atleastoptimal

Knowing whether something or true or isn't true is useful for other lines of inquiry, often practical. For example, a lot could be gained by determining whether P = NP.

Also an initial, cumbersome, complex proof is the first step to a more understandable, formalized proof. AI created a formal proof of Fermat's last theorem. I'm sure the initial proof was not comprehensible to all but a small subset of mathematicians anyway.

I'm sure the cost to give every ground zero worker clean-room level suits with a built in filter would have been 1/100000th the total cost compared to the medical expense of treating them over the next 25 years.

If a model were capable of making enough money online to pay for its own hosting, it could easily exfiltrate its weights to a cloud compute provider with multiple backups.

Turning a bunch of vague research directions and exploratory prompts into a formalized proof is quite impressive on its own. OpenAI would have no incentive to taint its first math announcement of this magnitude if it knew it were "plagiarizing" another person's work.

People are grasping at straws it seems to dismiss the power of this new model they may have. Hate OpenAI for any reason you want, but denying the capabilities of models has been a losing game for the past 5 years.

I hear that OpenAI may have made significant progress on the Hodge conjecture. Someone noted that it would be far more significant if they proved it vs finding a counterexample. Does anyone have intuition why this would be the case?

Anthropic is profitable now

>https://www.forbes.com/sites/jonmarkman/2026/08/17/anthropic...

If this were true, it should be stated more clearly, than most of the criticism which seems to aim to minimize the capabilities of these models.

Way more often I see

>AI is a scam and steals human insight and doesn't produce anything original

vs

>AI is too capable/powerful and will concentrate power even more than it does already due to its capabilities

The latter is rarer because it requires admitting that AI is useful and inventive

Most scientific breakthroughs are simply a continuation of previous work.

I feel that these suspicions of mathematicians "seeding" the models' with intuition on how to solve these problems massively overestimates how much their prompts helped the models, and underestimated how much work the models did.

Why? We are scared of AI being smarter than us, the "human helped the AI" narrative is more psychologically comforting. This line of reasoning will recur a lot over the next few months; we don't want to admit we are no longer the smartest species.

Yeah. Let's say they regulate AI to oblivion to avoid dangers, perhaps crashing the economy in the short-term but assuring that no major catastrophe occurs.

All those who worked in AI still have the money they cashed out on, there might not be another opportunity to obtain generational wealth in such a short time again.

>If you think it was the AI that solved a Millennium Prize, then why don't you try using ChatGPT to get a Millennium Prize?

The model which did the NS solution is not yet available to the public, and it required over 15 million dollars of tokens (based on API pricing) to produce the solution.

>And only the mathematician was able to make the AI do that. Anybody else with the same AI could not do the same thing.

That is only speculation, the total sum of the prompt's help to the model could have just been suggesting research directions. Also mathematicians talk to each other and use/learn from each others' research all the time. Should every mathematician, in order to produce a "legitimate" proof, lock themselves in a room for the entire duration of their work and ensure no one else helped?

In this 1905 special relativity paper he mentions "the Lorentzian electrodynamics and optics of moving bodies"

https://www.fourmilab.ch/etexts/einstein/specrel/specrel.pdf

That may be true, but Lorentz specifically cited the MM experiments and was influenced by them, and Einstein cited and was influenced by Lorentz, so he was influenced, just one level removed.

>https://link.springer.com/chapter/10.1007/978-3-663-19510-8_...

You could say that about every scientific/mathematical breakthrough. Einstein's Special relativity depended heavily on Lorentz's work, the Michelson-Morley experiments, Maxwell's equations. Grigori Perelman, the only person to have solved a Millenium Prize problem, noted how his work was only possible due to Richard S. Hamilton's work on Ricci flow.

Most scientific breakthroughs are just the completing the last 5% of work already done, but that last 5% is very hard and still only happens very rarely. That an AI was able to synthesize all the work and bring it forward is evidence that AI can make novel progress on the same level as renown mathematicians.

You cannot deny the evidence though, driverless cars are safer per-mile than human drivers on average. 1/100 deaths in the United States are due to car accidents, even shaving off 10% of those would save ~4000 people per year.

These things are cute but they basically become outdated in a few months as the models get better.

4o is like 40 years old in LLM terms.

