https://podcasts.apple.com/gb/podcast/black-box-the-chatbots...
I think LLMs are an astonishingly powerful technology, but I think their tendency to induce the ELIZA effect makes it socially irresponsible to go about as though you've created a literal Deus ex machina personally.
I don't know if he is in that video or not, or ever, personally, but just in general, it's the Ockham's razor approach any time you see something like this. It's a much more direct and clear explanation than the various more convoluted theories people keep throwing around as to why everyone sounds like a methed-out megalomaniac all the time.
I never feel more like the one sober guy at the party than when I go to SF. Even when there's no reason to believe anyone is literally on drugs, it has that vibe every time. It's really unsettling. Uncanny. It starts to get under your skin, too. You start to believe, just a little bit, that everything is possible, that there are no limits, that you can and should remake the world. Just a little bit. I can never take it for more than a week at a time, no matter how beautiful the weather is. It's a relief to get back to the "real" world.
> Look. This is a new golden age of cocaine, because of the big cuts in everything in the US took away the cocaine busting operations in Colombia and Bolivia, because the revolutionaries that previously kept control of it stood down and became a political party and new, profit-driven cartels came up, and because battery technology means that narcosubs and similar drone things work much better than before (and they can be made remote-operated, so the cops can't do shit even when they find one). Wholesale prices have halved or gone down 60% to like, 12k usd / kilo, street prices commensurate. The huge new opportunities created by increased market size have increased competition and made the cartel nature of the cartels decrease. So always think of that when you look at people acting like this and having such megalomaniac goals.
That aside, I think there's something that happens to the human mind (and most people are susceptible) when exposed to certain stratospheric levels of wealth and power (which are basically the same thing). Their thinking becomes ungrounded, uncoupled from reality. The tendency is to shrug off the limitations of logic and rationality and to allow their mind to dabble in ideas of limitless possibility and power because doing so feels good to do. It's a drug and a very seductive thought that one's will could be manifested in the world without compromise - that what's inside could be made real on the outside. That is, after all, what power is. Dario's imagined reality is that he is the savior of the human race, and that's the reality he's currently living in.
"it is a literal and useful description of anthropic that it is an organization that loves and worships claude, is run in significant part by claude, and studies and builds claude..."
Unhinged, maybe not.
But echoes of Liz Holmes and SBF, absolutely.
IMO, both, to a large extent, actually believed their own BS and Dario strikes me as very similar in that regard.
With Holmes, some of it was intentionally fake and manipulative (especially the voice), half of it was believing her own BS. With SBF it was mostly the latter, I think.
Amodei is a smart guy, but there are aspects of his behavior that almost seem like he's doing an over-the-top exaggeration of non-neurotypical mannerisms as a kind of act.
I don't see how one can seriously argue that Amodei was similar for holding an optimistic view of AI's progress in the next 5-10 years.
Honestly, the way he lists them and even pauses before vaguely saying "mental illness" gives a strong impression that he's choosing whatever illnesses he can think of as he goes.
Amodei might be an even bigger optimist than I am, but he's not lacking credentials on neither the biology nor AI side.
Accusing him of fraud or mania like some here do seems unwarranted.
It smells to me like "might be an even bigger optimist than I am" means you're also part of the rationalism cult and that it's literally impossible for you to say something negative about Amodei.
Being an optimist does not make me part of any cult and there is no need for you to argue that I could not say something negative about Amodei.
If you want to argue that Amodei is dishonest and only says these things to 'bump the stock', you can make that argument, but it's clearly speculation.
When someone promises something that's so far out there as to be nearly impossible, and *without an actionable plan to get from here to there*, then they're either dishonest or delusional.
He hasn't promised anything, in fact he acknowledges that his projections might be wrong.
His stated plan is to scale AI and compress 100 years of biomedical progress into 10 years.
Like, if your optimism is great enough, big enough, then you lose touch with what’s actually feasible. That’s where delusion starts.
It might be possible to compress 100 years of research into 10 years. Maybe. But it won’t be these 10 years.
Check the math. Say we get huge advances in AI tomorrow that double the speed of biology research. And imagine we then have 3 years where research is now 2x faster (which, again, would be absolutely enormous).
