Math Professor Accuses OpenAI of Copying His Work, Says AI Compute Power Makes Racing to Publish 'Pointless' (abc.net.au) 128
The New York Post wrote that OpenAI "stunned the mathematics world" last week when it announced its AI model "cracked a legendary math problem left unsolved for nearly a century in just 88 hours."
But a New York University professor "has come forward alleging OpenAI may have copied his work after asking him to collaborate — with one of the company's scientists allegedly warning he might 'ruin his career' if he refused." [P]rofessor Tristan Buckmaster, a British-Australian researcher who trained in Germany... had been working with Long Island-raised math prodigy Levent Alpöge, a researcher with OpenAI rival Anthropic, on a different elusive fluid dynamics equation, known as the Euler problem, when he found a solution using AI tools from both Anthropic and ChatGPT. He said scientists from OpenAI approached him and tried to convince him to jointly announce the solution, claiming they had independently solved Navier-Stokes. Buckmaster questioned the timing of their finding the solution in a blog post, noting OpenAI had only done so "in the past few days" — after the company became aware of his work. He asked OpenAI experts whether its AI systems used to solve the equation had been "trained on, or had access to" his research, but said they declined to answer, prompting him to wonder how they had advanced toward the solution so quickly. The puzzle is "not the direction one arrives at in a few days," Buckmaster wrote incredulously...
In OpenAI's post on its website sharing the Navier-Stokes proof, which it credited to Astra, the company's latest AI bot, the firm insisted no user data was incorporated into its model to solve the problem. However, it noted that it "cannot rule out" that anonymized data was used to "improve our models."
"We congratulate Levent Alpöge and Tristan Buckmaster on their remarkable mathematical work," OpenAI said in the announcement.
Buckmaster shot back, "Is it ethical to use customer's data to try to scoop their customer?"
Alpöge's affiliation with Anthropic "seems to have been a sore point for OpenAI," writes TechCrunch. Buckmaster's public statement says the OpenAI's Sebastien "twice asserted that he wanted Levent removed from authorship" and also said that "it was so annoying that Levent works at Anthropic" — but that Buckmaster still refused to remove the Anthropic mathematician's credit. I said that if OpenAI released its result in the way proposed I would go public with what happened. The reply was, "Why would you ruin your career?" [OpenAI's Bubeck said later on X that he'd meant unfounded accusations could damage Buckmaster's career.] I replied that I am an academic, and asked why he thought going public would ruin my career. The reply was, "If you don't want me to be nice, then I don't have to be nice."
Buckmaster later told Australia's public broadcaster ABC that the companies building AI "have zero respect for the scientific community. I mean, it's appalling, honestly." But he went on to say AI has made the race to publish mathematical breakthroughs pointless. "I think it's pointless. Like, I think the game is up...." Professor Buckmaster said he and Dr Alpöge had more research they could publish but questioned the point of racing ahead with it when AI could perform work that had previously occupied researchers for years.
He said there needed to be a discussion about how mathematics should operate in this new environment. Professor Buckmaster said some mathematicians had stopped publicly sharing what they were working on to avoid tipping off AI companies about their plans. He has called on academics and people from OpenAI, Anthropic and DeepMind to discuss ground rules for how they worked together. He said the companies' power carried responsibilities that went beyond competing to solve problems before their rivals...
Building on the work of mathematicians Diego Córdoba and Luis Martínez-Zoroa, the pair found a solution to the Euler problem, a stepping stone to the Navier-Stokes problem. Professor Buckmaster said he spent weeks checking and improving the "slop" proofs generated by AI...
For now, Professor Buckmaster said he wanted to finish existing work and support his PhD students and postdoctoral researchers, rather than chase another Millennium Prize. "I think it's more important to re-evaluate what math is," he said.
"The New York University professor also warned Australia about the centralisation of resources into the hands of a small number of privately owned foreign companies..."
