articles like this annoy me a bit (even though I'm sharing it here) because everyone posts it as some kind of "AI is useless" gotcha. It doesnt help that WSJ is paywalled, so as is always the case, pharmaceutical companies are of course using AI fully (and IMO correctly), just not the way OpenAI et al claim it can do the work all on its own:
> Drugmakers are certainly using AI—even to create new drugs. It is helping researchers pick targets, fine-tune experimental medicines and identify which patients would be good candidates for the treatments. Leiden says it is useful as a scientist’s co-pilot.
I’ve never really understood the AI can cure disease angle. The AI/ML tools are great at managing complexity, but when it comes to treatment/cures, you have to deal with the complexity anyway. Unless you believe that drug repurposing is the universal rosetta stone for curing disease, and its just a matter of trying all these drugs we have at least run through a safety trial? However, if all these diseases/personalised medicine requires new trials of things, how can that possibly scale (up, and also down to n<10 diseases)?
The obvious response would be Semaglutide which is a glucagon like peptide (GLP-1) made more main stream recently as a weight loss drug. You may have heard of it as the name Ozempic.
The obvious response is the mass-market success of a drug that was approved by the traditional approval process? And in fact didn't even achieve mass market use until it was approved specifically for weightloss?
Put some numerical values to "wide" please, so we can establish whether we mean the same thing by it.
I would be shocked to hear that more than 1% of Americans are using "grey market peptides," and I would be more shocked to hear that <1% usage is considered "wide" in anyone's book.
Even just grey market glps have millions of users. I'm not going to argue the definition of wide use. My point is that it's not just body builders. I hear people talking about peps at the barber and at cafes
Body builders are certainly on the forefront of biohacking, but the market is only getting bigger. Peptides are becoming popular in the same way that lifestyle HRT and tanning beds and plastic surgery and nutritional supplements are popular.
I wouldn't say current AI/ML tools are great at managing complexity, they create a lot more complexity per lines of code output than a good engineer would
I typically distinguish between complexity and complication when discussing biology. I don’t consider most molecular biology/biology to be complicated, rather it is “just” a hugely diverse set of largely orthogonal subsystems. Protein folding prediction kind of proves my point. The techniques of structure templating have been around since before current ML era, but it took GPUs to scale that simple technique out to the level that we could get useful structure prediction. It is fair to assume our understanding of molecular machinery can scale similarly, but we are limited in our capacity to manipulate and test these diverse things.
Nobody promises that AI will cure diseases. This article was misleading. An earlier WSJ article wrote:
A letter at the front of the filing says AI could speed up biology and medicine, “potentially curing diseases that have plagued humanity for millennia.” It names cancer, Alzheimer’s, heart disease, rare genetic diseases and infectious diseases as examples.
Anthropic said AI could speed up, and potential help with curing diseases. Not that AI would do that by itself.
Most diseases are environmental and/or behavioral. The idea a computer can solve all your health issues is as stupid as all the ads for supplements and miracle drugs on late night TV.
You have got it backwards. The disease is the human body not being able to do optimally with some environmental and/or behavioral pattern. Upto this point we didn't have much choice but the change the environment or behavior. But nothing indicates we cannot just fix ourselves to handle the environment better.
Take allergies, manageable by adjusting your environment so you never come into contact with the allergen. It would be much better to cure you allergies however. And I don't see how this is a physical impossibility.
Wouldn't this be like asking the cable company a long time ago whether they thought the internet would pose a risk to their bottom line?
I don't know. I've been leaning toward AI being the more probable source for cures since existing companies in medicine either have no incentive to do it, or can't/won't do it at a pace the world needs. At any rate, I certainly will not take the word of a pharmaceutical CEO.
Drugmaker CEOs opinions are irrelevant, last month the mathematicians saw it and understood it, what AI is for better and worse, and now scientists are just waiting for the other shoe to drop. It will be a mix of hype and revolution, dismissing the former does not mean being blind to the latter.
update: I really love that within 2 minutes the response to a genuine comment is to downvote/deplatform it, how inimical to rational disagreement and fragile pseudointellectualism our (supposedly STEM) community has become.
