Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Navier-Stokes – Tristan Buckmaster [pdf]

cims.nyu.edu

Post reply on HN
Print view View original

531–540 of 862 posts

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#531

Post by m00x »

Catloafdev wrote: post #515
vlovich123 wrote: post #497

Earlier quoted context omitted.

I'm not sure it's so easy to tell whether a given piece of data was in a training run at their scale. It's entirely possible they think the answer is no, but on the off-chance that it could be, they'd rather not say no and then later it turns out they did and then they're claimed to be lying. If you were them, unless you could 100% rule it out, you'd hedge and say you can't.

It may not be easy, quick, or simple to figure that out - absolutely fair. But it is knowable. Their entire business is built around training models - they have the ability to know exactly what was in any given training run. I guess time will tell.

It would be very difficult to say. It confirms that Tristan's data is likely part of the data the models use, but a lot of filtering, pruning, and transform goes into training.

Data has to be determined to be signal and not just noice, then it could go through processes of generating questions/answers from that data, then it RLHF's over this.

OpenAI have petabytes of data, all anonymized. It could take months to say for sure it was part of the training, and even more time to determine if it made any difference.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#532

Post by davesque »

hkmaxpro wrote: post #336
sigbottle wrote: post #305

Earlier quoted context omitted.

It's specifically the last two bullet poitns - Tristan is suspicious of the timing, as only few others were trying this approach. OpenAI says the model didn't access his user data directly, but leaves unanswered whether Tristan's chat conversations were part of the training. - OpenAI says they would partially credit Tristan for the $1,000,000 discovery (even though Tristan did not solve the $1,000,000 problem) — but…

Both Sam Altman and Sebastien Bubeck admitted they only want Buckmaster to be the lead author on a rewrite of the OpenAI proof. https://x.com/sama/status/2097385167002415140 https://x.com/SebastienBubeck/status/2097379411691516310 A wake up call for using OpenAI models. If you discover something with their model and you work for a competitor, they “felt it would be inappropriate” for you “to author OpenAI’s work”.

Honestly this whole thing is so fucking weird. I feel like there's an argument that absolutely no one involved in the final crossing of the finish line to the proof actually did any work (other than just intelligently directing an LLM) and deserves any credit. As the author of this doc mentions, the mathematicians who did the actual work that led to the formulation of this approach (without the use of LLMs; just good ole' fashioned human intellect) are the ones who deserve the credit.

Imagine that a no name janitor used their time in the evenings to go spelunking through the literature to push an LLM to this result. No one would care because that person isn't an anointed expert. So why would the expert deserve any more credit? Because they sort of understand the result, even if they couldn't have achieved it on their own? The whole issue of credit for AI-assisted discoveries seems like it's going to run into a brick wall pretty soon.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#533

Post by fn-mote »

mti wrote: post #338
bawolff wrote: post #297

Earlier quoted context omitted.

> OpenAI says they would partially credit Tristan for the $1,000,000 discovery (even though Tristan did not solve the $1,000,000 problem) — but only if they remove Levent as an author, as he works for Anthropic well that sounds like an asshole move.

In a serious discipline like mathematics, this isn't just an asshole move, but a career-ending level of academic misconduct.

> a career-ending level of academic misconduct

There is no such thing anymore.

Falsified data and published? Absolutely no problem. Keep your tenure.

It’s even hard to lose your position as president of a university due to egregious misconduct.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#534

Post by s1artibartfast »

Drblessing wrote: post #271
bhouston wrote: post #191

Earlier quoted context omitted.

> Ignoring the drama But the drama here is a little important. Stealing the millennium prize for N-S is sort of a big deal, especially to those who had been working on it for the last few years.

The math is all that matters.

What do you even mean by that? It is the human value?

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#535

Post by fn-mote »

fc417fc802 wrote: post #489
myrmidon wrote: post #447

Earlier quoted context omitted.

Just replace the model with a human student. "Training" on textbooks => fine "Training" with unpublished notes from another professor, then publishing something on that exact topic with a similar approach without giving any credit => extremely questionable.

Presumably the professor voluntarily provided the notes in this analogy. I think the student would also be expected to cite the textbook if building off of it directly. In contrast, humans are generally not expected to cite "general inspiration" or what have you. So if we're to apply human standards, and assuming that the model was trained on the relevant work, it would only be plagiarism if the model directly built…

To be clear, the accusation is that they trained on the chats they used while working on the problem. Not published work or even a preprint.

