It is normal, expected, and healthy for stakeholders in a regulatory environment to offer proposals about regulations. What's unhealthy is the proposition that the deliberation process is so fragile that a stakeholder needs to cover every angle, lest they corrupt the outcome.
Trustbusting should absolutely be included as well. One of the biggest immediate threats is the concentration of wealth into a very tiny number of companies.
DMCA-style fines should be retroactively + prospectively applied to copyrighted works reproduced by AI, paid for by the AI companies, paid out to the copyright holders whose work was used without permission.
It would not be prohibitively hard to do the math on this.
That would fix a lot of the problems with AI overnight, but it'll also never happen.
Dario's been beating the regulatory capture drum for several years at various intervals, always in the name of safety, but it's hard to not see how self-serving it is.
I'm personally very tired of reading the linear-algebra-median of every AI safety essay from lesswrong with the inserted opinion of "therefore all my competitors, especially those pesky open source ones from scary countries should be illegal, only I can be trusted to not abuse the computer god that I definitely will have in just a couple more releases and with a couple more trillion invested"
Its hard to read the first half of this as anything other than regulatory capture propaganda. It really all ties together as:
> AI has become a major commercial technology
>Frontier AI models, like airplanes, should be required to go through technical testing and auditing, and their release should be blocked or reversed as a threat to public safety if they do not meet high standards of safety
> AI companies that develop advanced AI models must have strong security standards that protect their model weights
Anyway Dario's financial interests aside. This is an interesting breakpoint for me.
> Second, any response to AI-driven job displacement needs to address both the need to provide for everyone economically, and the need for people to find meaning, purpose, and agency. The latter is ultimately more important
To me this reads as an out of touch statement. I think the majority of people on earth work to keep a roof over their heads. Of course work can be a source of meaning, purpose, and agency, but to call it the more important aspect on a societal level is a sort of rich person like Dario statement to make.
Can you explain how this is attempted regulatory capture? To start a lab now which can actually compete for the frontier (i.e. and pass the "Threshold of compute" needed to get regulated) a lab / company would need a ton of money. Surely a well funded operation of that kind can deal with the regulations.
For one thing, the financial barrier may not necessarily stay that high forever.
For another, such regulations could prevent a competitor from making the weights open for their model to try and disrupt the competition.
And finally, Amodei would no doubt want to be involved in designing the tests the AI needs to pass, and could (and likely would) design it in a way that Anthropic models would be able to pass easier than competing models.
> A wide range of pro-employment policy incentives can help to slow or reduce job displacement, including: wage insurance policies that compensate people when they have to take a lower-paying job, retention tax incentives to encourage employers not to make layoffs, workforce training grants, or infrastructure to facilitate matching of employers to employees to speed the rate of labor market adaptation. While the particulars of which interventions are best will depend on what kind of labor displacement AI brings, we should readily accept the costs and market inefficiencies that these policies could entail, particularly as they are likely to be offset by AI-driven productivity gains.
People get income from one of three places: capital income, labor income, or the welfare state. If this technology truly unlocks a holy panacea of productivity with a commensurate drop in employment then capital’s share of the national income can and should provide for a wider and deeper welfare state. Nothing new need be invented here. Dario’s long and only somewhat organized list of policy interventions makes appropriate preparedness sound like a manic pulling of any and all levers when a simple theory of distribution will suffice.
This and we already did a dry run of ad-hoc distributions with COVID relief. They had to use the data from tax filings but it did work in terms of getting the money out there.
> If this technology truly unlocks a holy panacea of productivity with a commensurate drop in employment then capital’s share of the national income can and should provide for a wider and deeper welfare state.
This isn't guaranteed in the tax system as it exists today, because reinvestments into further growth are often treated as expenses which cancel out the income for tax purposes.
> A nation that possesses powerful AI facing one without it—or even facing one that is behind in AI by 3 years—could be the equivalent of an army of World War II Marines facing an army of medieval swordsmen.
This is a somewhat ironic take from someone who very publicly feuded with the US government about whether their AI could be used for waging war.
I know this is likely just for IPO hype but when I read things like this I sometimes wonder if I must be missing something. I use agents everyday and find them really useful and they save me a lot of headache. At the same time I find that if I let it self-direct at a high level at all it generally makes bad choices that cause me headaches later so I can’t really give them autonomy. Enough people seem to believe this exponential line of thinking though that I keep having to wonder: am I the one missing something here? Is there some magic tool that I haven’t found yet that will cure cancer?
If we already were at the point that AI could self-direct effectively, then the world would already be very different (eg AI-driven technological progress and unemployment) in a way that we might have wished we prepared for more.
Do you mean policy-wise (like Dario is talking about), or more broadly?
I wonder about broad preparedness, but unfortunately there's not a lot that we "normal" people can do to prepare. Hoard savings and food? Learn physical trades?
You're talking quite statically though. I don't think anyone is worried about today's models being a serious threat, but next year's, three years' time? Just three years' ago these models were useful bumbling fools and it's hard to judge where on the S-curve we currently are.
I'd rather be thinking about these issues in advance rather than waiting until the problem becomes real.
> I don't think anyone is worried about today's models being a serious threat
Fable is essentially bricked for my areas of interest (even being a member of the cybersecurity program). It seems like they’re attempting to sell regulatory capture under the guise of safety. That’s more of the point.
It's nice that people are genuinely curious about this.
- All of your observations are absolutely dead on
- Yet, we have very very very robust scaling laws that as Dario points out we've had and validated for over a decade. This extends to downstream measures like METR time horizon and compsosite benchmarks like the epoch capability index.
- If you look at where you're at now, which is again dead on, you're looking at a point on a curve that is quite easy to extrapolate, but less easy to tell when exactly on the curve a certain capability or use case undergoes a step change from error rates dropping below a threshold that is hard to anticipate in advance.
So while Dario / other frontier CEOs are understandably unpalatable, they are absolutely spot on with a call out that all of this is bound to happen and happen quickly, and that's without solving several core problems that haven't been solved yet (e.g. continual learning). In 2023, coding agents were just laughable. Yet they followed the same predictable training curves. Anyone looking at the data can see the obvious, and anyone reading newspaper headlines or hacker news comments would get a very different impression.
This is a very tech-focused message board, populated by mostly tech-insiders, so perhaps a little outside perspective will help people understand.
Tech people are following a religious belief system whose utopian promise is the all-powerful computer that will end all suffering. I once read an article in reason magazine from over 30 years ago about how an advanced computer in the future will bring everyone who has ever lived back from the deat and let them live in paradise. They were completely serious. Atheists reading this may object to my description of the tech belief system as religious, but I believe it is accuarte. The idea that tech is an imrpovement and will improve people's lives is believed as an act of faith. Tech has its own moral systems based on some form of libertarian progressivism. And in the future, through the inevitable scientific magic of exponential something, a computer will ascend to godhood and judge all mankind for their actions before allowing some into eternal paradise.
To what extent any of this is true is up for debate, but most west coast tech elite are actively working towards this future, and these are the ideas that drive them. It's hard to talk to them about it because this is their woldview, and they imagine everyone to believe what they do.
> Tech people are following a religious belief system whose utopian promise is the all-powerful computer that will end all suffering.
Uh, I don't really think that's anywhere close to an accurate characterization of most people here. Everyone, including Dario and any researcher at any frontier lab, knows the situation is quite scary and unprecedented. There are problems that will be solved and diseases that will be cured, but will we be living in an Orwellian universe? Will a rogue drone swarm find you cowing in your basement and murder you? I mean the technology for this is already mostly here, it's a matter of the willpower and budget to roll out something really evil.
The comment's question is about capabilities and why the discussion about capabilities often times is far removed from todays capabilities.