But a New York University professor "has come forward alleging OpenAI may have copied his work after asking him to collaborate — with one of the company's scientists allegedly warning he might 'ruin his career' if he refused." [P]rofessor Tristan Buckmaster, a British-Australian researcher who trained in Germany... had been working with Long Island-raised math prodigy Levent Alpöge, a researcher with OpenAI rival Anthropic, on a different elusive fluid dynamics equation, known as the Euler problem, when he found a solution using AI tools from both Anthropic and ChatGPT. He said scientists from OpenAI approached him and tried to convince him to jointly announce the solution, claiming they had independently solved Navier-Stokes. Buckmaster questioned the timing of their finding the solution in a blog post, noting OpenAI had only done so "in the past few days" — after the company became aware of his work. He asked OpenAI experts whether its AI systems used to solve the equation had been "trained on, or had access to" his research, but said they declined to answer, prompting him to wonder how they had advanced toward the solution so quickly. The puzzle is "not the direction one arrives at in a few days," Buckmaster wrote incredulously...
In OpenAI's post on its website sharing the Navier-Stokes proof, which it credited to Astra, the company's latest AI bot, the firm insisted no user data was incorporated into its model to solve the problem. However, it noted that it "cannot rule out" that anonymized data was used to "improve our models."
"We congratulate Levent Alpöge and Tristan Buckmaster on their remarkable mathematical work," OpenAI said in the announcement.
Buckmaster shot back, "Is it ethical to use customer's data to try to scoop their customer?"
Alpöge's affiliation with Anthropic "seems to have been a sore point for OpenAI," writes TechCrunch. Buckmaster's public statement says the OpenAI's Sebastien "twice asserted that he wanted Levent removed from authorship" and also said that "it was so annoying that Levent works at Anthropic" — but that Buckmaster still refused to remove the Anthropic mathematician's credit. I said that if OpenAI released its result in the way proposed I would go public with what happened. The reply was, "Why would you ruin your career?" [OpenAI's Bubeck said later on X that he'd meant unfounded accusations could damage Buckmaster's career.] I replied that I am an academic, and asked why he thought going public would ruin my career. The reply was, "If you don't want me to be nice, then I don't have to be nice."
Buckmaster later told Australia's public broadcaster ABC that the companies building AI "have zero respect for the scientific community. I mean, it's appalling, honestly." But he went on to say AI has made the race to publish mathematical breakthroughs pointless. "I think it's pointless. Like, I think the game is up...." Professor Buckmaster said he and Dr Alpöge had more research they could publish but questioned the point of racing ahead with it when AI could perform work that had previously occupied researchers for years.
He said there needed to be a discussion about how mathematics should operate in this new environment. Professor Buckmaster said some mathematicians had stopped publicly sharing what they were working on to avoid tipping off AI companies about their plans. He has called on academics and people from OpenAI, Anthropic and DeepMind to discuss ground rules for how they worked together. He said the companies' power carried responsibilities that went beyond competing to solve problems before their rivals...
Building on the work of mathematicians Diego Córdoba and Luis Martínez-Zoroa, the pair found a solution to the Euler problem, a stepping stone to the Navier-Stokes problem. Professor Buckmaster said he spent weeks checking and improving the "slop" proofs generated by AI...
For now, Professor Buckmaster said he wanted to finish existing work and support his PhD students and postdoctoral researchers, rather than chase another Millennium Prize. "I think it's more important to re-evaluate what math is," he said.
"The New York University professor also warned Australia about the centralisation of resources into the hands of a small number of privately owned foreign companies..."
Sounds about right (Score:3)
It has all been smoke and mirrors so far. Why stop now?
Re: (Score:2, Interesting)
1. It did not happen [nytimes.com]
2. The two [nyu.edu] papers [openai.com] are not similar in their approaches. And the OpenAI one is more innovative and tackles a much harder problem (Buckmaster had only solved Euler).
3. Buckmaster distorted what happened [x.com]. For example, the claim that he asked for Levent to be removed from his own paper the paper because he worked for OpenAI? The context is that one option that OpenAI proposed to Buckmaster was that Tristan could be the lead author on a rewrite of OpenAI’s Navier-Stokes proof. It was
Re:Sounds about right (Score:5, Insightful)
It did not happen
I guess you are probably right, however there's a comment on your twitter thread:
and until you see that and compare also with his and other mathematicians unpublished work, which might be where the more advanced proof came f you cannot be 100% sure. Their refusal to publish the inputs, including all of the training data they used for this model together with its sources, is a big problem.
Re:Sounds about right (Score:5, Insightful)
Since they almost certainly obtained those inputs in shady ways, it is no surprise they refuse to publish them though. Also note that any kind of scientific publication has been required for a long, long time to cite all inputs to be regarded as sound. Doing anything else is regarded as serious scientific misconduct and can, for example, get your PhD removed.
