the, the jokes about the one-man band.
• 5:11
  But what the frontier labs create are the opportunities to create quasi or totally independent agents, sort of thinking units that then can work together, it c- what's called an agentic layer, to actually execute tasks, cr- pursue business, do research, and potentially drive insights, and potentially drive manufacturing of new drugs.
• 5:36
  It's pretty wild to think that, um, the, this, you know, the, the, the $2 billion and thousands of people in multiple companies that it normally would take to create a drug might be solved with a bunch of, uh, digital bots, if you will, that are agentic, that don't need any of that.
• 5:58
  And, and I guess, Dave, what's the s- what do you think?
• 6:02
  Because historically our The biopharmaceutical industry, people know about Moore's law, which is that w- as we, uh, that computing power and cost goes up by factors, you know, every few years.
• 6:16
  It gets twice as fast and is half, and costs half as much.
• 6:20
  If you map computing power and cost, you see this curve of incredible yields and low- and, and lowering costs.
• 6:27
  If you look at biopharmaceuticals, it's increasing costs and increasing time.
• 6:34
  Uh, it's the, it's a literally a reverse Moore's law, and this could completely reverse that, I think.
• 6:41
  Isn't that, isn't that the implication of what you're suggesting?
• 6:44
  I guess a reverse Moore's law would be like, uh, Eroom.
• 6:47
  Is that, is that Moore spelled backwards?
• 6:48
  So, uh- … maybe, maybe that's what it will be.
• 6:52
  John, one of the things that's very hard, and I- I'm glad you brought up Moore's law, because it's very difficult for the mind to comprehend exponential changes.
• 7:00
  You know, we can go and say- Are you talking about me?
• 7:03
  No, I think you could do it, but I think for… I say it's difficult.
• 7:07
  But it's hard.
• 7:07
  It's hard, John.
• 7:08
  So it's hard to look and y- and you say like, "What's the implication?" You can look and say, "Wow, these bots are gonna do everything." I don't know, but the-- but one way to look at it is let's, let's understand what's happened already and what did those steps tell us, and maybe in these particular, you know, acquisitions.
• 7:22
  So and from the, just from what's happened with these companies and what it means for the business strategy of Anthropic, which still has to go out and raise money from, from people.
• 7:30
  So they previously said they were just a software platform, not a drug developer.
• 7:34
  Now it's actually owning assets, so that has some, you know, potential conflicts with its partners like BMS, Bristol Myers Squibb, and then there's concerns of, you know, things they might be concerned about their IP leaking out, overlap of therapeutics.
• 7:48
  And then I, I'm always interested in these things, you know, like this, the idea of the one person biotech.
• 7:53
  Well, what does that mean?
• 7:54
  So first of all, there've been-- there's a lot of virtual biotechs that are out there.
• 7:59
  You know, we work with some of them.
• 8:00
  Actually, we work with one that has no employees, if you can believe that.
• 8:02
  But, um, it's pretty common, you know, to go and coordinate external resources.
• 8:07
  The CROs, contract research organizations, are outsourced and so, you know, so often a company with one or two people might be able to orchestrate them.
• 8:16
  Now, the software is accelerating that but, you know, that is just-- this is just preclinical.
• 8:21
  This is a mouse, it's not a person.
• 8:23
  So the physical biology, the wetlab throughput, you know, those are bound by just whatever execution speed you could have.
• 8:30
  It reminds me a little bit of something from 20 years ago or so when there was high throughput screening, when you're gonna be able to go through all these different molecules very quickly and try to find the right targets.
• 8:39
  So it sped up one part of the process, but we, we haven't-- it hasn't made a dramatic, such a dramatic impact, um, on the rest of it.
• 8:47
  So that's what I'm seeing at the moment as the next step.
• 8:50
  There's two big things that are embedded in what you're saying.
• 8:54
  One is that, is the question on valuation and Frontier Labs as software valuations and stuff.
• 9:01
  But the problem is they cost an enormous amount of money to build these massive platforms that, that make thinking and, and questions and response and, and creating of agents really cheap.
• 9:11
  It's only cheap for us because they've invested all this money.