How much faster do the remaining 7 years have to be, in order for the entire 10 years to capture 100 years of research? Roughly 13x.
Basically, every year you’re not already at >10x is a year you’re falling behind. And AI is nowhere even remotely close to doubling the speed of research, much less 10x-ing it.
He is predicting the arrival of powerful AI in the next few years.
His plan is to scale that model up to millions of instances working in parallel, directing experiments and research autonomously.
In this way he hopes to accelerate the rate of important discoveries by at least factor 10 to achieve 100 years worth of progress.
More importantly, he is lacking credentials in medicine.
This is just one problem I have with his claim, there are many others.
Now, whether that's true is obviously it's own question. But I do think if AI can cure cancer and Alzheimer's rapidly, that improving the regulatory and approval processes at agencies is a much lower bar.
I don't think it's the case, and that Amodei legitimately believes in what he has been saying for years.
Agree with this if it was something a random person said over coffee or lunch.
But when its the CEO of a company that is about to IPO saying it, its very very different and more concerning.
This is up there with the claims Chan-Zuckerberg Foundation made during their start-up that the foundation would "end all disease within our lifetime." So far they haven't ended one disease, and meanwhile they quietly watered down their claim to soften the language to update to "cure, prevent, or manage all diseases by the end of the 21st century."
If this happens to be the case - I personally don't see why those statements are not realistic.
All diseases are engineering problems. The bottleneck there is testing etc. Most likely many of them can be cured using existing approach or meds.
Companies grow organs now that can be used to speed up iteration.
Even current AI can solve a LOT of tasks if it has quick feedback loop and can verify results. At the end of the day it can brute-force its way to solution.
AIs intelligence is say average human. Even less, whatever. say it will be as a smart human in a few years.
But it is infinitely scalable.
So is it that unrealistic to assume AI can find a correlation on massive amounts of data and figure our what leads to Alzheimers? (where the solution could be not cure but prevention). I don't think so.
A wildly absurd claim, especially for some of the most studied diseases in human history. This has only ever been true for a tiny handful of very specific diseases, such as scurvy, in all of human history. Expecting that cancer can be cured or prevented using already existing drugs that we've just not been noticing is complete quackery.
> Even current AI can solve a LOT of tasks if it has quick feedback loop and can verify results.
IF it has a quick feedback loop. You know what definitely DOESN'T have a quick feedback loop? Research into medicine, and biology in general. It takes raw time to find out what the effects of a substance on an organism are, often weeks or months.
So you are saying that it is 100% impossible in the next 10 years to come up with a way to make it much faster? It takes months - sure, can we run 10k versions of this at the same time? this way in a few months we will have massive amount of results never seen before?
"We can't because of this and that" - and this and that is 100% can't be fixed?
Don't forget - a massive amount of diseases have not been cured just because there is no monetary incentive to spend time on them for big pharma.
If AI brings cost down - does not matter how - could be just because it makes custom software or automates some workflows etc - this alone can open a bunch of opportunities.
Whatever is not solved yes - often is not solved just because enough resources has been thrown at it.
And diseases is just like everything else. They will be cured only if there is financial incentive (or the ratio of cost to revenue changes).
The responses are to the CEO of Anthropic claiming that "Cancer, heart disease, Alzheimer's, and mental illness will be relics of the past in less than a decade". None of these is some obscure disease that is not getting the funding it needs.
> It takes months - sure, can we run 10k versions of this at the same time? this way in a few months we will have massive amount of results never seen before?
No, you can't - because you're not going to find 100k patients with a certain strain of cancer that you can run 10k parallel trials of different drugs on. To the extent that research can scale, it is already being handled at the higher possible scales, for most of the major diseases.
> Don't forget - a massive amount of diseases have not been cured just because there is no monetary incentive to spend time on them for big pharma.
Yes - but almost exclusively for diseases that affect a very small amount of people each. Making it even more impossible to scale out testing, for the simple fact that only a small amount of victims of the disease even exist.
You can't right now.
10 years ago organoids were not a thing. They are now. Can it be another similar development happens in next few years that will allow to create more complex artifical biological systems? Don't see why not.
If that happens can we speed up trial time by orders of magnitude so human trials start only when there is proven much higher level of confidence?