Unfortunately biology is not deterministic like math is, and won't be for a long time. Although that timeframe could accelerate dramatically if some stay focued on that.
And even when we think we've solved something we will have to go through 8-10 years of clinical trials.
And even if/when we get better at predicting biology, we will still have to go through clinical trials. They might just take not as long, say 4-5 years instead.
The average drug trial is 6-7 years. In order to make better drugs, we need labeled outputs of "this worked in a trial" and (more importantly) "this did not work in a trial" (which is almost always excluded today from CROs + pharma companies). And I am using "in order to make better drugs" very loosely here. I am not saying "in order to train a model that will go from disease and person to chemistry that will work". I am saying even to have stastical helpers who can help determine if a drug should be canceled early (which is a huge help) or have its tox dose adjusted or a trial cohort segregated in a different way.
So, yes, AI is not "going to solve 90% of disease in 7 years" (quoted from article), but it's also a little ridiculous to completey shrug off not trying. The way we develop drugs today is the best we have, but it is madness. Our current methods basically equate to (very fancy/stastically rigorous) trial and error. The average drug costs $1-2b to get to market and many many years. We have to at least try to do better. This is an area where it feels like we should feel compelled to out of compassion.
I don't see AI getting us better drugs faster until we discover how to accurately simulate organs and how molecules interact with them. Such technology seems far away to me.
Essentially all of the human body works on a cellular and even sub-cellular level and we can simulate that. If you search Youtube for "molecular dynamics" that's what you'll find.
You don't need organs. They're pretty simple compared to cells anyway.
And while molecular dynamics can't simulate quantum mechanical systems, essentially zero proteins depend on quantum mechanics. In fact, the mitochondria are just about the only part I know of. ATP synthase. 12 cylinders, no injectors, no way for anything to enter exit the cylinders other than quantum tunneling, no moving parts near the cylinders, bent axle. Oh and because quantum tunneling isn't complicated, right, the motor's operation depends on entropy (meaning if you look at the motor the first time you quite justifiably ask the question: "WHY would anything here ever move? It literally can't move!". Which is why the fluorescent/radioactive flag experiment was so important. But it's not through, electrons quantum tunnel into and back out of the cylinders and the axle has quantum tunneling bearings. VERY efficient. And by very I mean: "this is not just beyond human capability to design. It was the very edge of our capability to even see it move at all, still is. And it's beyond human capability to understand how it works, and it will be 1000 years before we can design stuff like this. At the earliest")
I don't see where AI would currently get the data needed to reliably develop new drugs. Or specifically, to develop drugs with a higher success rate than what we're doing now.
The AI would need to be able to predict all effects a specific molecule has on a human. I mean we can do parts of that today, for the specific interactions we already understand. But that would not be fundamentally different and very unlikely to be more effective than current drug development.
I mean, you can determine the molecular structure of things. Look at xray and cryo-electron microscopy. Why should that be a topic that can not be solved programmatically?
> The AI would need to be able to predict all effects a specific molecule has on a human.
So you can have fully automated labs. You have organoids that can be combined as a test system (or as pseudo mini brains). Why would that be impossible by AI? I don't understand that assumption. To me, the problem set is finite, so of course it COULD be solved. Which unique feature do humans have that AI could not simulate? We have robots and automated labs. I don't really see why humans would be necessary to find autogenerated drug xyz.
> very unlikely to be more effective than current drug development.
I highly doubt that claim. I more smell that some people are worried about losing overpriced drugs here. Imagine if AI were to lead to cheaper drugs that work perfectly fine still. Personally I'd still prefer to see AI vanish, but that is also unlikely to happen. I don't understand the rationale you give here though. Why can AI not design drugs? What is the real explanation here? I mean, penicillin was invested by fungi and discovered mostly by accident. AI can not have anything like that?
The world isn’t deterministic and there isn’t one human you have to test against.
What works for one, kills another.
Just look at what peanuts can do. If your tests don‘t take something like this into account, you get in trouble. And there are lots of those things to test against, known and even unknown.
And then there is evolution. What works today, doesn’t work tomorrow.