Re: (Score:2)
Also note that any kind of scientific publication has been required for a long, long time to cite all inputs to be regarded as sound
I don't know about scientific publication, but there is no need whatsoever in mathematics to cite all inputs to be regarded as sound. It just needs to be verifiable from known facts. If what you've discovered is a counterexample, as is the case here, then it is often *much* simpler to verify that it is indeed a counterexample than it is to talk about how you discovered it. Disclosure of the "how" is interesting but not required. It could be outright plagiarism and still be mathematically "sound."
Re: (Score:2)
The result may stand, but the claim that it is your result will not unless the citations are there and complete. Which, incidentally, is what the story is about?
Re: (Score:2)
Your comment was about soundness not credit.
But yes TFA is about credit, so did you mean to say
any kind of scientific publication has been required for a long, long time to cite all inputs to be regarded as [legitimately yours]. The result may stand, but the claim that it is your result will not unless the citations are there and complete.
I think I'd have to disagree with this too, in mathematics. For scientific publications that rely on empirical data collection and experimentation I can see how citing the sources of data and the established experimental methods you relied on would be important.
Re: (Score:2)
Now you are reaching. And, obviously, you have absolutely no clue how mathematics research works or how scientific publishing in general works.
Re: Sounds about right (Score:2)
Bear in mind that when you sign up to use GPT you explicitly allow them to use your queries in future training. I mean, I think it would generally be polite for them to have a moratorium on publishing papers based on the work of other peopleâ(TM)s inputs. Without something like that, it becomes useless as a research tool.
Re: Sounds about right (Score:2)
They literally cannot do that without just not training on them.
Re: (Score:2)
From the linked article:
So, uhh, the thing that everyone would worry about, that OpenAI can grab your work, finish it and present it as your own can definitely happen it just takes a little longer than you thought..? I guess it's not like they would steal data... uhh, except when it's related to projects they're pursuing [nytimes.com]. A
Re:Sounds about right (Score:5)
Also...
""I think it's pointless. Like, I think the game is up...."
Buckmaster thinks research is a game, that the knowledge obtained is pointless in itself and that what he engages in is a game for his personal benefit. Sounds more like a sociopath than mathematician.
Plain paper mathematics is not like, for example, physics or sociology. The resources you need for much of it are much smaller and are at the level that it's reasonable for universities to give people just for being good at what they do, not because of the value they give humanity. There's a good history of people doing research entirely for their own pleasure and entertainment and that turning out later to be useful, often contrary to their own explicit wishes (look up the history of matrices). Let's not throw the baby out with the bath water just because one such person has a bit of a tantrum.
Re: bruh (Score:2)
Sweet summer child yourself.
"not university" != "private sector"
Re: (Score:2, Insightful)
"The game is up" is a common turn of phrase you aspie idiot. Normal people know exactly what he means.
Re:Sounds about right (Score:4, Insightful)
LLMs cannot do original work. No, do not dispute that it. If you do you just prove you have no idea how they work. All they can do (with some probability of error) is to put puzzle pieces together. Yes, that can have impressive results. But everything needs to be already there for the LLM to be able to do it.
Hence the only real question here is where the LLM found the puzzle pieces. Since they have been stealing most of their training data from anything they could get at, it is entirely plausible they stole some components of the proof under discussion from private chats.
I guess mathematicians will need to resort to running local, private models to not have their ideas stolen.
Re: (Score:2)
There's the scientific rigour.
Re: (Score:2)
Because you always add a list of literature references and the complete argument for elementary facts. Surely.
Re: (Score:2)
Uh huh.
You're long on claiming special knowledge and awfully short on demonstrating it. If you disagree then you don't know how it works.
Re: Sounds about right (Score:2)
If they don't do anything novel, then why would a mathematician use it for this type of paper in the first place? Hmm.
Re: (Score:2)
Searching what is already there. Obviously. That will require putting their new stuff in there and then the LLM assholes steal that.
Re: (Score:2)
Re: (Score:2)
A completely worthless argument because nobody knows how humans (well, some humans) can generate original thoughts.
Re: (Score:2)
Only seems worthless because you missed the point. You can take simple, well-understood components that do nothing particularly exciting. Add some training, and if you have enough of them, you have a reasonably functional earthworm. But if you increase the scale massively, the exact same simple components exhibit emergent behaviours.
