Speaker 0:  anthropic built claude to answer questions. now it wants claude to help create actual medicine. the real test begins when ai generated hypothesis leaves the computer and enters anthropic's new wet lab. so anthropic, and claude is now becoming a drug maker. anthropic has just launched claude science, an ai research workbench that connects scientific databases, computational tools, specialist agents, and laboratory infrastructure in one environment. at the same time, the company has confirmed they will run its own drug discovery programs, reportedly concentrating on neglected and rare diseases. so anthropic is no longer merely selling scientists and ai assistant. it is becoming the first major frontier model company to test its technology against one of science's most expensive realities, producing a drug candidate that actually works. what do you think about this, john?

Speaker 1:  i'm not surprised because anthropic has already taken over for canva. we've already taken over whatever. a lot of different industries. pharmaceutical, to me, makes sense because if we're talking about developing a drug, like, we know it typically takes ten to fifteen years. it costs 1 to $3,000,000,000, and failure rates exceed ninety percent of this. right? and the biggest bottleneck in all of this isn't manufacturing. it's discovering which molecules might actually work. and so modern ai attempts, they try to compress years of laboratory expiration, into weeks by predicting protein folding, identifying drug targets, generating new molecular structures, simulating interactions, ranking the highest probability candidates. and instead of testing millions of molecules physically, now researchers can can narrow candidates computationally before even entering the lab, which reduces experimental cost dramatically. but the thing is, as of right now, though, ai can't replace laboratory validation or clinical trials. so every prediction, it's gotta still be experimentally verified, but we're cutting down the biggest bottleneck, which is the discovery phase of it. right? and, hopefully, the, the failure rates of it too. but, again, i'm not surprised. i'm just curious as to what industry are they gonna take on next. what are your thoughts on this?

Speaker 0:  well, i they've got 60 plus curated scientific skills and connectors and diet scientific databases. they can scale from one gpu to hundreds using laboratory infrastructure or external compute. they have up to 50 external ai, for science projects that will receive support. each selector project can receive $30,000 in quad credits and up to $2,000 in, model computing credits. so one u, us ucsf research group reported that completing certain analysis is roughly one tenth of the previous time as you indicated. a report case, reported case study, not a universal benchmark. so this is really cool, but i wanted to do a little, comparison to great. we've been talking about google's deepmind, doing some things very similar, and i'd like to, us to understand the differences. so google deepmind built a specialist ai that understands molecular structure. anthropic is building a general ai workforce that can coordinate much more of the scientific process. so to understand the anthropic, versus, google deepmind, anthropic is using a general purpose scientific reasoning where google deepmind is specialized biological and molecular models. the flagship technology is cloud science for anthropic where deepmind is using alphafold two and alphafold three and isomorphic drug design engines. anthropic's primary strength is our reading, research, analyzing data, writing code, and planning experiments, and coordinating tools where google's is predicting molecular structures and interacting and designing drug candidates. and there's a bunch more, differences here. but, basically, the biggest advantage anthrophic has is the breadth and agenda coordination. and for google, it's depth of biological modeling and accumulated drug development infrastructure. just you know, we talk about them, and i wanted to kinda separate the two a little bit.

Speaker 1:  go ahead. yeah. there's, like, there's other companies. like, openai has their gpt rosalind. nvidia has one. microsoft has one. amazon have one. so, like, it makes sense that anthropic would get into this. i know we covered alphafold before, which kinda like, i know google is probably ahead of the curve on most of this. but, again, just goes to show that it doesn't matter if we look at, the ai race nowadays. anyone could take the lead at any time.