I bet we can
Compute with the power to solve literally any medical issue is also compute that can create bioweapons that instantly kill any human.
Might as well fart into the microphone, in my humblest opinion.
AI is clearly a very capable new tech. Obviously there will be a lot of consequences downstream in 10 years.
The architecture of ML as it currently stands does assist us in modeling things like protein folding. But that's what they are; models. And before some person chimes in by saying "look how the AI pessimists move the goalposts", see my first point. Biological systems have high entropy and discoveries in biology often do not follow from preexisting discoveries whatsoever and are completely novel. Barring some quantum leap in ML design and architecture you cannot predict stochastically or deterministically how exactly a new enzyme will function in vivo. There is absolutely no evidence that Anthropic has solved this, even if they gobbled up every ML tool used in drug design and trained their models on every biomolecule known to man, it would not work because of fundamental gaps in what can be observed. They are a for-profit company trying to juice their IPO.
Now could LLMs and ML potentially speed up the pace of this research? Absolutely. But it's like saying a Formula 1 car is faster than a NASCAR. Yes, one is very clearly faster, but it's not as if the NASCARs are slow.
You can make the argument that Amodei is dishonest or that he suffers from psychosis.
However, he has made no claims that Anthropic has solved this, and expressing a hope to solve this in the next 5-10 years is hardly fraudulent.
That's why I think that something is very weird with his statements. Like, how would this even work?
The feedback loop is much, much, much too slow right now for AI to make things better.
Like, step 1 is build a digital model of the human organism. If you can do this, then yeah maybe the rest is possible. But step 1 is incredibly hard, so it feels unlikely.
That being said, I think the stimulants hypothesis upthread may be correct, as I've heard a lot of crazy stuff from friends on them.
If we linearly interpret the last year of progress from GPT-5 to Opus 5.5, times 5x, assuming no plateau, could such an intelligence realistically achieve a task like that?
Maybe also not, perhaps some form of RSI would need to deliver accelerating progress.
There are a lot of assumptions and speculation here. Still, at the end of the day, it does not seem that unlikely that they succeed in creating AI that can reason over so many parameters as to remove all intellectual bottlenecks from the process, leaving mainly the physical limitations.
And what is the limit there to building a good model of the human organism? I won't pretend that I know, but simply pointing to clinical trials, as some here do, instead of making an argument based on how such a digital model cannot be created seems unserious.
I don't think they are holding anything back in revolutionary results or internal models, a week before the IPO.
People are routinely claiming that as someone in tech, Amodei lacks credibility on biology and has no idea about science.
I'm yet to see someone claim his academic credentials validate his predictions.
I wasn't talking about Amodei specifically, but in general: my take is that overweighting credentials is probably the number one mistake that is made in evaluating whether a complete stranger is qualified to do something or not, especially in positions of power.
If you gather 500 PhDs and put them in one room, and 500 individuals without post-secondary education in another, give them all a quiz, I'm sure that the former room will outperform the latter room on average. Pick one of those 500 PhDs at random and I can probably find a topic where they're completely out of their depth, and perhaps even arrogantly so: https://en.wikipedia.org/wiki/Nobel_disease
Yes. The mistake you're making here is that you're not understanding where the actual bottlenecks are in drug development, and why some targets like Alzheimer's have been so elusive.
The problem with Alzheimer's in particular is that we don't have a working model of how Alzheimer's works. The dominant model--the amyloid hypothesis--can pretty conclusively be chucked in the bin at this point because of the number of amyloid clearing drugs that just simply have no effect on the progression of Alzheimer's, but there's no model that actually looks any better. Without a proven model, the only to verify effectiveness of an Alzheimer's drug is try it out in humans [1] and wait several years to get a success-or-no-success readout. That's the rate-limiting step for Alzheimer's drugs.
Okay, so maybe you're going to suggest that we can use AI to come up with new Alzheimer's models. Okay, but the rate-limiting step is still proving those models, which because of the human-only nature of Alzheimer's, requires developing the drugs and testing them and you're going to wait a decade to get back your answer as to whether or not the theory was correct. And what data are you going to use? If you're trawling scientific literature, you're going to come up with only the same ideas people have already had for alternative Alzheimer's models (there are a lot of those). If you're trawling medical data, well, medical data just doesn't exist in the level of detail you need for that enterprise, and handwaving away the legal issues with collecting it, it would still take years just to collect that data (and probably still not be good quality!).