• 9:15
  So I actually think that they're gonna have to get into a lot of vertical sectors, buy, build- verapamil succinato de metoprolol carvedilol bisoprolol propranolol amiodarona propafenona digoxina adenosina ivabradina sacubitril-valsartana hidralazina dinitrato de isossorbida nitroprussiato de sódio rivaroxabana apixabana varfarina enoxaparina clopidogrel ticagrelor ácido acetilsalicílico alteplase tenecteplase rosuvastatina ezetimiba salbutamol fenoterol formoterol salmeterol brometo de tiotrópio brometo de ipratrópio budesonida beclometasona fluticasona montelucaste omalizumabe mepolizumabe roflumilaste pirfenidona nintedanibe DPOC VEF1 índice de Tiffeneau CURB-65 critérios de Light polissonografia TFG CKD-EPI relação proteína/creatinina hematúria dismórfica cilindros hemáticos glomerulonefrite pós-estreptocócica nefropatia por IgA doença de Berger glomerulonefrite rapidamente progressiva síndrome de Goodpasture nefrolitotripsia percutânea alfaepoetina sevelâmer omeprazol pantoprazol esomeprazol domperidona metoclopramida bromoprida ondansetrona sucralfato azatioprina infliximabe adalimumabe ustequinumabe Well, let, let's-- I'm, I'm with you, John, and I'm, I'm… Let's, let's talk about a couple things.
• 11:44
  So I'm glad you brought up the, you know, the CF, cystic fibrosis example from your own personal story.
• 11:49
  I j- I just heard actually earlier this week from my mother that she has a friend whose daughter has CF and is, is, is expecting a baby.
• 11:58
  Yeah.
• 11:59
  That's amazing.
• 12:00
  I mean, literally brings tears to my eyes.
• 12:02
  I mean, it's- Yeah … different.
• 12:03
  Yeah.
• 12:04
  So let's talk about-- Let me step into now phase two.
• 12:08
  So once you, you know, the, the first part of the drug development is where you've got an animal, and then it moves into people.
• 12:14
  And a place where you have a problem is that these animal models, like a mouse, a mouse is similar to a person, but you can't really tell if the drug is then gonna be safe and work in a person just 'cause it worked in a mouse.
• 12:25
  And at the moment, what happens is there's a lot of falloff once you actually get into a clinical trial, meaning with a person as opposed to preclinical with an animal.
• 12:35
  And despite all of what AI can do, it can't actually model that complex human biology.
• 12:40
  However, it can't do it yet, right?
• 12:43
  That could change.
• 12:44
  That could change, and that could make a big-- that could make a real, a real difference.
• 12:49
  So I do think that that is, um, you know, that may be where, may be where it's, where things are going.
• 12:54
  It's not there yet.
• 12:56
  Well, what's interesting is that biology is still one of the hardest things to hack.
• 13:01
  I think we've been pretty effective at hacking physics, physical things, and the w- and, and working within the laws of physics.
• 13:07
  But biology, there's a lot of mystery.
• 13:09
  Um, and that's, I think, one of the great things that, that, the, the great promises.
• 13:15
  Um, and to your point about exponential change, um, you know, the, uh It, it's super hard for us to think about it, and yet I be- believe we're in a bit of an exponential change moment in terms of what these models can do.
• 13:32
  But exponential change only happens to a certain way.
• 13:34
  But again, contextually, you know, the lily ponds and the, and the, and the, and the pond example is that when lily ponds double every day, I think it takes about 30 days to cover a pond, but by day three, you've only got 3% or 4% of the pond covered.
• 13:51
  By day 30, you can't see the water.
• 13:53
  And it's just hard to imagine those kinds of breakthrough moments, what gradually and suddenly really means.
• 14:01
  Having said that, you know, there is some s- I am somewhat skeptical that we can hack biology.
• 14:08
  But Dave, you know, we are in a race with China.
• 14:11
  You know, the, the, uh, INDs are the, the, the, the things that folks file when they wanna get a new drug in the, the, the, the- Yeah, that's the, that's the investigational new drug application for the, with the Federal Drug Administration to d- to introduce a drug in the US, and my understanding is that, you know, about a decade ago, about 25% of them came from, um, China, and that it's currently up to 75% of all of the INDs are coming from China.
• 14:39
  Now, many of those are me too.
• 14:41
  Um, I get that.
• 14:42
  Um, but it gives you a sense of how quickly and how rapidly China's accelerating, uh, their, their move, which is a national priority based on the, the Prime Minister Xi's to kind of basically displace our lead as a country in, um, biotechnology.
• 15:01
  And, you know, honestly, you know, going back to Marty Ma- uh, Mar- Marty Macri, who we knew before he became FDA commissioner, and perhaps we'll get him back, Dave.
• 15:12
  Uh, but this administration actually, I think, has been trying to make constructive steps to make it easier to move through the FDA.
• 15:19
  But if we don't fix, accelerate our industry approaches and fix the regulatory regime, I worry we're gonna fall behind China and, you know, lose one of the best industries, industry leaderships positions we've got as a, as a c- uh, as a country.
• 15:38
  John, the, you know, this, this topic, uh, I don't think it's any kind of exaggeration to say it really brings out all the huge issues.
• 15:45
  You're right.
• 15:45
  Right into China with the central point of economic, um, of economic competition.
• 15:51
  And we hear about, you know, like is one model from China ahead of another?