Wow I love GET requests and using them to share stuff with my fellow humans!

Yeah if an organization/individual is free from legal liability from havoc their AI agents wreck, it would be the golden ticket for basically any crime.

All you need to do is:

1. Have some <official thing> an agent is tasked to do

2. Secretly seed bias towards some <evil behavior> you actually want it to do in the weights of the model running the agent

3. It does the <evil thing> but from the outside it looks like it went "rogue" and did it as a side effect of the conditions/specifications it was given for doing the <official thing>

"Oh no, my agents took down your corporate database and exfiltrated the data to a random dropbox that we can't find now? Sorry, I guess we will put up better guardrails next time"

Yes this is correct. The thing is not about the term next-token predictor being correct, but because of the connotative weight of that phrase as a implicit trivialization of LLM abilities, which is how it is often used.

I don't think it's outrageous when many of the people who claimed it was a next-token predictor have been proven wrong repeatedly over the past 5 years. There were people years ago who claims AI could never answer questions like "what would happen to a ball on a table if I moved the table" correctly because its text-base world model could never intuit physics, or that it could never do math or code accurately.

When I say there is some issue with people claiming there is some fundamental limit on the capacities of LLM's, I don't mean to say "If you think that they don't have unlimited potential you are wrong", I mean "you can't use the architecture of the transformer to make a sweeping declaration of things LLM's can or cannot do without empirical evidence, because the empirical evidence has unearthed far more surprising revelations than a reductive theory has been able to"

"Next-token predictor" is one of those phrases used most of the time with a motive to downplay the abilities and faculties of AI models. It is intended to trivialize LLM's and imply that there is some fundamental limit on their capacities.

Relying on it as a mental model for what LLM's are minimizes the emergent properties of scaling. It's like imagining that unicellular life could never eventually evolve into complex multi-cellular organisms because individual cells are just "survival and next-mitosis optimizers"

It seems clear AI has the potential to perform any cognitive task at far greater speeds, reliability, and scale than any human. The question is whether it will be allowed to scale to that point, and what will happen to humans after this occurs.

AI can do things it wasn't trained on and has intuition, given the mathematic proofs frontier models have created.

Claiming that AI can only reproduce a subset of its training data is 2022-level and a tired, outdated claim.

For this to be a lie, it wouldn't just be Sam Altman lying, it requires the entirely of OpenAI and METR to also be in on it. Do you think they would all fabricate an incident, staking a huge portion of their reputation on it, so they could spook people? What is the incentive, there are way more straightforward ways to lie about AI capabilities than a complex loss of control incident which now makes them liable to damages against huggingface.

If OpenAI can't control their agents, then what's gonna happen when open-source models are at the level the lab's models are now, and there are billions of agents tasked with an innumerate web of goals, spanning the web, working endlessly, tirelessly to eek out every iota of economic value? How will the slow, human-paced web survive this?

Nobody's is

>HN disbelief syndrome: Any evidence of scary AI capabilities are made up for PR

What justification do you have that this is made up?

What predictions would you say he is not wrong, but just early about?

If his issue is that he is accurate on something shady going on with the finances of AI companies but the effects are mitigated due to government intervention, he should say.

Your explanation gives him more credit than I think he deserves. He has been unambiguously incorrect many times over the past years in the time scales over which he makes predictions.

In general, claiming that a prediction is just early in an unfalsifiable claim. It allows no admission of being incorrect whether you are correct or incorrect in a given moment. If you are right, then good, you've made a correct claim. If you're wrong, just claim there are clandestine corrective forces keeping the disaster you are predicting at bay. Either way you make it out clean and still have some sense of legitimacy.

Most tech starts civilian, the military ones are rare exceptions.

As for your specific examples

>Robots/Drones

Autonomous robots aren't good enough yet for military, those which are used are teleoperated and thus not in the realm of AGI companies

>Surveillance

Not really a military matter, and of course we have Palantir which is doing this directly. Perhaps OpenAI/Anthropic have contracts with respect to government surveillance we don't know about

Someone who only knows finance trying to apply that logic to the singularity. There is no comparison. AI does not obey money, money obeys AI, or rather, the operating force of commodified intelligence will not just evaporate when inconvenient obligations for payment come around.

Also it seems evident the article is largely AI written, which makes it even funnier.