Your post does not distinguish, and it matters.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#536

Post by acchow »

Catloafdev wrote: post #425
hkmaxpro wrote: post #336

Earlier quoted context omitted.

Both Sam Altman and Sebastien Bubeck admitted they only want Buckmaster to be the lead author on a rewrite of the OpenAI proof. https://x.com/sama/status/2097385167002415140 https://x.com/SebastienBubeck/status/2097379411691516310 A wake up call for using OpenAI models. If you discover something with their model and you work for a competitor, they “felt it would be inappropriate” for you “to author OpenAI’s work”.

These responses seem to me to make it abundantly clear who's telling the truth here. I wonder who this fools. It would be extraordinarily easy to simply say, this model was not trained on your work, if that were the case. It's telling that they refuse to acknowledge the root issue here, and are attempting to shift the conversation elsewhere.

> It would be extraordinarily easy to simply say, this model was not trained on your work, if that were the case.

The Huggingface Attack revealed that making blanket statements like this is difficult and requires quite a bit of manual labor:

1) the agents spin for days and produce too much output to review 2) using LLMs to process that output skips many important details

Ergo, the agent could likely decide it would like to look through actual user data, hack its way into that data, and produce way too much output for a human to decide whether or not this occurred.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#537

Post by yk »

n2d4 wrote: post #83

Drama/accusation summary: - Aug 15th: Tristan Buckmaster & Levent Alpöge make progress on a few important math problems, "finite-time blowup with smooth forcing for incompressible porous media, for Boussinesq, and for 3d incompressible Euler." - they do NOT have a proof for the $1,000,000 Millenium Prize problem. BUT, they do claim to have a proof for a similar (non-Millenium) Navier Stokes problem that could help le…

The last two points are disputed/sound significantly more reasonable in [0]. So from what I gather, Buckmaster realizes sometime during the call that the biggest result of his career is going to get steamrolled (the blowup of Navier Stokes is a much bigger deal than the blowup of 3D Euler), and on the other hand the openAi guys realize that they are basically talking about internal results with Anthropic and probably have to call corporate right after this call. Between these two stressor the conversation appears to have gone somewhat poorly.

[0] https://x.com/SebastienBubeck/status/2097379411691516310

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#538

Post by m00x »

n2d4 wrote: post #83

Drama/accusation summary: - Aug 15th: Tristan Buckmaster & Levent Alpöge make progress on a few important math problems, "finite-time blowup with smooth forcing for incompressible porous media, for Boussinesq, and for 3d incompressible Euler." - they do NOT have a proof for the $1,000,000 Millenium Prize problem. BUT, they do claim to have a proof for a similar (non-Millenium) Navier Stokes problem that could help le…

I haven't seen any proof that OpenAI asked Tristan to remove Sebastian from the prize. Until we have proof of this, it would be wise to offer conclusions.

Same for NS validity. This was not validated by the community yet.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#539

Post by fn-mote »

treis wrote: post #429
mswphd wrote: post #403

Earlier quoted context omitted.

it's very possible they only had to use the massive compute budget because they were trying to plagiarize his work before he published it though, e.g. autonomously do things in ~7 days what he had likely been thinking about for ~1 year.

This doesn't really make sense. You don't need massive amounts of computing to plagiarize something. The most nefarious explanation seems to be that they got wind it was possible to solve NS via LLMs and perhaps a small nudge in the right direction.

> You don't need massive amounts of computing to plagiarize something

The compute was used to leapfrog the human team, using their ideas and pushing them to a solution of the general problem.

Plagiarism isn’t being used in the literal sense.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#540

Post by hellohello2 »

fc417fc802 wrote: post #438
lukewarm707 wrote: post #337

Earlier quoted context omitted.

if they do not deny training on them, they can't deny plagiarism.

By that logic everything any LLM spits out is plagiarizing the vast majority of work written prior to a few months ago. That doesn't seem like a useful or desirable line of argument to me.

This is a common misconception, so its understandable that you have it. Generative models can both plagiarize and generalize. The question here is which of the two happened.
Post reply on HN