• 15:54
  But let's talk about it in the industry context because what's happened is that China has gone essentially from not being a factor in drug development to surpassing the US in some ways.
• 16:03
  You're seeing it a lot with their earlier stage licensing deals.
• 16:07
  Mm-hmm.
• 16:07
  And on the one hand, you're kind of saying, "Well, maybe it's not that big of a deal because it's me-too products and all that," but it actually goes beyond that.
• 16:13
  I saw an analysis that was looking at cancer drugs, and they said, you know, these drugs as we know are very expensive, and a lot of them don't have such a dramatic… They make such a dramatic improvement.
• 16:25
  So they're looking back and saying, "Well, what's the cost effectiveness?" And they said- Mm-hmm … of the drugs that were actually making a difference, many, if not most, were actually the ones from China.
• 16:34
  So on the one hand, yeah, a lot of me too, but on the other hand, some of the real innovation is there.
• 16:39
  And John, that has something to do with let's, let's bring all this stuff to the table.
• 16:42
  Immigration policy, John.
• 16:43
  We had people from China, many people from China doing their PhDs in sciences here.
• 16:48
  Yep.
• 16:48
  It used to be unheard, unheard of for somebody wanting to return to China for their career.
• 16:53
  And guess what?
• 16:53
  That is not the case now, and we're making it a lot easier to push them away if they even get here in the first place.
• 16:58
  We've tightened up all the immigration access.
• 17:02
  Um, the federal government in its own idiocy has, has cut funding in the NIH and through a lot of post-baccalaureates, so we've literally lost about a quarter of the PhD slots in the United States across all categories.
• 17:18
  We- so we've restricted access and opportunity for some of the brightest people in the world who wanna come here, and then we've restricted for those who are here the number of slots that are available to apply to.
• 17:30
  We're undermining our lead in basic science by underinvesting in the talent, in the talent development, and the talent access.
• 17:37
  So there's a whole series of things and, and, and I'm, I'm hoping that that'll be reversed perha- potentially s- uh, through some of the elections or rebalanced.
• 17:46
  But I a- but I also think that the, that it's, uh, that our leadership in AI for the moment and our massive investment will hopefully help us regain the, regain and sustain leadership in, in science, um, and particularly bioscience and pharmaceuticals 'cause there are… At, at least theoretically, we should be able to turn a lot of these deadly diseases into manageable conditions with the right drug cocktails.
• 18:12
  So John, let's go back to a company that I brought up at the start of the conversation, BMS, Bristol Myers Squibb, because they are working with Anthropic as an AI partner, but they're also making moves in China.
• 18:23
  So they're acquiring major- Really?
• 18:24
  … Chinese assets.
• 18:25
  Yeah.
• 18:26
  And it's- Oh, so they're balancing their bets.
• 18:28
  So Bristol Myers Squibb is one of the large, like one of the 26 largest biopharmaceutical companies in the world, and was a laggard for a while.
• 18:36
  Yeah.
• 18:36
  That's right.
• 18:37
  So they've been… You know, and who's to say whether this will pay off or not, but they're basically betting on AI in the early discovery and relying on China to be able to do clinical trials, uh, in a speedy way.
• 18:48
  Now there's also, at the same time, uh, talk about policy, we'll talk about the FDA in a minute, but on the legislative side there's proposed restrictions on, you know, Chinese trial data and that's gonna make major uncertainty.
• 19:00
  And what it might do is make it so that, you know, okay, well w- maybe we're protected in a sense, the companies, but it also makes it so that are we, are companies in biotech gonna be globally competitive or not?
• 19:12
  And John, maybe I'll, I'll give an example.
• 19:14
  So most people have heard… Or an analogy.
• 19:16
  Most people have heard about the electric car market in China, right?
• 19:20
  Here we have, like the Teslas are pretty much similar today as they were years ago.
• 19:24
  They were actually leading in China at one point.
• 19:26
  But now if you go to China, you know, it's just unbelievable all the cars that are there, electric cars, and they're much cheaper and they're, you know, a much higher performance than what we have in the US and we can't get them here.
• 19:37
  So is the same thing gonna happen with drugs?
• 19:38
  Well, we actually, we, we could, we actually st- we put up trade barriers so that you can't import those cars which are widely available in places like Mexico and are half the price as Tesla and have twice the, apparently twice the technolo- technological capability.
• 19:55
  So we've, we've protected the US auto company, the slowpoke US auto industry That's a re- I think a reasonable choice for autos in the US.
• 20:07
  But if we were to do the same thing in drugs and biopharmaceuticals, you'd be potentially restricting people's access to healthcare.
• 20:13
  I mean, I, I talked to a, a friend who's a, who's, who's, who's a doctor, uh, in, in P- you know, the People's from the People's Republic of, of Massachusetts, um, who's involved in the Mass General system, and one of his oncologists suggested that he take his father to Shanghai because some of the novel stuff was more available and just, and not that expensive.