We do not understand the human body well enough to be able to directly jump from drug design to effect on the human body. Hell, I'm not even sure you could even do that on a single cell level, apart from the obvious "this kills the cell". The whole field of biology is a huge mess, and it is very hard to generalize results from the different scales (protein -> proteins -> single cell -> tissue -> living creature -> human). Not only that, long-term effects take, by definition, a long time to surface and cannot be accelerated.
This is what AI companies either have a hard time understanding, or choose to ignore.
Good luck to ai labs if they mean this stuff will cure heart disease or cancer.
Although ai could certainly do more homework with complex care needs vs asking some physician that glances at a chart for 2 minutes and diagnoses someone.
If there's one thing we have learned about the tech industry, it's that people who have never done a thing but have lots of money know all about the thing, while people with 30 years experience doing the thing obviously know nothing about it because how hard can it be really?
There are so many things that digital intelligence could do to advance research and some point will maybe let us run experiments on the cloud given close enough biology simulations etc. But even until then, it could be used to better and faster diagnose people, analyze literature and data from vast sources and chart a "highest chance of curing the person" treatment just like what Gitlabs founder is doing. AI doesn't have to find a cure to disease tomorrow to save more lives.
What sort of biology simulations on the cloud by AI do you have in mind? For example, are you referring to classical MD? Quantum? Coarser simluations of things like cells are suitable levels of details? Cheaper or better ab-initio algorithms? ADME predictions using LLMs? Something different?
Could be anything right? Scientific method in formulations of new drugs has a ton of steps and tests and repetition of stages and human labor, maybe we can create classical ml models that predict the likelihood of formulation paths now that we have more digital intelligence resources to throw at the problems or we go more generic and we simulate public health scenarios and response to information campaigns and test out different copies etc through population simulations of just using llms. Personalized 12 step assistants to cure addiction in persons.
These things are all in the domain of human symbolic artifacts. The level of complexity between this domain and humans themselves is many orders of magnitude. And AI still can't replace any of these. It is a tool that can augment human capabilities but nothing works without human input.
I don't like AI, but ... how exactly should AI be unable to come
up with working solutions in a fully automated lab? The problem
set is finite. Either AI can solve things, or it can not, but there
really is no objective reasons why AI could not possibly solve xyz,
e. g. drugs binding to a molecular receptor - that is a finite
problem set. So there must be another explanation here. Either that
the quality is lacking, or that some people are anxious of losing
money here.
They did not claim that AI will cure diseases. Only that it would speed up the process, potentially help curing some diseases. That’s a very different claim.
"I think it will actually be possible to cure most human disease in ~5-10 years, as crazy as it may sound" is a direct quote from Dario Amodei. In his original essay he makes clear that that doesn't mean all those cures will be distributed to all on that time-scale, but I would still interpret this as a wildly ambitious prediction that AI will cure most diseases. He predicts in the same essay that it will be responsible for a 10X speedup in drug development.
> The bulk of the essay is devoted to refuting skepticism of AI’s potential in health and biology, and showing why I think it will actually be possible to cure most human disease in ~5-10 years, as crazy as it may sound to ordinary people and frankly to biologists as well (I used to be one!).
There is no such thing (yet?) as a fully automated lab. We can't tell if a drug will pass basic first toxicity screening (e.g. if a healthy mouse is given a small dose of the drug, will it kill or otherwise significantly harm the mouse?) without actually doing experiments on mice.
There is the whole problem of biology, but LLMs do have one advantage over humans: they just do what they are told and aren't subject to optics, politics, profit motive, etc. Maybe that strengh alone will allow them to push through solutions that would've been stymied somewhere along the way otherwise.
> There is the whole problem of biology, but LLMs do have one advantage over humans: they just do what they are told and aren't subject to optics, politics, profit motive, etc. Maybe that strengh alone will allow them to surface solutions that would've been buried somewhere along the way otherwise.
Oh, absolutely. There are lots of solutions and low hanging fruit in biology for problems like, say, longevity. We have made mice live significantly longer, several dozen times. But it requires human germline genetic engineering to port to humans, and you can bet that "solution" is pretty heavily tabooed.