It's crazy how all these tech CEOs develop the same sense of ethics that seeks to make the foundation of open research and development that made their efforts possible and may threaten their market position illegal in the name of safety against nebulously-defined risks
While I do understand the risks, I don't understand the solution. Essentially, Dario is saying that powerful model weights can't be distributed (ban open weights), and governments should coordinate and agree on standards, and block any dangerous model from being used at all, with government deciding what dangerous means.
Okay, I don't understand how legitimate access is granted then. Surely, Dario isn't saying to ban Sonnet, because I can definitely make it do cyber harm, as most exploits that I've seen in the wild with my own eyes were trivial.
So the only way I see his proposal working is:
- No open weights, AI is centralized in the hands of few
- We get AI-FAA that sets the rules and monitors
- If I want to do a security scan of my codebase, I get a time and scope limited license from AI-FAA that I upload to claude that will allow it to run the security scan in cloud with their models - Claude Mythos Scanner(TM).
Dario's proposal ultimately requires that people lose direct access to inference via API. Is this why they've been building SaaS clones with AI bolted on?
It is impressive how well they've scheduled all their releases, posts, and other news to dominate the tech news cycle almost every day in this pre-IPO phase.
It's very hard those days to think of companies/people more arrogant than Anthropic/Dario, which is quite the achievement as the bar is very high.
If that arrogance was well placed at least you could somewhat excuse it, but the fact that it is so overtly hypocritical and based on false premises just makes it so much worse.
>> Of all tyrannies, a tyranny sincerely exercised for the good of its victims may be the most oppressive. It would be better to live under robber barons than under omnipotent moral busybodies. The robber baron’s cruelty may sometimes sleep, his cupidity may at some point be satiated; but those who torment us for our own good will torment us without end for they do so with the approval of their own conscience. They may be more likely to go to Heaven yet at the same time likelier to make a Hell of earth.
I read this essay, and it feels like lying behind a mask of moral responsibility and safety for humanity.
They are asking for FAA style preclearance and third party audits. That literally means no new AI startup can emerge. Do they not know that audits cost money?
Protect your own monopoly, protect your customers' regulations. They want strong regulation like the FAA to raise barriers to entry for the foundation models they themselves build, but then why do they want to loosen FDA regulations? While at the same time driving token consumption from their own customers.
They talk about permanent job displacement and UBI. I usually call this "a morally packaged safe landing."
They are doing something unpopular (destroying jobs) and getting criticized for it. But they do not want to be criticized further, and they want to ask for social sympathy. So they claim a 'noble cause' that everyone can sympathize with and that is safe for themselves
AI will generate astronomical productivity gains and capital profits, which AI companies privatize. So why should the social costs be paid by national taxes? In my opinion, something like "We will donate all of our AI companies' revenue for the next 10 years to society" would show genuine sincerity.
Then they say, if we do not develop AI, China will eat our lunch, and they go after China. But is not this really about preventing Chinese dumping, maintaining our own token prices, and asking the world to beat down China so that they can preserve global tech hegemony?
But by blocking China from the CUDA ecosystem, now the CANN ecosystem has emerged, has it not? If China develops techniques that reliably reduce inference costs, who knows how things will turn out then.
Honestly, I like Anthrpic's Claude, but the Anthropic CEO's rhetoric is so stale. It is not that it feels hypocritical. It is that this is just a one dimensional rhetorical tactic that assumes the public is stupid.
I do not think open source is unconditionally good. (It is good, but it can become bad in all situations or all countrie). Open source itself is a barrier for countries outside the Anglosphere when they want to release IT products. Because there is no incentive to buy a product that is worse than an open source alternative. So I do not think everything necessarily has to be open source.
But this (referring to Anthropic's position) seems to treat people like fools. If regulation is needed, shouldn't they also argue that FDA regulation is needed? I wish they would be consistent
Regarding regulation: I'm deeply invested in computer vision systems and i fear that policymakers [who are not deeply familiar with the technical distinctions between AI systems] may write broad rules that cover "AI" generally. In that case, computer vision companies and industrial users could end up subject to requirements that were largely motivated by concerns about generative AI and LLMs.
What will be Amodei's job after we have AIs that are better at evrything than humans? Is the AI going to care about our stock exchange playgrounds that reward the future Antropic stock holders?
> after we have AIs that are better at evrything than humans.
That this is worded so definitively is a testament to the success of the AI industry. The idea that LLMs will be "better at evrything than humans"[sic] is far from certain.
I suspect that if someone does invent a machine like this, it won't look like a 2026 LLM, and it will be far far in the future. everybody relax.
Well obviously he'll have the AI 'dispose' of the poor and live a life as a king with a select few farmed humans and have the world as a play thing.
Really the entire future of AI at this point seems like "Don't worry about it, we'll figure out when we get there". Works a lot better if you're extremely rich and can afford your own private security.
Can't find the article mentioning it but apparently it's an open problem they're thinking about.
But yeah if society collapses these billionaire nerds are the first to go. Quietly, in their bunkers, while the team leader of their seal mercenary team takes over.
Even before the rest of us realizes what's happening.
> Finally, the CEO of a brokerage house explained that he had nearly completed building his own underground bunker system, and asked: “How do I maintain authority over my security force after the event?” The event. That was their euphemism for the environmental collapse, social unrest, nuclear explosion, solar storm, unstoppable virus, or malicious computer hack that takes everything down.
This single question occupied us for the rest of the hour. They knew armed guards would be required to protect their compounds from raiders as well as angry mobs. One had already secured a dozen Navy Seals to make their way to his compound if he gave them the right cue. But how would he pay the guards once even his crypto was worthless? What would stop the guards from eventually choosing their own leader?
The billionaires considered using special combination locks on the food supply that only they knew. Or making guards wear disciplinary collars of some kind in return for their survival. Or maybe building robots to serve as guards and workers – if that technology could be developed “in time”.
I tried to reason with them. I made pro-social arguments for partnership and solidarity as the best approaches to our collective, long-term challenges. The way to get your guards to exhibit loyalty in the future was to treat them like friends right now, I explained. Don’t just invest in ammo and electric fences, invest in people and relationships. They rolled their eyes at what must have sounded to them like hippy philosophy.
Even loyal security would hardly be sufficient. If a million people have decided they want your head on a pike, even a billionaire cannot afford a big enough army. And what exactly is a billionaire in that world anyway?
> Members of the trusted coalition should freely share chips and semiconductor manufacturing equipment (SME) with each other, while working together to deny it to adversaries. US export controls on frontier chips and SME to China have been a major contributor to the US’s overall lead in AI, and these policies need to be expanded, tightened, and coordinated with other likeminded states.
I understand why Dario thinks this is crucial, but it's a very dystopian view of the medium-term future.
I'm not an optimist to the point that I believe that AI will lead to global Star Trek-style utopia (although it theoretically could), but ongoing disparity between "allied" and "enemy" powers relating to hardware technology and software models is both not really possible to enforce in the long term, and a pretty dismal state of global affairs even if successful.
I'd be interested in an expert geopolitical opinion on what the long tail of this would really look like in any sort of reasonable reality.
>The government should have the power to block or deter deployment of the model if it is determined, in light of third-party assessment, to present unacceptable risks. This power must be scoped to the above four specific risks and there must be protective measures against political favoritism or arbitrary decisions.
I feel significantly less sympathy for Anthropic's Supply Chain Risk designation if they believe the government should have this power over them. You get what you sign up for.
Ooof a Yann LeCun quote from '24, a risky time period to be mining his quotes from lol. But I think broadly he's right, but the argument is: we don't see any evidence for the curve flattening, and we should plan for it _not_ to flatten any time soon. "It will eventually slow down" is true but its meaningless if you dont put any sort of time period on this.
"AI is advancing at a lightning pace—in only four years, AI models have gone from barely being able to write a coherent line of code to writing most of the code at major AI companies. Similar gains have been made in biology, physics, math, finance, law, translation, and many other fields."