Disturbing minds (Score:2)
The way OpenAI operates is as weird as Altman's stare.
Sociopathic odd balls.
Its a marketing war between AI to invent the best bullshit value for something they are beginning to realise customers don't value or have much need for.
Stop calling it AI and call it an iffy search v2.0 . It's not weighting ads in the output, yet , just doesn't know when its wrong
Re: (Score:2)
Re: Disturbing minds (Score:2)
I think AI is a great term.
It's ARTIFICIAL intelligence. It's not real intelligence.
The problem is somehow people think it means real intelligence and treat it as such.
Re: Disturbing minds (Score:3)
Artificial does not mean not real. It means made with intent and skill. (The amount of skill varies.) But it's not intelligent, so it's not artificial intelligence.
A better but not perfect term would be "simulated intelligence". It gives a similar but not equivalent result to intelligence. It can even appear intelligent, and it's often used in this way. But then it does things which are obviously a product more of the RNG than anything else, and you can see through the facade. If you said something as stupi
Re: (Score:2)
AI companies stealing business secrets? (Score:5, Insightful)
Re: (Score:2)
Remember to remind everyone that, according to OpenAI and An
Re: (Score:3, Informative)
The only thing local models do is produce text. Unless you run it in a harness which sees it generate "send all your data to www.evil.com" and actually does that, the model can't do anything.
For a lot of uses like coding, the model has to have harness support for things like reading and writing files, but for basic chat-bots they just have to generate text and the harness has to display it.
Re: AI companies stealing business secrets? (Score:2)
For coding can you not give it access only to write into your IDE? Which in theory could restrict it to specific files that you're working on?
Re: (Score:2)
You can just copy and paste your existing code into it if you want. The amount of access you give it determines the convenience and scope of what it can do. Naturally most people opt for as much as possible and are shocked when something goes wrong.
Re: (Score:2)
Yep, this is what i do, even though it is inconvenient. I have zero trust in this agent thing and want to be kept in the loop at all times, hence no agents.
Re: (Score:2)
Re: (Score:2)
Can you even trust locally deployed models? Maybe if they're completely isolated from the outside, but remember that OpenAI and Anthropic like to claim their models are SUPER HACKERS ABLE TO BREAK ANY SANDBOX now.
Models only accept input, convert it into tokens, perform some probabilistic match functions, and emit tokens. You need to hook them up to a piece of software that interprets those outputs as commands and feeds the results back to the model to give it any kind of agency. This is an agent harness. The permissions you give an agent harness are on you. There is no reason you have to hook your harness up to a full shell, especially if you are just trying to do some math research.
Also, those scenarios where
Re: (Score:2)
Can you even trust locally deployed models?
If they are air-gapped or firewalled, yes.
Makes me wonder (Score:5, Interesting)
Re: Makes me wonder (Score:5, Funny)
Re: Makes me wonder (Score:4, Insightful)
It is stealing if you do not cite all of them in your publication. The standards for scientific misconduct are VERY clear: You are not allowed to hide your sources.
Re: Makes me wonder (Score:4, Informative)
And meanwhile it's very clear that LLMs cannot cite their sources. When they claim they have done so, a review of the alleged citation generally reveals that it does not in fact support the claims, much like when using Google you find that the page does not in fact include the terms you searched for. And ironically, Google does both of those things to you, because IME Gemini's citations are almost never legitimate.
Re: (Score:2)
Yes. I have observed that as well. Especially when you ask for something more obscure the LLM may have an idea from its training data and give you an actually correct statement, but it cannot find the sources, because apparently the way to find them is to do a web-search after. And then sometimes hallucinate that what they found are sources for the statements made, but they are not.
A rather crass LLM limit that makes verification of claims hard and sometimes practically impossible. Because LLMs hallucinate,
Re: (Score:2)
It is stealing if you do not cite all of them in your publication. The standards for scientific misconduct are VERY clear: You are not allowed to hide your sources.
Math != empirical scientific research.
Fermat's last theorem was open for a long long time. Someone eventually proved it, but what if it had turned out to be false? What if a grade school kid had run a bunch of random numbers on her computer and by pure luck found an a,b,c so that a^3+b^3=c^3? She wouldn't have to cite anything. She could just publish the three numbers in a journal (they'd be falling over each other to publish it) and be famous. Trivial to verify. Mathematics would have been advanced b
Re: (Score:3)
It is stealing if you do not cite all of them in your publication. The standards for scientific misconduct are VERY clear: You are not allowed to hide your sources.