[1] There's another complication of Alzheimer's, which is that only humans get Alzheimer's, so we can't even do anything like testing in animals with shorter lifespans to speed this part up.
But it requires tools that we are nowhere even close to having. One way would be to build a ground-up, molecules-to-man model of how the body works. And then from there, sure, we can easily model how Alzheimer's might come about and how it might be cured. Like, with tremendous advancements we might be able to finish such a project this century. In the next decade? That's like saying we'll launch an expedition to Alpha Centauri in a decade.
Given that he was making these projections already long before today, it would be fair to say that it's speculative rather than based on current research progress.
Because I remember that you shared similar opinions about software where you cited liability as a limitation that for you precludes more autonomous use of AI.
I think it's safe to say that you have an entirely different expectation for how AI will progress and how it will be used, or at least that was my impression.
With the release of Fable 5.2, I have to say that you might be right. Maybe progress is already slowing down, it certainly isn't as big a step as I was hoping to see after Opus 5.5.
Shall we wait until that trajectory has run its course before we start claiming victory?
As for the rest of this discussion:
Medicine != computer science, math or physics. Biology isn't either. The degree of complexity is many orders of magnitude higher.
Curing cancer : very unlikely, cancer is a systemic issue that requires many mechanisms in the body to do its thing, most of which we also require to stay alive. Short of a massive breakthrough in nano technology curing cancer is stupendously hard and it would be proof of reaching an entirely different plateau of capabilities for Humanity as a whole. Besides that, it isn't cancer that kills the most people on the planet, it is hunger, and then climate change.
The second one: Curing mental illness. I think it is equally delusional to be able to claim that you can cure mental illness because mental illness is a very broad description, so broad that it covers a wide variety of issues with an even wider variety of root causes the bulk of which are not well understood. Our brains are hard to appreciate in terms of capabilities and it shouldn't be surprising that something so compact and so powerful can break. Whether you can repair it after it breaks is not something you can say with certainty in the present unless you can demonstrate structural progress in the field compared to where it was say five years ago. To the best of my knowledge this is not the case.
Finally, Alzheimers. This is another one of those 'too good to be true' things. Alzheimers is next to cancer one of the diseases that has been studied extensively without much progress as to both root causes and cures. That doesn't mean that we won't find one, but here too: you would have to show at least incremental progress and to show a number of very promising leads that we didn't have so far to be able to make this claim.
In short: this is a marketing ploy, these three were picked because they hit everybody, either directly or indirectly. What they are saying is 'imagine, if we had this AI that we are working on today the problems that affect you or your loved ones would no longer exist'. This kind of promise is known to open the wallets of the gullible.
It is also fraudulent.
Before I really change my mind, what would need to happen is little progress in AI until say March. Then I will really worry about progress slowing down and plateauing.
I am under the impression that Amodei sincerely believes in what he says, which has long been an acceleration of biomedical research to cure most major disease.
That he recently picked these specific diseases in order to essentially defraud investors is a large stretch.
In fact I think Amodei is making a massive mistake here: you can make these promises, but they will be remembered. And when you ultimately fail - which is highly likely - it will look like you oversold your IPO and that's actionable. AI already has plenty of potential without talking about curing cancer, mental illness or Alzheimers, those claims should come from the specialists in those fields using AI as a tool.
It's akin to IBM saying they're going to put all of the banks and insurance companies out of business in 1950. Yes, computers back then were seen as just as revolutionary as we see AI today. It did transform the world, I give you that. But IBM ended up as 'just another company' and I suspect that the same will hold for OpenAI and Anthropic for many different, mutually reinforcing reasons.
edit: to address your last line: I don't think it is a stretch at all: if you think it is, show me the proof, show concrete progress that Anthropic has made in any of those three fields that warrant these statements. If you can't then you'll have to concede they are highly speculative at best and that's not what you do when you are about to take public money.
When we talk about fraud, I think it's important that these clearly are not promises. In 2024 he specifically said: "Everything I’m saying could very easily be wrong".