Speaker 0:  yeah. any like, they can jump ahead. i think, deepmind and google's alphafold, we've been talking about a little bit more because i think they've been public about it. they're already on alphafold three. we're hearing talks about alphafold four, and this is claude's kind of first public foray into it. so, they're all kinda getting into it, and i think it only means good things for the future of medicine. but we're starting. i am skeptic. you know, that because biologic biology is not software. when i say skeptic, i mean, i think it's gonna help a lot, but, an ai can generate a plausible target molecular molecular experiment very quickly. but animal studies, toxicity problems, manufacturing constraints, and human clinical trials can still invalidate years of apparent promising work. so they haven't disclosed a specific drug targets, candidates, or preclinical results or a commercialization plan for anything, but that's because, i guess, we haven't come up with it. we just caught up with the thing. so i we we seem to be trying to shorten and shorten the long term studies that we used to do, which i guess is a good thing. but at the same time, then i think we're failing to understand the long term side effects of some of these things because we're rushing them in the market. now the long term benefit may well exceed any sort of negative, which is perfectly fine. i as an old school person, i'm thinking, where's the five year, ten year, study that shows side effects, that shows any, you know, weird cancer, weird whatever growth, or any of that kind of stuff? we're just not getting it anymore, at or maybe not. so those are my concerns, but i do like that we're being able to do these. my other concerns is how long does it take to work creating designer drugs, you know, not pharmaceuticals, recreational, whatever with these tools. and then how, good or dangerous are those?

Speaker 1:  yeah. well, what i like about what anthropic's doing is they're focusing on neglected diseases because they're illnesses that disproportionately affect poor populations where traditional pharmaceutical companies often see limited financial return, which i love. again, another check mark in anthropics, good guy box kind of thing. because, again, they've taken hard stances on certain things. like, they're coming across as the more ethical company, and now they're focusing on more neglected diseases. so, for those of you that don't neglected diseases, we're talking about tropical diseases, rare infectious diseases, and certain rare genetic disorders. so i guess they're they're sort of demonstrating a more humanitarian impact while avoiding direct competition with other, pharmaceutical markets initially, which smart i think is nice. what do you think about that approach?

Speaker 0:  i like them going after rare neglected diseases because why not? the second thing is what are they gonna do with it? are they actually gonna produce the drugs? because the pharmaceutical companies probably won't. because some of these things, one, it's not profitable enough to create something that's gonna cure things. one, they prefer a treatment because then you got a recurring payment. you know, say what you wanna say, but that's we like we like subscription plans in this world, so we want your recurring treatment in the drug world rather, not necessarily a cure. and you say underprivileged, or unrepresented, or whatever. i just think it's just not big enough for there to be a market, you know, because neglected and rare diseases are neglected and rare. like, they're not sweeping through the populations because in the end, if they were larger or if there was a larger population affected, somebody would pay for it. we see, you know, the bill gates foundation going in and creating drugs and giving them to africa and doing whatever. whether you're a conspiracy theorist and think that's evil because they're testing weird stuff or you're on the other side saying that's great philanthropic. they're helping everybody. the point is somebody's spending the money to create these drugs and getting them distributed, hopefully, for the betterment of mankind. it'd be great to have a faster track, cheaper, getting there, and, cheaper results so it's more distributed. so it's not controlled by the pharmaceutical companies so the whole world gets this information. so the question is, is this actually going in the good guy box? what's anthropic gonna do when they actually come to the cure? who gets the information? who gets to sell the stuff? the that is we're gonna have to wait and see.

Speaker 1:  yeah. well, here's the thing, though. right? and this would be kind of my takeaway is the first generation of ai, the company sold software. and this second generation, they're beginning to own the outcomes. so, like, tesla builds ai in cars. waymo builds ai in robotaxis. anthropic now wants ai in pharmaceuticals. so they're extending their tentacles into other things, which makes sense. they're they're diversifying and becoming kind of vertically integrated into everything. they're not waiting for people to build solutions. they're building the solutions themselves because now they have a working ai model. right? why not build upon that?

Speaker 0:  yeah. and tools for the labs and, you know, funding for the labs at least in the way of credits and whatnot. but the question is, what's the long game? i i'm happy with everyone doing it. it's only gonna lead, like, we're talking about, this could cut development from approximately twelve years to seven, eight years or double approval, success rates in that time as well. so there's a lot of good to be had here. the question is, what are they doing with it long term? and what's it gonna be, you know, in the wrong hands? what can it do? yeah. so anthropic has built an unusually complete bridge from ai reasoning to physical biology, but the breakthrough only becomes real when the bridge produces valid treatment. so should anthropic become a drug developer itself, or should it remain the ai infrastructure provided for science and pharmaceutical companies? let us know what your thoughts are in, below and where you think they fit in the overall, drug research and production pipeline.