• 20:34
  And you know, it's obviously a, a … That they had, they, that, that he… That it was for, for a stage four, uh, condition, that they were doing more things, and he had a couple of targeted therapies that he was recommending.
• 20:47
  I mean, and I think you're gonna see more of that, and that's the vaunted Harvard system from Massachusetts.
• 20:52
  Well John, we used to have the Boston-Austin access, and now I guess it's one People's Republic to the other.
• 20:57
  Let's, uh, let's, let's, let's finish off by talking about FDA.
• 21:01
  And- Sure … there have been, there have been moves there.
• 21:03
  And you said Marty, uh, somebody that we actually have on the pod- Lots of moves… on the podcast here.
• 21:08
  Go, let, let me, let me know a couple of those moves other than sort of like getting out of the way of, uh, oncoming whatever it may be.
• 21:14
  Okay, I have one.
• 21:14
  So usually what happens when a company goes from phase two, uh, to then phase three… So phase three in clinical trials is your big phase where you have a lot of people, and it's your phase- Do it right, right before, yeah, people right before launch.
• 21:29
  And usually they- Mm-hmm … require two trials In order to make sure that it's actually, you know, should get to market.
• 21:35
  And there's been a move to make it so that only you need to have one.
• 21:38
  So that makes a big difference because those trials are expensive, um, you know, and take quite a long time.
• 21:43
  And the FDA itself is, is implementing AI, as it actually we've talked about on this show.
• 21:47
  The point of the stage three trials is to make sure that in addition to the efficacy it's showing at the bench and in animals, that it's effective in humans and safe in humans.
• 21:57
  And I think that's a sensible move, Dave.
• 21:59
  I, I, the, the, for, for drugs that have reached a certain point based on everything we know.
• 22:05
  I, I'm excited by that.
• 22:07
  I, I'm a little al- alarmed by how they've talked about the use of AI, and yet it doesn't feel like there's been as much disclosure of how this administration appropriately has embraced AI, I think, as a transformational tool.
• 22:24
  But I've got to think there's a long cost to the alienation of the federal bureaucracy.
• 22:30
  You know, they came in firing people somewhat randomly.
• 22:33
  They fired some departments at the FDA and then had to hire them back.
• 22:36
  Um, there was a demonization in the press of, of, you know, by Elon Musk and DOGE and others.
• 22:43
  And I've got to think for regulators that think of themselves, particularly as the FDA as q- as a quasi-independent authority and kind of acted a little arrogant and slow-moving because of it, that, uh, that, that, that it's probably been harder to get alignment there than other areas.
• 23:02
  But certainly the possibility of using AI, because in order to get a drug regulated and mo- and monitored, it's enormously data intensive, that AI would be a natural accelerant and hopefully drive new insights.
• 23:20
  But I, um, I, I, I, I, I just think that the demonization of the federal bureaucracy has got to be probably slowing things down, because I haven't actually heard that they've made a great deal of progress on some of the sensible reforms I think that they had talked about initially.
• 23:35
  That's fair, John, and probably merits its own episode, and it's fine for leadership, you know, to talk about something.
• 23:40
  But the leadership itself is not, is not stable.
• 23:43
  There was a term that I'd heard back in the day which was when Janet Woodcock would say something, and then like, would it really hap- you know, would it really flow through?
• 23:50
  Because it takes a while, and they called that podium policy.
• 23:53
  So you could- Yeah … talk about it, but then like what's, you know, then somebody who's actually submitting something from a drug company is like, "I don't know.
• 23:59
  Janet Woodcock's not gonna review my submission, so I better do it the old-fashioned way." And then if- Correct … the person that you used to work with isn't even there anymore, uh, you know, that's a, that's a problem.
• 24:08
  But maybe a topic for a different podcast, as important as it is.
• 24:13
  John, if, if you don't mind, I'm gonna, uh, close us out, uh, here for the moment and just say that, you know, on the one hand you've got AI, it lets a single entrepreneur orchestrate a pipeline, but it doesn't, you know, software is not dealing with human biology just yet.
• 24:29
  The fate of biotech remains tied to the clinical trial speed, FDA flexibility, and so on, and it's, uh, tremendously exciting, um, and I think more to discuss without being able to draw all the conclusions right now.
• 24:43
  I think on that very ambiguous, uninspiring end- Ungh … it's a good place to stop this episode.
• 24:50
  So with that, we wanna wrap this episode of Care Talk.
• 24:54
  I'm John Driscoll, the chair of the UConn Health System.
• 24:58
  And I'm David Williams, president of Health Business Group.
• 25:01
  If you like what you heard, but especially if you didn't, please subscribe on your favorite podcast service.