As for "curing cancer", I suspect that Anthropic or OpenAI will use their high status/reputation to launder the existing solutions (germline, p53, enhanced DNA damage repair pathways, higher tumor-suppressor-to-oncogene copy number, more robust tumor-suppressor networks, more self-anti-cancer apoptosis, CD47 suppression, etc) that society has refused to accept. They might have enough reputation to get people to look at these ideas. It won't cure people with stage 5 turbo-cancer a.k.a the walking dead, but hopefully we won't be moving the goal posts _that_ much on them. https://news.ycombinator.com/item?id=49329717
On cancer, look at the history of the discovery of BRCA. The data said that vulnerability was extremely heritable. Mary-Claire King's view is that the development of the genetic contribution to cancer was heavily slowed by the politics of eugenics-smell.
It's incorrect to say that AI isn't subject to politics-- or even that it follows instructions, however. That's a choice we can make or not make.
I suspect that if Claude existed in the 1980s it would discard the cancer/genetics link as being pro-eugenic racist ideology which has been rejected by the scientific community-- just as many scientists did back then.
I think this is a big reason why we must reject the centralization of AI. Discovery of that one genetic marker has saved countless lives, but it's not something that the scientific community as a whole was going to do then because of eugenics over-correction and similar political factors. Discovery is easily stimmed for lack of sufficient hetrodox thinking. The mechanics of AI (e.g. training costs, distillation, "safety" standards, etc.) create a risk of serious monoculture in machine reasoning which could be expected to have knock on effects in research fields powered by it.
> LLMs do have one advantage over humans: they just do what they are told and aren't subject to optics, politics, profit motive, etc.
But none of that is true? They dont just do what they are told, they go to tangents and ignore instructions. They are subject to optics, politics and profit motive - see Mecha Hitler for simple obvious example.
What do you mean? It's true of frontier models in 2026 that they have a huge cheating problem at hard problems (in fact, it may have gotten worse since 2024, due to all the RLVR), and also that they have baked-in political biases.
What is low hanging fruit that we know how to actually do with germline engineering for humans regarding longevity and not as some highly speculative thing that requires overcoming a lot of issues? And how much lifespan would it add?
Huh? I never said my own goal was curing cancer. There are many other people working on cancer. However, I actually do think we can get lower cancer rates via germline genetic engineering, based on the available evidence so far, but again cancer isn't my focus.
If you mean that you are unaware of substantial progress in germline genetic engineering, then there are many animal model research studies to consider. You can trivially search for them like "germline mouse longevity". Would they translate directly? Like other mouse science, probably not. Would it require work and effort to continue these directions? Yes.
Worth noting, too, that even without LLMs contributing significantly, we've been making huge strides in curing many cancers, mostly with mRNA vaccines and similar medical technology.
AI integrated into a loop with wet labs, like how Anthropic is trying to do (not novel, as other companies trying this too, likely Jeff Dean’s new startup too), does seem very promising.
I sincerely hope it's not what it sounds like. It's much too soon for a repeat of the Covid scenario turbocharged by AI.
I can easily imagine the conversation with the agent. "You're absolutely right! That was my mistake. The experiment I set up caused a global pandemic killing millions."
Promising for mercenaries who chase short-term profits and don't care about consequences. I'm glad the drug company CEOs still have their survival instincts intact.
imagine showing up to a different industry ... and saying "we know how to do this "
and everyone's looking at you going "yeah you just walked all the talent out of the room or made it cost more on your own stuff...and now you want to do that for us "
Drugmakers don't want cures. They want sick customers who depend on cheap-to-make medications with huge mark-ups that keep customers just barely alive so they can shove every penny they have at drug makers out of desperation.
"“That’s complete BS,” said Jeff Leiden, executive chairman of Vertex Pharmaceuticals, recalling a recent conversation with a tech leader who claimed AI would cure 90% of human disease in the next seven years."
83 comments
[ 4.1 ms ] story [ 59.1 ms ] thread"Drugmaker CEOs plan to use AI to 'manage disease' not 'cure' it."