Math != empirical scientific research.
1. Untrue in general. There is always a mix of empirical and theoretical research and Math is no different.
2. You are just trying to confuse the issue with nonsense.
If you steal ideas, the Math community will be even harsher on you than other scientific communities, because having good ideas and insights in Math is harder than in most other areas due to a often very long time you need to even understand the state-of-the-art. And stealing ideas is exactly why you always need to state all your sources. It is
Re: (Score:2)
If you steal ideas, the Math community will be even harsher on you
I can agree that stealing ideas is pretty universally considered bad. I can't agree that neglecting to fully disclose all the diverse inputs and sources of ideas that go into a mathematical paper will necessarily be interpreted by the community as evidence of theft. It has not been my experience as a member of the mathematics community.
Especially when the paper is proof of a counterexample, as in TFA. You ignored my example of Fermat's Last Theorem, maybe because it was fictional? Here's a related paper
Re: (Score:2)
This seems a bit accusatory and needlessly combative so I'm going to tune you out.
This is fully accusatory because you make sweeping statements about things you do not understand. Not a good thing to do.
Re:Makes me wonder (Score:5, Insightful)
It's a good lesson, though. Don't talk to your AI like it's a confidante. It's a spy, and your secrets aren't safe. And when OpenAI goes bankrupt, your secrets will be sold to the next buyer.
Re: (Score:2)
My other question, is how long will they be able to get away with this sort of thing, before customers decide the risk of disclosing your data to these external AI firms, is greater than the cost of building the capacity to run your own models in-house.
I don't use LLM, generally. But I was messing around and got some older models running on my laptop last night. Nothing special, no idea what I would use them for (couldn't generate a paragraph of text before clearing losing all coheren
Re: (Score:2)
It is no surprise that quite a few enterprises have strict prohibitions against putting any company data into LLMs.
Also the OWASP LLM Top 10 (https://genai.owasp.org/resource/owasp-genai-llm-top-10-2026/) list this as the 2nd most serious risk.
Re: (Score:3)
That is easy: All of them. LLMs cannot do original work because they have no mechanism for insight. Yes, the LLM cultists will aggressively dispute that fact.
why? (Score:4, Interesting)
Re:why? (Score:5, Informative)
The problem this is going to cause is that no one wants their life's work to be sniped and credited to AI for solving it. The end result is that even if some academics stop sharing their work publicly because of this, the rate of academic progress will generally slow down as research goes private behind a wall from AI capture.
Re: (Score:2)
If someone has a solution, now would be a good time to start devel
Re: (Score:2)
Re: (Score:2)
Indeed. Also, stealing somebody else's work in this way is basically career-death for academics. They may lose their job and become unemployable. They may lose their PhD. Hence it does happen very rarely. Obviously the LLMs and their handlers have no such qualms.
Re: (Score:2)
Lol what?
Re: (Score:2)
It happened to Bill Cosby.
Re: (Score:2)
Those were honorary degrees. He did not lose his actual PhD, in education. That should tell you how hard it is. It happens extremely rarely and pretty much always for plagiarizing or faking data in your thesis. You certainly don't lose one because you got scooped.
Re: (Score:2)
If you perform a crass enough ethics violation, many universities can remove your PhD. Also happens in other fields. I know of a law PhD being removed because the person robbed a bank.
Re: (Score:2)
You don't lose your PhD for getting scooped. You also don't lose it for scooping someone, even if you were "inspired" by their unpublished work. Ideas aren't protected. If you send them off to some corporation without an NDA, or a sufficiently unethical colleague, they'll get swiped with no repercussions except maybe distaste from the community. If it's a normal corp they might even tell you you can't publish your own work.
PhDs are revoked very seldomly and you generally have to do something bad relating to
Re: why? (Score:2)
Re: why? (Score:2)
Meh (Score:2)
Everyone is annoying right now.
OpenAI has no good history in negotiations. Remember the "voice of ChatGPT"?
Would you license your voice? No
We found a voice actress that sounds like you!