This is a massive exaggeration. The advancement in the automation of computer code writing has been impressive and is obviously, at least in the short term, changing the software engineering industry substantially. Most other fields have not been affected to nearly the same degree. Certainly not biology, physics, finance, and law (I don't know enough about the math and translation fields to speak to those).
---
"3. Accelerating AI’s positive impact..."
This whole section is the type of thing that often comes out of the mouths of Silicon Valley tech executives without a pharmaceutical science background. It indicates a thorough lack of understanding of the realities of pharmaceutical research. What he is describing here is removing many of the solid, evidentiary rules that are in place to make sure that the drugs reaching the market actually work and replacing them with proxy predictions. Look, my least favourite part of the job is the animal testing, and I would be hugely grateful if that could be eliminated from the drug discovery pipeline. People have been trying to do that for a long time. But it's extremely difficult. Biology is very, very complicated. Our understanding of how processes in organisms work are vague and approximative. This is not computer code. Even if Anthropic somehow got all of Big Pharma to hand them their proprietary data, it would only scratch the surface of the understanding that is needed to solve these kind of problems. Due to these realities, the program Amodei is describing here would, effectively, open a floodgate of drugs on the market that don't actually do what they are supposed to and are more likely to have unidentified toxicity.
I appreciate the critical perspective on political and economic power, as long as it's consequently followed, and every willingness for cooperation and the creation of fair rules is good.
What makes me doubt that Dario Amodei has really internalized the problem is the lack of humility, the stance that it's just important that the "good guys" keep the technology away from the "bad guys".
If you really want to provide AI with public benefit, you need to prevent power concentration. How? Some unpolished ideas, I'd be happy to hear yours:
- Avoid getting too close to an administration that is openly attacking democracy and is not interested in the benefit of humanity or mutually beneficial cooperation.
- Don't support surveillance. Non-(US-)Americans have human rights and privacy, too. Prepare for a situation where a government tries to convert your compute infrastructure into surveillance infrastructure.
- Support the creation of community data centers. In other words, build data centers together with local communities and make sure they profit from them.
- Advocate for laws that require transparency about resource usage and emissions of data centers.
- If you don't want an AI race, make sure that other countries don't need to fear the US concentrating too much power. Create institutions that can be trusted by other countries, too.
EDIT: I forgot:
- If qualified labor will actually turn out to get devalued, we also need a plan to prevent states from turning into rentier states that don't depend on a well-educated society any longer.
(On the other hand, I have been fooled by too many billionaires claiming to act in the interest of democracy and freedom. I once fell for a billionaire buying a social media platform, claiming to be a free speech fundamentalist, and it didn't age too well.)
I feel completely baffled by the other responses on this thread. People viewing this purely as a marketing stunt, regulatory capture or attack on their freedoms, with seemingly no appreciation of the real threat that AI could pose to society and even humanity given its current rate of progress.
I'm not going to claim that the CEO of pre-IPO company has no incentive to bolster the claims of his tech, but to completely disregard everything he is saying based on that seems awfully binary.
I don't know whether people are just high on copium, spouting "it's just fancy autocomplete" or "only humans can really be creative" on every LLM-related thread, but it is impossible to deny that in a span of a few years we've gone from models that could barely put together a sentence, to something maybe not equivalent to a junior developer, but at least resembling it.
And sure, you can point out every flaw that current day LLMs have, just how everyone pointed out that Stable Diffusion couldn't generate accurate hands (until it could 6 months later!). But the gradient is pretty clear and I am yet to see a well-argued narrative from anyone why scaling laws should fail in the next year or two (by which point it feels like we're going to have a real problem, extrapolating the current trajectory).
I'm very glad this discussion is at least being had, and I wish everyone would get off their high-horse and take things a bit more seriously.
Or simply we use AI and see on the ground what it can and can't do. I can generally trust an agent for solved problems, but the more something deviates from established industry standards (i.e. what was relentlessly scraped) I have an increasingly harder time not having constant oversight of what it's doing, no matter the specs I put on the md.
Personally I feel most of the improvement in the last year comes from tooling/integration (MCPs, realtime documentation access, treesitter support, orchestration) than from the models themselves, in the last year. And still frontier models would routinely come up with bs until you tell them to actually use those tools.
You're talking as if this is a static thing though. It's the God of Gaps [1] but for humanity's special sauce.
Two years ago, I couldn't trust an LLM to do anything that wasn't straight forward boiler plate.
One year ago, I was pretty solid at writing algorithms that were combinations of existing ideas.
Now, Fable is outputting stuff that I would genuinely consider to be creative and original if a colleague had presented it to me.
Yes, maybe the code style still isn't great, but given the pattern of the last few years, it feels correct (a priori) to assume that this gap isn't going to keep closing.
In terms of aeronautics, went from the Wright brothers to the moon in 40 years. After that, everyone understandably thought that we'd be living in space and flying everywhere with personal jet packs in another couple years. Little did they know, it was the top of the S-curve, not the middle.
In the 60 years since, we've barely been able to adapt the 737 to fly longer routes.
Sure, I have no doubt that AI progress will follow an S-curve. The question is, where are we on it and is the plateau at a level safe for humanity? That's a very difficult thing to estimate without a crystal ball and not a risk I want to take.
> I'm not going to claim that the CEO of pre-IPO company has no incentive to bolster the claims of his tech
I am going to say it. The CEO of a pre-IPO company has extreme incentive to bolster the tech he is selling, to the point where his every action should be viewed as only in service to that goal. Every word he says should be viewed critically through that lense. He is not making this post out of the goodness of his heart, he is doing it in service to the IPO. If it happens to align with your views that's great, but it's still just a marketing stunt to get people with your views to buy in. Don't be fooled. Buy in if you feel it's a good deal, not because of the CEO's marketing.
I agree that AI poses a threat to society. I act on this by not developing world-leading AI models and offering them to anyone willing to pay top dollar, while funneling that money back into accelerating AI capabilities. Maybe Dario would consider taking a similar ethical position? Maybe he would support restrictions and taxes on data center construction, in order to slow down the pace?
If Anthropic were not developing these models, one of many other companies would be. I think it's good that the CEO of the current world-leader is at least considering these discussions and platforming possible solutions.
The fact that he doesn't support more restrictive approaches that don't align with his incentives doesn't invalidate the points he is making.
It’s a guy asking to make an end-run around the constitution and the APA regulatory framework based on a flimsy sci fi premise. Naturally it provokes a negative reaction.
Why are they not preaching for protected weights, but public, ie, under state control? What do you feel his posture will be if that is starting to be discussed?
Also, on an unrelated note, why would you have an account for 5 years and only now post your second comment? AI has been an existential threat for years, why only now?
This is a pattern I am seeing all over the place on HN in the last year in AI threads, and I have to admit that I am starting to become paranoid and my feels need some assuaging.
Your first point is very reasonable, and I agree that that is something Dario would likely be more opposed to.
However, my point isn't that I think Dario is our saviour who we should follow the every word of. As with everyone, his opinions should be filtered through the lens of his incentives. That said, I don't understand the knee-jerk reaction by many commenters to completely disregard the many important points he's making.
As for the lack of my account use, I can't comment for others, but I'm just quite shy. I've opened up the comment box many times to write a reply but rarely commit to actually posting it, especially because I feel like I'm not on the side of the general HN consensus.
Humans just aren't very good at dealing with threats that aren't immediate concerns. 'Safety regulations are written in blood' is a saying for a reason. A significant chunk of the population shrugs off climate change, and nearly all fertility rate crises threads are filled with dumb 'hurr lower population good' and/or 'See what Capitalism gets you!' rhetoric - They fundamentally don't even understand what the problem is. So is it really all that surprising that a technology like this would be shrugged off until it's too late ? Especially one with such existential issues for humanity? Some people are still too loathe to admit the clearly intelligent machine is intelligent, devolving into increasingly nonsensical and absurd (and ironically more human demeaning) arguments as model capabilities get better. I'm afraid you're asking for too much.
