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

681–690 of 871 posts

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#681

Post by johnnienaked »

adastra22 wrote: post #573
djokkataja wrote: post #556

Earlier quoted context omitted.

The first part of the sentence is important: > Importantly it was admitted that internal Anthropic models had been used in their proof of Euler blowup; I therefore felt I could not consider Levent to be an independent academic. If an Anthropic employee is doing independent research, but with models that aren't available to the public (because they're internal models), then . . . idk. It's not clear to me why that sho…

This is being reported as OpenAI wanting to strip an Anthropic employee of academic credit for the work they did. What the OpenAI person involved is claiming is that they wanted the outside researcher(s) to put their name on OpenAI's work: to headline OpenAI's publication of what they earnestly believed to be an independent result. If true, that's generous and beyond the level of generosity one should expect. Extendi…

It's a ridiculous explanation that makes no sense.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#682

Post by Gud »

piloto_ciego wrote: post #677

Is this being astroturfed? Like, to me this looks like academic slap-fighting from Bubeck and Levent. People working at OpenAI are saying, "hey, we don't have that particular data in our models," others are saying, "we used a different approach to do it with Navier-Stokes" this feels like much ado about nothing. Then in these comments I see some wild accusations. If OpenAI is telling the truth (I don't really see a r…

You’re making a lot of assumptions not based on facts. Looks more like this to me: math wizards uses ChatGPT to assist solving a math problem. OpenAI gobbles up the prize.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#683

Post by tovej »

adastra22 wrote: post #573
djokkataja wrote: post #556

Earlier quoted context omitted.

The first part of the sentence is important: > Importantly it was admitted that internal Anthropic models had been used in their proof of Euler blowup; I therefore felt I could not consider Levent to be an independent academic. If an Anthropic employee is doing independent research, but with models that aren't available to the public (because they're internal models), then . . . idk. It's not clear to me why that sho…

This is being reported as OpenAI wanting to strip an Anthropic employee of academic credit for the work they did. What the OpenAI person involved is claiming is that they wanted the outside researcher(s) to put their name on OpenAI's work: to headline OpenAI's publication of what they earnestly believed to be an independent result. If true, that's generous and beyond the level of generosity one should expect. Extendi…

Nothing ethically justifiable about it, even if you take their word for it.

If OpenAI thinks someone should be the lead author for a paper, that person should have full discretion to decide who the co-authors are.

Even if the co-author's contributions were non-technical

So no, there's no world where you can ethically extend the right to publish a result and decide who the authors are from the outside.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#684

Post by rlt »

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.

"which is when I said that I did not understand why one would risk their career [over unfounded accusations]. Genuinely, at that moment, I was trying to care for him"

"Our aim was to see whether our system was also capable of this impressive feat"

"OpenAI's intention was to do everything possible to celebrate their mathematical achievements and the heroic efforts that they made on Euler"

For some reason I have a hard time believing people when they use language like this.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#685

Post by remus »

scott_weber wrote: post #543
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 should be quite easy: if they don't leak the user session data publicly, and don't commingle it with training data internally, how could it possibly end up in the training data? What surprises me is they're not more boldly/plainly lying about it.

How would they know for sure that some details were not part of some other training data they use? The authors may have discussed some tangential details on a forum for example, in which case you might argue that the model picked up on these details the authors assumed were benign but novel and worked out how to apply them to the problem.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#686

Post by tristanj »

20k wrote: post #21

>I asked whether the model had been trained on, or had access to, our sessions in Codex, into which we had been putting all our drafts for the whole of this project. I was told the model did not look up user data. I asked again, about training, and I did not get an answer. If you think these companies are not training on your prompts you are incredibly naive. These models were built by stealing and pirating literally…

OpenAI cannot give a definitive answer here, because it is genuinely unknowable if Buckmaster's data is in the training set.

OpenAI explicitly uses user feedback (the thumbs up or thumbs down ratings), as RLHF to train models. However, this feedback is anonymized and stripped of user identifiers. If Buckmaster ever used this feature, then that conversation would be anonymized, saved, and used for training, but not tied back to him.

They cannot issue a blanket denial (which people so desperately desire), and instead repeat that "it's very unlikely" (which pisses people off), because they cannot in good faith claim to have zero data at all.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#687

Post by Tanjreeve »

m00x wrote: post #494
20k wrote: post #451

Earlier quoted context omitted.

That doesn't make it fine. We should not excuse this behaviour just because its rampant already, especially when it comes to such a serious prize

You're getting a massive discount because you're helping to train the model. If you want to have ZDR, you have to pay API rates. This is well-known to anyone in the industry.

> If you want to have ZDR, you have to pay API rates.

That's a pinky swear. Especially as data gets harder to come by I'm curious how long till there's a scandal on that too.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#688

Post by 405error »

rochansinha wrote: post #663

People here do not seem to be considering the second-order effects of these series of events. No academic institution or enterprise will trust OpenAI, Anthropic or any other non-local AI model with their core IP. There will be severe restrictions on what employees at these companies/institutions can share with AI services even from their personal accounts. (Or I am just overthinking it)

Many academics and grad students I know have closed source their in progress work, and started being really careful about what they chat with LLMs (or using local ones) because of the drama around this. No one wants four years of their life getting sniped by ten million dollars worth of tokens.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#689

Post by rzmmm »

epistasis wrote: post #391
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…

> OpenAI says the model didn't access his user data directly, but leaves unanswered whether Tristan's chat conversations were part of the training. This sort of cagey half-answer is highly suspicious and indicates that yes OpenAI did actually "access user data directly" because they are only willing to say that the "model did not access user data." That has a very specific meaning, the model looking up user chats, th…

Isn't it obvious? Web scraping and even scanning written books is at record levels because the data is so useful for training. They are using every byte of user data.

Re: Navier-Stokes – Tristan Buckmaster [pdf]

#690

Post by aurareturn »

rochansinha wrote: post #663

People here do not seem to be considering the second-order effects of these series of events. No academic institution or enterprise will trust OpenAI, Anthropic or any other non-local AI model with their core IP. There will be severe restrictions on what employees at these companies/institutions can share with AI services even from their personal accounts. (Or I am just overthinking it)

It's competition after all. If your academic colleague doesn't care and is leveraging ChatGPT in a big way and is making progress, you'll start to feel the pressure.
Post reply on HN