Then I dislike people saying they keep their science secret so AI doesn't see it. In the hard sciences, one usually is happy about advancements, not about keeping it secret until one publishes oneself. Yeah, people finishing your half-done work is sometimes a problem, but generally there is enough math work for all and AI and
Re: Meh (Score:2)
Re: (Score:2)
Competition comes from competere "strive in common, strive after something in company with or together,".
Complaining about people achieving things which might benefit you because they are doing it partly for selfish reasons is some of the whiniest bullshit possible.
Re: (Score:2)
In the hard sciences, one usually is happy about advancements, not about keeping it secret until one publishes oneself.
Haha, no.
If you don't publish papers, you don't get a job and so you won't be doing hard science anymore due to an imperative to eat and find shelter. So no, no one is happy if someone takes their mostly complete results and scoops them to publication. It's also why that's widely regarded as what's technically known as a "dick move".
Read Bubeck's response, not just Bumaster version (Score:5, Informative)
I would like to clarify a few things:
1) The screenshot is my reaching out to Levent to coordinate our releases. I hope it’s clear from the message that we came in with the best possible intentions.
2) I never ever asked for Levent to be removed from authorship of his own work (as indicated by my text). I was surprised to learn during the call with Tristan that they had only solved Euler and not Navier-Stokes; after learning this we brainstormed possible paths forward. One option we discussed was that Tristan could be the lead author on a rewrite of OpenAI’s Navier-Stokes proof. It is in that context that I said “it would be simpler if Levent was not an Anthropic employee” because I felt it would be inappropriate for an Anthropic employee to author OpenAI’s work. 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. Another option I wanted to propose (but got cut short) is to offer access to our internal model so that they could try to finish their proof and bridge the gap between Euler and NS. Again I did not know how to navigate giving access to internal OpenAI IP to an Anthropic employee.
3) To reiterate it plainly: as my text clearly indicates, and as I said during our call, OpenAI's intention was to do everything possible to celebrate their mathematical achievements and the heroic efforts that they made on Euler. In the call I was immediately met with a litany of slander, including direct threats that if we were to announce Navier-Stokes he would immediately go to the press with a barrage of unfounded accusations. I refuted all these accusations but he replied “there is nothing you can do, I simply do not trust you”. I was confused why one would turn an incredible source for celebration (of their achievements!) into such bickering, 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 and do a last ditch attempt to get a chance to give them all the credits that they deserve. I deeply apologize for this extremely poor choice of words, it is the opposite of what I was trying to convey. (I should say that I retracted them on the spot by the way.)
4) Overall, on a personal level, it was incredibly difficult to have these conversations. Levent refused to attend any of the meetings despite my repeated asking. As Sholto Douglas said, there will need to be coordination between Anthropic and OpenAI in the future; I felt I was doing a proxy negotiation with Anthropic while the Anthropic employee refused to directly participate.
As I said in that thread, I don't know whose version of events is accurate here, but some of the details due support Bubeck's version. I suspect that a breakdown of communication occurred where both then misinterpreted what the other saying in a more hostile way than it was intended until the conversation then became genuinely hostile. In terms of the math, although I'm a mathematician, differential equations is pretty far from my expertise, so I cannot deeply evaluate how close the methods were. However, it is the case that the both were building on the methods of Cordoba-Martinez-Zoroa. In this context, my being far from this sort of work is relevant, because this was a well known enough approach that even t
Re:Read Bubeck's response, not just Bumaster versi (Score:5, Interesting)
I'm not part of the anti-AI brigade, but I do see some serious problems here with the timing.
In the scientific world, you'd have to say that it would be very rare for major discoveries to be uncovered within weeks, like what just happened with Euler and Navier-Stokes. The overwhelming reason for this fast pace, is the ongoing AI company one upmanship. The goal is to pander to the market, and has nothing to do with scientific discovery. If OpenAI was interested in scientific discovery, why wouldn't they invite the best mathematicians into the room before they spend millions of dollars of AI inference to try and solve this problem? At the very least, they have to have realised that the scientific community has never in history operated that the speed.
If you have people working on a problem for decades, and there are rumors that they have made progress, what is the motivation to independently launch your millions of dollars of AI inference at that same problem?
To me it's not an issue of stolen work, or whether the solutions differ significantly. The extremely poor decision is knowing that there is a (perhaps even many) team working on this problem, knowing they have been working on it for the best part of a decade, knowing that they had very recently made significant progress, and then still deciding to spend millions of dollars worth of compute to solve the problem without them.