It's denialism, same as climate change, the subconscious fear to really grapple with the actual "what if" alternative scenario. Anthropic are true believers. They got to $1T in 5 years by being exceptionally smart and ahead of the curve here. Meanwhile HN just continually devolves into reactionary cynicism. "must be marketing, they just want to be rich, impossible AI advances much further." Meanwhile at every step of the way, Anthropic and X-riskers / "doomers" are vindicated in their correct predictive beliefs. We're headed to a future far dangerous than nukes very soon. We're in an arms race to detonate one 100x the size
Honest question: is there a reason for the naming conventions for these models? Anything that makes it better than giving them names with model numbers, like “Claude 3” or such?
Each of the nouns is a “size class” in literature. From small lines poem (haiku, sonnet) to larger story (fable) to very large story (opus) to culture-defining foundational (myth).
It’s a fun way to say how many parameters are in the model without revealing a number like 405B or 17B which isn’t really comparable vs other models.
What about punishment on AI exponential for security failures, electric chairs for CEO of high ranked frontier models that gets jailbroken by geopolitical adversaries.
This reads like an AI with an overflowing context window wrote it; or in the alternative it’s a list of statutes written by an arrogant and delusional king. It is this type of arrogance that will lead to an unfavorable reaction by Congress.
This was slightly more thoughtful than I feared it would be. However, what I still do not see in any posts from any of these AI pushers is any genuine consideration of the possibility that the best thing might be for them to do less of what they are currently doing. Not to do it differently. Not to mitigate it. Not to do something else in addition. But to actually reduce their current activities.
> AI models have gone from barely being able to write a coherent line of code to writing most of the code at major AI companies.
Gasoline has gone from barely being able to power stationary farm machines to now being the fuel that underpins our entire economy. So, great news all around, right?
> which predict an exponential increase
And was that actually delivered?
Real question: If a model goes from 80% accurate to 85% accurate is that an exponential increase in "cognitive capabilities?" Are we considering training costs and effort?
I have to be honest, I am tired of reading these arrogant, self-absorbed posts from Dario and Anthropic in general. Opening with this lord of the rings reference just feels like they are trying too hard and are untrustworthy.
As annoying as their tone is, the real big danger is what they are setting up for. All this fear-mongering around Mythos, the overly aggressive controls on Fable, and these manifestos they keep writing, are part of setting up for REGULATORY CAPTURE. Even collaborating with the Pope and the Interfaith Alliance (https://iafsc.org/our-work/faith-ai-covenant) are part of creating a vast support network for regulations and restrictions. Those regulations will help those faith organizations or the government or whatever, but will also help Anthropic’s bottom line.
Those regulations will not support your civil liberties. They will restrict speech, access to AI, and allowed uses of AI. They will lead to bans on use of models from some countries like China, and also bans on open-source or open-weight models.
If Dario wants to be trusted, he needs to explicitly say in writing that Anthropic will not support any legal or regulatory restrictions on open-source AI, open-weight AI, or Chinese models. Otherwise, what he is really saying - even as he claims he is trying to ‘defend democracy’ - is that he and Anthropic do not truly support fundamental rights like our right to speech.
It’s not just Anthropic either. OpenAI had their own recent polemic, pushing for regulations like mandatory safety reviews by agencies for “frontier” models (https://news.ycombinator.com/item?id=48387246). It’s a dead giveaway that these companies have no moats, are in serious danger of being a commodity, and are now in the process of using regulations and enshittification to hold onto money and power.
I mean he wants to be trusted but he also absolutely doesn't support freedom of speech in that sense, and I doubt anyone with power or influence over ai policy does?
If you find a good lobbying group with money who can push for it let HN know.
To be honest, at this point I’d take any group at all willing to usefully and non-hypocritically push for free speech (as a general societal value, not just a US-specific legal notion).
How predictable. The company currently on top wishes to use the regulatory power of the state to prevent competitors from encroaching on their market dominance. It’s a tail as old as time, although their CEO’s rarely publish blog posts about it.
> "Models above a threshold of compute should undergo mandatory testing by a qualified third party for their level of risk in four specific areas: cybersecurity, biological weapons, loss of control of AI systems, and automated R&D that could accelerate these other risks."
AKA: Make it as expensive and untenable as possible for any open source model to jump through the regulation red tape so we can pull the ladder up behind us.
Disgusting.
All the marketing talk about "this model is so dangerous" "omg we can't release this to the public its so dangerous" etc. is just priming for the incoming lobbying for protectionism from foreign competition, and regulation preventing the development of any other model that could threaten their dominance in the name of "safety"
If you mean https://news.ycombinator.com/item?id=48481131, users flagged it. We can only guess why users flag things. Perhaps it was because they thought the comment broke the site guidelines by being snarky and/or fulminatey?
I don't think it was an extreme case of that, so I've turned off the flags now.
"Democracies should seek to form a global coalition centered on building AI according to their common values, iteratively trying to draw in the rest of the world by making it more and more attractive to be part of the coalition and less and less attractive to be outside it."
It's not clear to me on which side of the coalition USA is meant to be in this divide. And as an European I'm not sure whether being in China's or USA's coalition is better in the long term.
In general, this deliberate mongering of ever more geopolitical division is extremely harmful. As is the Trump bootlicking.
The comments here. They make me feel that we are so doomed.
We all want to nuclear codes so badly. We are addicted to intelligence and labour so badly that we simply can't concieve that a pro-social actor might want us all not to have it, and for good reason.
I mean... Obviously, insiders like Oppenheimer (who dedicated their lives to considering the implications of the technology under discussion), they just feared nuclear proliferation because they wanted all the profits for themselves, right :(
I like that he comes up with new laws and regulations for AI companies. Can I suggest some more?
- You shall not embed copyrighted material in your models.
- You shall not bombard every little website in existence with 1 million scraping queries per day.
- You shall not use your political influence to pump and dump your AI (or rocket?) company.
- You shall not imperill the whole IT sector by buying all CPU and memory chips.
These new rules will affect every society directly in a positive way. Thanks.
"We need an approach to make sure AI doesn't destroy the world and wipe humanity to extinction."
"Yeah, and quotas on web scrapers!"
It is normal, expected, and healthy for stakeholders in a regulatory environment to offer proposals about regulations. What's unhealthy is the proposition that the deliberation process is so fragile that a stakeholder needs to cover every angle, lest they corrupt the outcome.
What you say is true but completely ignores the obvious ways in which what he is proposing benefits his company.
Its like saying it’s normal for a taxi driver to drive people places while he’s got you handcuffed in the trunk.
Trustbusting should absolutely be included as well. One of the biggest immediate threats is the concentration of wealth into a very tiny number of companies.
Yep we need new laws for this. The current laws end up in years long lawsuits and no real change.
DMCA-style fines should be retroactively + prospectively applied to copyrighted works reproduced by AI, paid for by the AI companies, paid out to the copyright holders whose work was used without permission.
It would not be prohibitively hard to do the math on this.
That would fix a lot of the problems with AI overnight, but it'll also never happen.
You missed a couple ofvery important ones:
- Your AI data centres will run only on renewable energy
- Your AI data centres will not use evaporative cooling
> AI companies that develop advanced AI models must have strong security standards that protect their model weights
So, basically, make open-weight models illegal. It's nice for Dario to come out and say this so explicitly.
Dario's been beating the regulatory capture drum for several years at various intervals, always in the name of safety, but it's hard to not see how self-serving it is.
I'm personally very tired of reading the linear-algebra-median of every AI safety essay from lesswrong with the inserted opinion of "therefore all my competitors, especially those pesky open source ones from scary countries should be illegal, only I can be trusted to not abuse the computer god that I definitely will have in just a couple more releases and with a couple more trillion invested"
People ask what the 'open' part of 'OpenAI' even means as they're a for-profit company.