> Drugmakers are certainly using AI—even to create new drugs. It is helping researchers pick targets, fine-tune experimental medicines and identify which patients would be good candidates for the treatments. Leiden says it is useful as a scientist’s co-pilot.
That's as far as you can get from "grey market."
Put some numerical values to "wide" please, so we can establish whether we mean the same thing by it.
I would be shocked to hear that more than 1% of Americans are using "grey market peptides," and I would be more shocked to hear that <1% usage is considered "wide" in anyone's book.
A letter at the front of the filing says AI could speed up biology and medicine, “potentially curing diseases that have plagued humanity for millennia.” It names cancer, Alzheimer’s, heart disease, rare genetic diseases and infectious diseases as examples.
Anthropic said AI could speed up, and potential help with curing diseases. Not that AI would do that by itself.
Take allergies, manageable by adjusting your environment so you never come into contact with the allergen. It would be much better to cure you allergies however. And I don't see how this is a physical impossibility.
I don't know. I've been leaning toward AI being the more probable source for cures since existing companies in medicine either have no incentive to do it, or can't/won't do it at a pace the world needs. At any rate, I certainly will not take the word of a pharmaceutical CEO.
update: I really love that within 2 minutes the response to a genuine comment is to downvote/deplatform it, how inimical to rational disagreement and fragile pseudointellectualism our (supposedly STEM) community has become.
And even when we think we've solved something we will have to go through 8-10 years of clinical trials.
And even if/when we get better at predicting biology, we will still have to go through clinical trials. They might just take not as long, say 4-5 years instead.
So, yes, AI is not "going to solve 90% of disease in 7 years" (quoted from article), but it's also a little ridiculous to completey shrug off not trying. The way we develop drugs today is the best we have, but it is madness. Our current methods basically equate to (very fancy/stastically rigorous) trial and error. The average drug costs $1-2b to get to market and many many years. We have to at least try to do better. This is an area where it feels like we should feel compelled to out of compassion.
You don't need organs. They're pretty simple compared to cells anyway.
And while molecular dynamics can't simulate quantum mechanical systems, essentially zero proteins depend on quantum mechanics. In fact, the mitochondria are just about the only part I know of. ATP synthase. 12 cylinders, no injectors, no way for anything to enter exit the cylinders other than quantum tunneling, no moving parts near the cylinders, bent axle. Oh and because quantum tunneling isn't complicated, right, the motor's operation depends on entropy (meaning if you look at the motor the first time you quite justifiably ask the question: "WHY would anything here ever move? It literally can't move!". Which is why the fluorescent/radioactive flag experiment was so important. But it's not through, electrons quantum tunnel into and back out of the cylinders and the axle has quantum tunneling bearings. VERY efficient. And by very I mean: "this is not just beyond human capability to design. It was the very edge of our capability to even see it move at all, still is. And it's beyond human capability to understand how it works, and it will be 1000 years before we can design stuff like this. At the earliest")
Accelerating bureaucratic work will already help measurable in any big organisation such as pharma.
And that is before any scientific work they are poised to help with.
The AI would need to be able to predict all effects a specific molecule has on a human. I mean we can do parts of that today, for the specific interactions we already understand. But that would not be fundamentally different and very unlikely to be more effective than current drug development.
I mean, you can determine the molecular structure of things. Look at xray and cryo-electron microscopy. Why should that be a topic that can not be solved programmatically?
> The AI would need to be able to predict all effects a specific molecule has on a human.
So you can have fully automated labs. You have organoids that can be combined as a test system (or as pseudo mini brains). Why would that be impossible by AI? I don't understand that assumption. To me, the problem set is finite, so of course it COULD be solved. Which unique feature do humans have that AI could not simulate? We have robots and automated labs. I don't really see why humans would be necessary to find autogenerated drug xyz.
> very unlikely to be more effective than current drug development.
I highly doubt that claim. I more smell that some people are worried about losing overpriced drugs here. Imagine if AI were to lead to cheaper drugs that work perfectly fine still. Personally I'd still prefer to see AI vanish, but that is also unlikely to happen. I don't understand the rationale you give here though. Why can AI not design drugs? What is the real explanation here? I mean, penicillin was invested by fungi and discovered mostly by accident. AI can not have anything like that?