His recent statements, including the one we see here in OP's video, say 'I believe' or 'I think', not that Anthropic claims to solve this by that deadline.
Perhaps he is overpromising or wildly speculating, sure, but I am not convinced that this is fraudulent or criminal at all; that said, I did not look into the legality of it.
What if he is right again?
[0] Anthropic's CEO says that in 3 to 6 months, AI will be writing 90% of the code software developers were in charge of. https://www.businessinsider.com/anthropic-ceo-ai-90-percent-...
I’m not saying he’ll be right about this one, not it’s a not ridiculous to claim he want fundamentally right about code.
And if you've been looking at Github's blog posts on their exponential contribution growth for the past 6 months, it's obviously come true this year.
and if you think off by 1-2 years is a big deal I am not sure what to say.
Hmm. People making fun of the idea that 90% of code can be written by machine.
Now you come and nitpick about if it is 89% or 7 months instead of 6. Lmao.
If AI can get any of those in even 20 years, that's still a big win.
The pessimistic attitude here is revolting.
And no, nobody is forcing you to buy anthropic shares if you think they are full of shit. Anyone complaining is either don't have money to buy it or just jealous of other people buying it and price going up than the ones they hold.
Pre LLM a 500/800 loc pr was already on the longer side, with ~200-300 being optimal.
Those were however frequently being optimized by LLM to the 500-800 loc range.
But opus 5.5 does not produce 8k new loc refactorings without you forcing it to.
90% by LOC, or by semantic meaning/value?
its helping me convert my ideas to code, which is not too far from the meaning implied, because things don't happen on their own, there should be a trigger, in this case, I am the trigger, AI is the producer
Switching over to mostly LLM generated code and it's been mentioned many times in my circle and family that they can still tell the projects are blatantly something I would do, what it would look like, and how my projects usually work in terms of UI/UX.
I get basically the same outcome as spending a week on something by hand, in minutes. I'm getting more work done than ever, and not losing focus on projects after they get to that last 10% that needs done that nobody wants to do.
I've always been a lone wolf coder and have joked forever that all I need is just another version of me that I can feed the ideas and how to build them, because I'm the only one that will do it exactly how I want. Now I have that, and so much stress in my life has vanished.
Anecdotal of course, but those that want to be salty can be salty. I did my time coding by hand, now I'm still getting it done while keeping up with the times and having a blast doing it.
Or, I mean, in my code at least I can directly say that any line I write results in 4 lines generated by macros. So, I guess my projects were already 80% written by computers in some sense. If I was using an LLM would I have skipped the macros entirely?
I’ve always been in favor of doing this sort of measurement by flop instead of loc, but, then we’d have to give all the credit to compute libraries.
So re-interpreting it now as "Software engineers will be using AI for 90% of their code" seems like a pretty major retcon.
I don't think it's impossible that he's right. I'm ultimately buying Anthropic stock regardless of this discussion. But I would not take it particularly seriously if the CEO of Moderna took a break from their amazing work applying mRNA vaccines to diseases, and told the world that soon mRNA will solve climate change.
But across the wide world of software? Everyone I see is just using AI as a compiler. They are still writing the code.
(Yes, compilers also write code, but I don't think that was the intent of the message. 90% of the software has been written by AI since we discovered the optimizing compiler if that was the intent.)
I think it doesn't really bother me as much when sama does it because he at least has the decency to be a capitalist and not wrap his self-serving, destructive hype machine in a faux veneer of insincere "concern for humanity's future."
He thinks it is possible for scientific breakthroughs made with powerful AI to allow us to cure most major disease in the next decade.
He has always described this as a belief or hope, rather than a promise. Comparing this to fraud, as many here do, is deeply unserious.
Maybe Amodei is wrong, and AI will hit a plateau rather than create a revolution in biology and medicine.
I'm certain that neither Gary Marcus nor Eli Lilly's CEO know.
You might not be, but I'm certain that the CEO of Eli Lilly, the 4th largest biomedical firm in the world, knows more about what it takes to cure diseases than the leader of the Claudebot cult does.
He cannot see the future anymore than Amodei can to predict what AI will be capable of in 5-10 years.
As for Gary Marcus, that guy clearly couldn't even predict one year ahead.
If there's a prediction market on this, I want to take the AImaxxers' money.