It turns out it means 'opener than Anthropic.'
Opener Anthropic Imitator
"We Have No Moat, And Neither Does OpenAI"
Its hard to read the first half of this as anything other than regulatory capture propaganda. It really all ties together as:
> AI has become a major commercial technology
>Frontier AI models, like airplanes, should be required to go through technical testing and auditing, and their release should be blocked or reversed as a threat to public safety if they do not meet high standards of safety
> AI companies that develop advanced AI models must have strong security standards that protect their model weights
Anyway Dario's financial interests aside. This is an interesting breakpoint for me.
> Second, any response to AI-driven job displacement needs to address both the need to provide for everyone economically, and the need for people to find meaning, purpose, and agency. The latter is ultimately more important
To me this reads as an out of touch statement. I think the majority of people on earth work to keep a roof over their heads. Of course work can be a source of meaning, purpose, and agency, but to call it the more important aspect on a societal level is a sort of rich person like Dario statement to make.
Can you explain how this is attempted regulatory capture? To start a lab now which can actually compete for the frontier (i.e. and pass the "Threshold of compute" needed to get regulated) a lab / company would need a ton of money. Surely a well funded operation of that kind can deal with the regulations.
For one thing, the financial barrier may not necessarily stay that high forever.
For another, such regulations could prevent a competitor from making the weights open for their model to try and disrupt the competition.
And finally, Amodei would no doubt want to be involved in designing the tests the AI needs to pass, and could (and likely would) design it in a way that Anthropic models would be able to pass easier than competing models.
> A wide range of pro-employment policy incentives can help to slow or reduce job displacement, including: wage insurance policies that compensate people when they have to take a lower-paying job, retention tax incentives to encourage employers not to make layoffs, workforce training grants, or infrastructure to facilitate matching of employers to employees to speed the rate of labor market adaptation. While the particulars of which interventions are best will depend on what kind of labor displacement AI brings, we should readily accept the costs and market inefficiencies that these policies could entail, particularly as they are likely to be offset by AI-driven productivity gains.
People get income from one of three places: capital income, labor income, or the welfare state. If this technology truly unlocks a holy panacea of productivity with a commensurate drop in employment then capital’s share of the national income can and should provide for a wider and deeper welfare state. Nothing new need be invented here. Dario’s long and only somewhat organized list of policy interventions makes appropriate preparedness sound like a manic pulling of any and all levers when a simple theory of distribution will suffice.
This and we already did a dry run of ad-hoc distributions with COVID relief. They had to use the data from tax filings but it did work in terms of getting the money out there.
> If this technology truly unlocks a holy panacea of productivity with a commensurate drop in employment then capital’s share of the national income can and should provide for a wider and deeper welfare state.
This isn't guaranteed in the tax system as it exists today, because reinvestments into further growth are often treated as expenses which cancel out the income for tax purposes.
You’re conflating firm level taxable income with the national income.
> A nation that possesses powerful AI facing one without it—or even facing one that is behind in AI by 3 years—could be the equivalent of an army of World War II Marines facing an army of medieval swordsmen.
This is a somewhat ironic take from someone who very publicly feuded with the US government about whether their AI could be used for waging war.
I know this is likely just for IPO hype but when I read things like this I sometimes wonder if I must be missing something. I use agents everyday and find them really useful and they save me a lot of headache. At the same time I find that if I let it self-direct at a high level at all it generally makes bad choices that cause me headaches later so I can’t really give them autonomy. Enough people seem to believe this exponential line of thinking though that I keep having to wonder: am I the one missing something here? Is there some magic tool that I haven’t found yet that will cure cancer?
If we already were at the point that AI could self-direct effectively, then the world would already be very different (eg AI-driven technological progress and unemployment) in a way that we might have wished we prepared for more.
> we might have wished we prepared for more
Do you mean policy-wise (like Dario is talking about), or more broadly?
I wonder about broad preparedness, but unfortunately there's not a lot that we "normal" people can do to prepare. Hoard savings and food? Learn physical trades?
I've experienced the same.
That said Claude Code has a million features like loops that I know exist but never use.
I imagine that spending a lot more time creating an initial spec goes a long way towards independence, I just don't usually do that.
N+1. This is my experience and for the most part the people that I work with share the same feeling.
A highly enthusiastic concussion enthusiast with 10 hands is how one person put it.
These are people in different fields but highly accomplished so I’m feeling comfortable sharing their assessment.
You're talking quite statically though. I don't think anyone is worried about today's models being a serious threat, but next year's, three years' time? Just three years' ago these models were useful bumbling fools and it's hard to judge where on the S-curve we currently are.
I'd rather be thinking about these issues in advance rather than waiting until the problem becomes real.
> I don't think anyone is worried about today's models being a serious threat
Fable is essentially bricked for my areas of interest (even being a member of the cybersecurity program). It seems like they’re attempting to sell regulatory capture under the guise of safety. That’s more of the point.
It's nice that people are genuinely curious about this.
- All of your observations are absolutely dead on
- Yet, we have very very very robust scaling laws that as Dario points out we've had and validated for over a decade. This extends to downstream measures like METR time horizon and compsosite benchmarks like the epoch capability index.
- If you look at where you're at now, which is again dead on, you're looking at a point on a curve that is quite easy to extrapolate, but less easy to tell when exactly on the curve a certain capability or use case undergoes a step change from error rates dropping below a threshold that is hard to anticipate in advance.
So while Dario / other frontier CEOs are understandably unpalatable, they are absolutely spot on with a call out that all of this is bound to happen and happen quickly, and that's without solving several core problems that haven't been solved yet (e.g. continual learning). In 2023, coding agents were just laughable. Yet they followed the same predictable training curves. Anyone looking at the data can see the obvious, and anyone reading newspaper headlines or hacker news comments would get a very different impression.
This is a very tech-focused message board, populated by mostly tech-insiders, so perhaps a little outside perspective will help people understand.
Tech people are following a religious belief system whose utopian promise is the all-powerful computer that will end all suffering. I once read an article in reason magazine from over 30 years ago about how an advanced computer in the future will bring everyone who has ever lived back from the deat and let them live in paradise. They were completely serious. Atheists reading this may object to my description of the tech belief system as religious, but I believe it is accuarte. The idea that tech is an imrpovement and will improve people's lives is believed as an act of faith. Tech has its own moral systems based on some form of libertarian progressivism. And in the future, through the inevitable scientific magic of exponential something, a computer will ascend to godhood and judge all mankind for their actions before allowing some into eternal paradise.
To what extent any of this is true is up for debate, but most west coast tech elite are actively working towards this future, and these are the ideas that drive them. It's hard to talk to them about it because this is their woldview, and they imagine everyone to believe what they do.
> Tech people are following a religious belief system whose utopian promise is the all-powerful computer that will end all suffering.
Uh, I don't really think that's anywhere close to an accurate characterization of most people here. Everyone, including Dario and any researcher at any frontier lab, knows the situation is quite scary and unprecedented. There are problems that will be solved and diseases that will be cured, but will we be living in an Orwellian universe? Will a rogue drone swarm find you cowing in your basement and murder you? I mean the technology for this is already mostly here, it's a matter of the willpower and budget to roll out something really evil.
The comment's question is about capabilities and why the discussion about capabilities often times is far removed from todays capabilities.
It's crazy how all these tech CEOs develop the same sense of ethics that seeks to make the foundation of open research and development that made their efforts possible and may threaten their market position illegal in the name of safety against nebulously-defined risks
While I do understand the risks, I don't understand the solution. Essentially, Dario is saying that powerful model weights can't be distributed (ban open weights), and governments should coordinate and agree on standards, and block any dangerous model from being used at all, with government deciding what dangerous means.