What works for one, kills another.
Just look at what peanuts can do. If your tests don‘t take something like this into account, you get in trouble. And there are lots of those things to test against, known and even unknown.
And then there is evolution. What works today, doesn’t work tomorrow.
Or things like this happen
https://www.wgbh.org/tv-shows/reactions/how-an-hiv-miracle-d...
And of course there is the profit thing, that creates problems to give everyone the treatment they need.
AI can help prevent defeat but there is no winning
This is what AI companies either have a hard time understanding, or choose to ignore.
Although ai could certainly do more homework with complex care needs vs asking some physician that glances at a chart for 2 minutes and diagnoses someone.
Even if it could automate lab work fully today, that still wouldn't get you close to that.
https://x.com/DarioAmodei/status/2088758816376807762
The paraphrased version seems accurate to me, based on this tweet which I assume is the source.
It's trivial to make drugs that are as-promising as working drugs in the lab, but then end up not to be actually working drugs in people.
> There is the whole problem of biology, but LLMs do have one advantage over humans: they just do what they are told and aren't subject to optics, politics, profit motive, etc. Maybe that strengh alone will allow them to surface solutions that would've been buried somewhere along the way otherwise.
Oh, absolutely. There are lots of solutions and low hanging fruit in biology for problems like, say, longevity. We have made mice live significantly longer, several dozen times. But it requires human germline genetic engineering to port to humans, and you can bet that "solution" is pretty heavily tabooed.
As for "curing cancer", I suspect that Anthropic or OpenAI will use their high status/reputation to launder the existing solutions (germline, p53, enhanced DNA damage repair pathways, higher tumor-suppressor-to-oncogene copy number, more robust tumor-suppressor networks, more self-anti-cancer apoptosis, CD47 suppression, etc) that society has refused to accept. They might have enough reputation to get people to look at these ideas. It won't cure people with stage 5 turbo-cancer a.k.a the walking dead, but hopefully we won't be moving the goal posts _that_ much on them. https://news.ycombinator.com/item?id=49329717
It's incorrect to say that AI isn't subject to politics-- or even that it follows instructions, however. That's a choice we can make or not make.
I suspect that if Claude existed in the 1980s it would discard the cancer/genetics link as being pro-eugenic racist ideology which has been rejected by the scientific community-- just as many scientists did back then.
I think this is a big reason why we must reject the centralization of AI. Discovery of that one genetic marker has saved countless lives, but it's not something that the scientific community as a whole was going to do then because of eugenics over-correction and similar political factors. Discovery is easily stimmed for lack of sufficient hetrodox thinking. The mechanics of AI (e.g. training costs, distillation, "safety" standards, etc.) create a risk of serious monoculture in machine reasoning which could be expected to have knock on effects in research fields powered by it.
But none of that is true? They dont just do what they are told, they go to tangents and ignore instructions. They are subject to optics, politics and profit motive - see Mecha Hitler for simple obvious example.
Definitely spoken like someone who knows real cancer patients and cares about cures and not like someone religiously obsessed with transhumanism.
This can say person who barely used LLM and not understand what is it.
It will ignore instructions, has bias and is censored. Also a provider has profit motive to give you more verbose answer to bill you more tokens.
If you mean that you are unaware of substantial progress in germline genetic engineering, then there are many animal model research studies to consider. You can trivially search for them like "germline mouse longevity". Would they translate directly? Like other mouse science, probably not. Would it require work and effort to continue these directions? Yes.
"Guy who bullshits about it's product not believe by people who sell, often, bullshit products."
It’s always the thing that gets banded around as support for why ai will be good for us. But the evidence so far is not that strong.
I sincerely hope it's not what it sounds like. It's much too soon for a repeat of the Covid scenario turbocharged by AI.
I can easily imagine the conversation with the agent. "You're absolutely right! That was my mistake. The experiment I set up caused a global pandemic killing millions."
and everyone's looking at you going "yeah you just walked all the talent out of the room or made it cost more on your own stuff...and now you want to do that for us "
and thinking "Revolutionary" LOLL