Okay, I don't understand how legitimate access is granted then. Surely, Dario isn't saying to ban Sonnet, because I can definitely make it do cyber harm, as most exploits that I've seen in the wild with my own eyes were trivial.
So the only way I see his proposal working is:
- No open weights, AI is centralized in the hands of few
- We get AI-FAA that sets the rules and monitors
- If I want to do a security scan of my codebase, I get a time and scope limited license from AI-FAA that I upload to claude that will allow it to run the security scan in cloud with their models - Claude Mythos Scanner(TM).
Dario's proposal ultimately requires that people lose direct access to inference via API. Is this why they've been building SaaS clones with AI bolted on?
It is impressive how well they've scheduled all their releases, posts, and other news to dominate the tech news cycle almost every day in this pre-IPO phase.
It's very hard those days to think of companies/people more arrogant than Anthropic/Dario, which is quite the achievement as the bar is very high.
If that arrogance was well placed at least you could somewhat excuse it, but the fact that it is so overtly hypocritical and based on false premises just makes it so much worse.
>> Of all tyrannies, a tyranny sincerely exercised for the good of its victims may be the most oppressive. It would be better to live under robber barons than under omnipotent moral busybodies. The robber baron’s cruelty may sometimes sleep, his cupidity may at some point be satiated; but those who torment us for our own good will torment us without end for they do so with the approval of their own conscience. They may be more likely to go to Heaven yet at the same time likelier to make a Hell of earth.
-C.S. Lewis
I read this essay, and it feels like lying behind a mask of moral responsibility and safety for humanity.
They are asking for FAA style preclearance and third party audits. That literally means no new AI startup can emerge. Do they not know that audits cost money?
Protect your own monopoly, protect your customers' regulations. They want strong regulation like the FAA to raise barriers to entry for the foundation models they themselves build, but then why do they want to loosen FDA regulations? While at the same time driving token consumption from their own customers.
They talk about permanent job displacement and UBI. I usually call this "a morally packaged safe landing."
They are doing something unpopular (destroying jobs) and getting criticized for it. But they do not want to be criticized further, and they want to ask for social sympathy. So they claim a 'noble cause' that everyone can sympathize with and that is safe for themselves
AI will generate astronomical productivity gains and capital profits, which AI companies privatize. So why should the social costs be paid by national taxes? In my opinion, something like "We will donate all of our AI companies' revenue for the next 10 years to society" would show genuine sincerity.
Then they say, if we do not develop AI, China will eat our lunch, and they go after China. But is not this really about preventing Chinese dumping, maintaining our own token prices, and asking the world to beat down China so that they can preserve global tech hegemony?
But by blocking China from the CUDA ecosystem, now the CANN ecosystem has emerged, has it not? If China develops techniques that reliably reduce inference costs, who knows how things will turn out then.
Honestly, I like Anthrpic's Claude, but the Anthropic CEO's rhetoric is so stale. It is not that it feels hypocritical. It is that this is just a one dimensional rhetorical tactic that assumes the public is stupid.
I do not think open source is unconditionally good. (It is good, but it can become bad in all situations or all countrie). Open source itself is a barrier for countries outside the Anglosphere when they want to release IT products. Because there is no incentive to buy a product that is worse than an open source alternative. So I do not think everything necessarily has to be open source.
But this (referring to Anthropic's position) seems to treat people like fools. If regulation is needed, shouldn't they also argue that FDA regulation is needed? I wish they would be consistent
Regarding regulation: I'm deeply invested in computer vision systems and i fear that policymakers [who are not deeply familiar with the technical distinctions between AI systems] may write broad rules that cover "AI" generally. In that case, computer vision companies and industrial users could end up subject to requirements that were largely motivated by concerns about generative AI and LLMs.
What will be Amodei's job after we have AIs that are better at evrything than humans? Is the AI going to care about our stock exchange playgrounds that reward the future Antropic stock holders?
> after we have AIs that are better at evrything than humans.
That this is worded so definitively is a testament to the success of the AI industry. The idea that LLMs will be "better at evrything than humans"[sic] is far from certain.
I suspect that if someone does invent a machine like this, it won't look like a 2026 LLM, and it will be far far in the future. everybody relax.
Well obviously he'll have the AI 'dispose' of the poor and live a life as a king with a select few farmed humans and have the world as a play thing.
Really the entire future of AI at this point seems like "Don't worry about it, we'll figure out when we get there". Works a lot better if you're extremely rich and can afford your own private security.
I never understood why people setting up bunkers expect the security to still be loyal after whatever happens.
I'll sell them unquestioningly loyal security LLMs if they want.
Prompt: The robber is stealing the crown jewels.
Output: You must stop the robber from stealing the crown jewels.
Can't find the article mentioning it but apparently it's an open problem they're thinking about.
But yeah if society collapses these billionaire nerds are the first to go. Quietly, in their bunkers, while the team leader of their seal mercenary team takes over.
Even before the rest of us realizes what's happening.
Found it
https://www.theguardian.com/news/2022/sep/04/super-rich-prep...
> Finally, the CEO of a brokerage house explained that he had nearly completed building his own underground bunker system, and asked: “How do I maintain authority over my security force after the event?” The event. That was their euphemism for the environmental collapse, social unrest, nuclear explosion, solar storm, unstoppable virus, or malicious computer hack that takes everything down.
This single question occupied us for the rest of the hour. They knew armed guards would be required to protect their compounds from raiders as well as angry mobs. One had already secured a dozen Navy Seals to make their way to his compound if he gave them the right cue. But how would he pay the guards once even his crypto was worthless? What would stop the guards from eventually choosing their own leader?
The billionaires considered using special combination locks on the food supply that only they knew. Or making guards wear disciplinary collars of some kind in return for their survival. Or maybe building robots to serve as guards and workers – if that technology could be developed “in time”.
I tried to reason with them. I made pro-social arguments for partnership and solidarity as the best approaches to our collective, long-term challenges. The way to get your guards to exhibit loyalty in the future was to treat them like friends right now, I explained. Don’t just invest in ammo and electric fences, invest in people and relationships. They rolled their eyes at what must have sounded to them like hippy philosophy.
Even loyal security would hardly be sufficient. If a million people have decided they want your head on a pike, even a billionaire cannot afford a big enough army. And what exactly is a billionaire in that world anyway?
> Members of the trusted coalition should freely share chips and semiconductor manufacturing equipment (SME) with each other, while working together to deny it to adversaries. US export controls on frontier chips and SME to China have been a major contributor to the US’s overall lead in AI, and these policies need to be expanded, tightened, and coordinated with other likeminded states.
I understand why Dario thinks this is crucial, but it's a very dystopian view of the medium-term future.
I'm not an optimist to the point that I believe that AI will lead to global Star Trek-style utopia (although it theoretically could), but ongoing disparity between "allied" and "enemy" powers relating to hardware technology and software models is both not really possible to enforce in the long term, and a pretty dismal state of global affairs even if successful.
I'd be interested in an expert geopolitical opinion on what the long tail of this would really look like in any sort of reasonable reality.
>The government should have the power to block or deter deployment of the model if it is determined, in light of third-party assessment, to present unacceptable risks. This power must be scoped to the above four specific risks and there must be protective measures against political favoritism or arbitrary decisions.
I feel significantly less sympathy for Anthropic's Supply Chain Risk designation if they believe the government should have this power over them. You get what you sign up for.
>AI Exponential
How much of the policy prescription changes if the exponential is actually just a series of sigmoids[1]?
[1]: https://x.com/ylecun/status/1799064075487572133
Ooof a Yann LeCun quote from '24, a risky time period to be mining his quotes from lol. But I think broadly he's right, but the argument is: we don't see any evidence for the curve flattening, and we should plan for it _not_ to flatten any time soon. "It will eventually slow down" is true but its meaningless if you dont put any sort of time period on this.
"AI is advancing at a lightning pace—in only four years, AI models have gone from barely being able to write a coherent line of code to writing most of the code at major AI companies. Similar gains have been made in biology, physics, math, finance, law, translation, and many other fields."
This is a massive exaggeration. The advancement in the automation of computer code writing has been impressive and is obviously, at least in the short term, changing the software engineering industry substantially. Most other fields have not been affected to nearly the same degree. Certainly not biology, physics, finance, and law (I don't know enough about the math and translation fields to speak to those).
---
"3. Accelerating AI’s positive impact..."
This whole section is the type of thing that often comes out of the mouths of Silicon Valley tech executives without a pharmaceutical science background. It indicates a thorough lack of understanding of the realities of pharmaceutical research. What he is describing here is removing many of the solid, evidentiary rules that are in place to make sure that the drugs reaching the market actually work and replacing them with proxy predictions. Look, my least favourite part of the job is the animal testing, and I would be hugely grateful if that could be eliminated from the drug discovery pipeline. People have been trying to do that for a long time. But it's extremely difficult. Biology is very, very complicated. Our understanding of how processes in organisms work are vague and approximative. This is not computer code. Even if Anthropic somehow got all of Big Pharma to hand them their proprietary data, it would only scratch the surface of the understanding that is needed to solve these kind of problems. Due to these realities, the program Amodei is describing here would, effectively, open a floodgate of drugs on the market that don't actually do what they are supposed to and are more likely to have unidentified toxicity.
The only effective action is push back on Athropic.
I appreciate the critical perspective on political and economic power, as long as it's consequently followed, and every willingness for cooperation and the creation of fair rules is good.
What makes me doubt that Dario Amodei has really internalized the problem is the lack of humility, the stance that it's just important that the "good guys" keep the technology away from the "bad guys".
If you really want to provide AI with public benefit, you need to prevent power concentration. How? Some unpolished ideas, I'd be happy to hear yours:
- Avoid getting too close to an administration that is openly attacking democracy and is not interested in the benefit of humanity or mutually beneficial cooperation.
- Don't support surveillance. Non-(US-)Americans have human rights and privacy, too. Prepare for a situation where a government tries to convert your compute infrastructure into surveillance infrastructure.
- Support the creation of community data centers. In other words, build data centers together with local communities and make sure they profit from them.
- Advocate for laws that require transparency about resource usage and emissions of data centers.
- If you don't want an AI race, make sure that other countries don't need to fear the US concentrating too much power. Create institutions that can be trusted by other countries, too.
EDIT: I forgot:
- If qualified labor will actually turn out to get devalued, we also need a plan to prevent states from turning into rentier states that don't depend on a well-educated society any longer.
(On the other hand, I have been fooled by too many billionaires claiming to act in the interest of democracy and freedom. I once fell for a billionaire buying a social media platform, claiming to be a free speech fundamentalist, and it didn't age too well.)
https://xcancel.com/cuimao/status/2058458683781365873
I feel completely baffled by the other responses on this thread. People viewing this purely as a marketing stunt, regulatory capture or attack on their freedoms, with seemingly no appreciation of the real threat that AI could pose to society and even humanity given its current rate of progress.
I'm not going to claim that the CEO of pre-IPO company has no incentive to bolster the claims of his tech, but to completely disregard everything he is saying based on that seems awfully binary.
I don't know whether people are just high on copium, spouting "it's just fancy autocomplete" or "only humans can really be creative" on every LLM-related thread, but it is impossible to deny that in a span of a few years we've gone from models that could barely put together a sentence, to something maybe not equivalent to a junior developer, but at least resembling it.
And sure, you can point out every flaw that current day LLMs have, just how everyone pointed out that Stable Diffusion couldn't generate accurate hands (until it could 6 months later!). But the gradient is pretty clear and I am yet to see a well-argued narrative from anyone why scaling laws should fail in the next year or two (by which point it feels like we're going to have a real problem, extrapolating the current trajectory).
I'm very glad this discussion is at least being had, and I wish everyone would get off their high-horse and take things a bit more seriously.
Or simply we use AI and see on the ground what it can and can't do. I can generally trust an agent for solved problems, but the more something deviates from established industry standards (i.e. what was relentlessly scraped) I have an increasingly harder time not having constant oversight of what it's doing, no matter the specs I put on the md.
Personally I feel most of the improvement in the last year comes from tooling/integration (MCPs, realtime documentation access, treesitter support, orchestration) than from the models themselves, in the last year. And still frontier models would routinely come up with bs until you tell them to actually use those tools.
You're talking as if this is a static thing though. It's the God of Gaps [1] but for humanity's special sauce.
Two years ago, I couldn't trust an LLM to do anything that wasn't straight forward boiler plate.
One year ago, I was pretty solid at writing algorithms that were combinations of existing ideas.
Now, Fable is outputting stuff that I would genuinely consider to be creative and original if a colleague had presented it to me.
Yes, maybe the code style still isn't great, but given the pattern of the last few years, it feels correct (a priori) to assume that this gap isn't going to keep closing.
[1] https://en.wikipedia.org/wiki/God_of_the_gaps
In terms of aeronautics, went from the Wright brothers to the moon in 40 years. After that, everyone understandably thought that we'd be living in space and flying everywhere with personal jet packs in another couple years. Little did they know, it was the top of the S-curve, not the middle.
In the 60 years since, we've barely been able to adapt the 737 to fly longer routes.
Sure, I have no doubt that AI progress will follow an S-curve. The question is, where are we on it and is the plateau at a level safe for humanity? That's a very difficult thing to estimate without a crystal ball and not a risk I want to take.
> I'm not going to claim that the CEO of pre-IPO company has no incentive to bolster the claims of his tech
I am going to say it. The CEO of a pre-IPO company has extreme incentive to bolster the tech he is selling, to the point where his every action should be viewed as only in service to that goal. Every word he says should be viewed critically through that lense. He is not making this post out of the goodness of his heart, he is doing it in service to the IPO. If it happens to align with your views that's great, but it's still just a marketing stunt to get people with your views to buy in. Don't be fooled. Buy in if you feel it's a good deal, not because of the CEO's marketing.
I agree that AI poses a threat to society. I act on this by not developing world-leading AI models and offering them to anyone willing to pay top dollar, while funneling that money back into accelerating AI capabilities. Maybe Dario would consider taking a similar ethical position? Maybe he would support restrictions and taxes on data center construction, in order to slow down the pace?
If Anthropic were not developing these models, one of many other companies would be. I think it's good that the CEO of the current world-leader is at least considering these discussions and platforming possible solutions.
The fact that he doesn't support more restrictive approaches that don't align with his incentives doesn't invalidate the points he is making.
It’s a guy asking to make an end-run around the constitution and the APA regulatory framework based on a flimsy sci fi premise. Naturally it provokes a negative reaction.
Why are they not preaching for protected weights, but public, ie, under state control? What do you feel his posture will be if that is starting to be discussed?
Also, on an unrelated note, why would you have an account for 5 years and only now post your second comment? AI has been an existential threat for years, why only now?
This is a pattern I am seeing all over the place on HN in the last year in AI threads, and I have to admit that I am starting to become paranoid and my feels need some assuaging.
Your first point is very reasonable, and I agree that that is something Dario would likely be more opposed to.
However, my point isn't that I think Dario is our saviour who we should follow the every word of. As with everyone, his opinions should be filtered through the lens of his incentives. That said, I don't understand the knee-jerk reaction by many commenters to completely disregard the many important points he's making.
As for the lack of my account use, I can't comment for others, but I'm just quite shy. I've opened up the comment box many times to write a reply but rarely commit to actually posting it, especially because I feel like I'm not on the side of the general HN consensus.
You write well & HN is better when there are more well-written people on the opposite side of the consensus
Humans just aren't very good at dealing with threats that aren't immediate concerns. 'Safety regulations are written in blood' is a saying for a reason. A significant chunk of the population shrugs off climate change, and nearly all fertility rate crises threads are filled with dumb 'hurr lower population good' and/or 'See what Capitalism gets you!' rhetoric - They fundamentally don't even understand what the problem is. So is it really all that surprising that a technology like this would be shrugged off until it's too late ? Especially one with such existential issues for humanity? Some people are still too loathe to admit the clearly intelligent machine is intelligent, devolving into increasingly nonsensical and absurd (and ironically more human demeaning) arguments as model capabilities get better. I'm afraid you're asking for too much.
It's denialism, same as climate change, the subconscious fear to really grapple with the actual "what if" alternative scenario. Anthropic are true believers. They got to $1T in 5 years by being exceptionally smart and ahead of the curve here. Meanwhile HN just continually devolves into reactionary cynicism. "must be marketing, they just want to be rich, impossible AI advances much further." Meanwhile at every step of the way, Anthropic and X-riskers / "doomers" are vindicated in their correct predictive beliefs. We're headed to a future far dangerous than nukes very soon. We're in an arms race to detonate one 100x the size
Honest question: is there a reason for the naming conventions for these models? Anything that makes it better than giving them names with model numbers, like “Claude 3” or such?
Yes.
Each of the nouns is a “size class” in literature. From small lines poem (haiku, sonnet) to larger story (fable) to very large story (opus) to culture-defining foundational (myth).
It’s a fun way to say how many parameters are in the model without revealing a number like 405B or 17B which isn’t really comparable vs other models.
What about punishment on AI exponential for security failures, electric chairs for CEO of high ranked frontier models that gets jailbroken by geopolitical adversaries.
This guy can’t stay a day without posting something more or less “ban open source AI”. We keep you safe
This reads like an AI with an overflowing context window wrote it; or in the alternative it’s a list of statutes written by an arrogant and delusional king. It is this type of arrogance that will lead to an unfavorable reaction by Congress.
This was slightly more thoughtful than I feared it would be. However, what I still do not see in any posts from any of these AI pushers is any genuine consideration of the possibility that the best thing might be for them to do less of what they are currently doing. Not to do it differently. Not to mitigate it. Not to do something else in addition. But to actually reduce their current activities.
AI is a lethal risk to mankind and should be totally rolled back for a century of close research before we try again.
> AI models have gone from barely being able to write a coherent line of code to writing most of the code at major AI companies.
Gasoline has gone from barely being able to power stationary farm machines to now being the fuel that underpins our entire economy. So, great news all around, right?
> which predict an exponential increase
And was that actually delivered?
Real question: If a model goes from 80% accurate to 85% accurate is that an exponential increase in "cognitive capabilities?" Are we considering training costs and effort?
Yeah I’m lost by the constant use of “exponential”. What is the x and y axis when the talking heads say this?
Can we not open up every article talking to working professionals as if they’re children?
I like to stay up to date on things but more and more I’m finding myself pointing codex at a URL and saying “get to the point”.
He wrote it with ai that’s why
I think we should treat this with smirking suspicions until their IPO happens.
My god this guy is insufferable. Stop mis-using the term exponential
Looks pretty exponential to me [1]. From a fully independent, non-profit research group.
[1] https://metr.org/time-horizons/
I have to be honest, I am tired of reading these arrogant, self-absorbed posts from Dario and Anthropic in general. Opening with this lord of the rings reference just feels like they are trying too hard and are untrustworthy.
As annoying as their tone is, the real big danger is what they are setting up for. All this fear-mongering around Mythos, the overly aggressive controls on Fable, and these manifestos they keep writing, are part of setting up for REGULATORY CAPTURE. Even collaborating with the Pope and the Interfaith Alliance (https://iafsc.org/our-work/faith-ai-covenant) are part of creating a vast support network for regulations and restrictions. Those regulations will help those faith organizations or the government or whatever, but will also help Anthropic’s bottom line.
Those regulations will not support your civil liberties. They will restrict speech, access to AI, and allowed uses of AI. They will lead to bans on use of models from some countries like China, and also bans on open-source or open-weight models.
If Dario wants to be trusted, he needs to explicitly say in writing that Anthropic will not support any legal or regulatory restrictions on open-source AI, open-weight AI, or Chinese models. Otherwise, what he is really saying - even as he claims he is trying to ‘defend democracy’ - is that he and Anthropic do not truly support fundamental rights like our right to speech.
It’s not just Anthropic either. OpenAI had their own recent polemic, pushing for regulations like mandatory safety reviews by agencies for “frontier” models (https://news.ycombinator.com/item?id=48387246). It’s a dead giveaway that these companies have no moats, are in serious danger of being a commodity, and are now in the process of using regulations and enshittification to hold onto money and power.
I mean he wants to be trusted but he also absolutely doesn't support freedom of speech in that sense, and I doubt anyone with power or influence over ai policy does?
If you find a good lobbying group with money who can push for it let HN know.
To be honest, at this point I’d take any group at all willing to usefully and non-hypocritically push for free speech (as a general societal value, not just a US-specific legal notion).
https://en.wikipedia.org/wiki/Foundation_for_Individual_Righ... has been doing alright recently?
How predictable. The company currently on top wishes to use the regulatory power of the state to prevent competitors from encroaching on their market dominance. It’s a tail as old as time, although their CEO’s rarely publish blog posts about it.
Yep
> "Models above a threshold of compute should undergo mandatory testing by a qualified third party for their level of risk in four specific areas: cybersecurity, biological weapons, loss of control of AI systems, and automated R&D that could accelerate these other risks."
AKA: Make it as expensive and untenable as possible for any open source model to jump through the regulation red tape so we can pull the ladder up behind us.
Disgusting.
All the marketing talk about "this model is so dangerous" "omg we can't release this to the public its so dangerous" etc. is just priming for the incoming lobbying for protectionism from foreign competition, and regulation preventing the development of any other model that could threaten their dominance in the name of "safety"
Yeah, they know their moat is evaporating as the Chinese models continue to catch up, and at a much cheaper price.
Also, why has my comment been flagged? It is sitting at positive votes but has suddenly been flagged for no apparent reason?
If you mean https://news.ycombinator.com/item?id=48481131, users flagged it. We can only guess why users flag things. Perhaps it was because they thought the comment broke the site guidelines by being snarky and/or fulminatey?
I don't think it was an extreme case of that, so I've turned off the flags now.
I mean it’s pretty tame and actually adds to the conversation.
It feels somewhat obvious that someone just disagreed and flagged it.
Is there a system for holding people accountable who abuse flags?
Sit back and watch the stock market implode as these lot double down on safety theatre. The hype is coming to an end.
They have to keep it up at least until the IPO dust settles.
> The hype is coming to an end
I'm sure you have evidence for this
"Democracies should seek to form a global coalition centered on building AI according to their common values, iteratively trying to draw in the rest of the world by making it more and more attractive to be part of the coalition and less and less attractive to be outside it."
It's not clear to me on which side of the coalition USA is meant to be in this divide. And as an European I'm not sure whether being in China's or USA's coalition is better in the long term.
In general, this deliberate mongering of ever more geopolitical division is extremely harmful. As is the Trump bootlicking.
[flagged]
Please don't post unsubstantive comments to Hacker News, regardless of who is a goober or you believe they are.
The comments here. They make me feel that we are so doomed.
We all want to nuclear codes so badly. We are addicted to intelligence and labour so badly that we simply can't concieve that a pro-social actor might want us all not to have it, and for good reason.
I mean... Obviously, insiders like Oppenheimer (who dedicated their lives to considering the implications of the technology under discussion), they just feared nuclear proliferation because they wanted all the profits for themselves, right :(