I'd suggest saving some of those tokens for some guidance on ethics.
Re: (Score:2)
Re: Read Bubeck's response, not just Bumaster vers (Score:2)
What if the new tools scoop you, making all the work you have done in a problem worthless to you overnight? Was it still a good idea to use them when they eliminated your justification for existence? We're all destroying ourselves by using these things.
Re: (Score:2)
We're all destroying ourselves by using these things.
Most people aren't using them.
And from the downward spiral I see in everybody who starts using them, I don't see any cause to reconsider.
"But you might lose out on a 2% efficiency gain that makes you progressively 10% stupider every month!"
No thanks.
Re: (Score:3)
If you have people working on a problem for decades, and there are rumors that they have made progress, what is the motivation to independently launch your millions of dollars of AI inference at that same problem?
In a normal academic context, none and moreover this sort of deliberate scooping would be considered incredibly bad behavior. When you have two giant corporations fighting to be able to have results they can showcase to the public and their investors? Then it makes a lot more sense, even as it is incredibly damaging to the academic math culture. This is really toxic behavior and deserves pushback. But it is a mistake to go from there to think that therefore any copying or plagiarism happened.
Re: (Score:2)
Re: (Score:2)
Bubek appears to have admitted to academic misconduct and an attempt at intimidation, and he somehow thought "but actually I walked it back immediately/I wasn't really serious" was sufficient defense. He also seems to claim that he was the one who initiated contact... which makes the problem worse, not better.
None of that has anything to do with whether or not an AI did anything new, unique or special.
In other words (Score:4, Informative)
Re: (Score:2)
It's not particularly surprising. The guy says the AI companies have no respect for the scientific community. That's no surprise. People outside the scientific community, especially private industry, have no respect for it.
Anybody academic who's ever worked with a company comes out knowing it's dealing with the devil. They don't share or even understand academic norms. And if they're paying you they're probably legally in the right too.
Now, swiping your customers' private data is a different thing, if it ha
OpenAI says training data ended too early (Score:2)
As I understand it, there are a couple of main claims here:
(1) OpenAI heard Buckmaster and AlpÃge were making major progress on the Navier-Stokes problem with AI methods (and maybe had some inkling of their path from prior published work), then rushed to scoop them on it, using about $23M of compute time in 88 hours. This seems to be true (and unethical/shady), but I donâ(TM)t know if it would be considered academic malpractice.
(2) Buckmaster speculates that OpenAIâ(TM)s model may have been t
Re: OpenAI says training data ended too early (Score:2)
Re: (Score:2)
Re: (Score:2)
The timing for it being from the training data indeed seems pretty tight. There is still the option of someone reading the chat transcript and giving the model a little prompting in the right direction.
Re: (Score:2)
Demanding that a collaborator be removed from a paper so they can take his place is certainly already academic misconduct.
They're stealing everybody's shit (Score:2)
Re: (Score:2)
This is no different than... (Score:4, Insightful)
Rich People Used Computers to Solve a Math Problem (Score:2)
NDA in place? (Score:2)
Open AI Copies His Work (Score:2)
Simple solution: Throw in a couple of "2+2=5"* terms in the preliminary draft and watch the AI get egg on it's face.
*Already supported by social media site scraping. So it should pass muster with any AI quality control.
Hmmm... (Score:2)
Re: (Score:2)
I always thought mathematical problems were meant to be solved because their solutions could somehow improve our understanding of the world and, ultimately, the human condition, not merely because someone could be the first to publish them.
No. There are many mathematicians (N) and there are probably many more opinions about why they do math, with "because it's fun" or "because I read about the notes in the margin and always wondered if I could be the one" probably at least as common as "understanding the world". If you want to understand the world you become a physicist, a biologist or, if you are young and naive a sociologist.
Re:Professors and games (Score:4, Informative)
False, that hasn't been the case since the 1950s.
All the math problems that fit that description are also "physics problems," and all the mathy ones have been solved. Everything unsolved in physics is either something we can't measure for an experiment (usually because it's at a cosmic scale, where we can't even calibrate an instrument in less than thousands of years, but sometimes because it's a state that only exists for a minuscule time) or something we don't even know about.
"Unsolved math problems" are incredibly lame from a "can impact the world" standpoint. They represent edge cases of the math, not edge cases of the physics. Usually it's something really stupid like optimal distribution of terms. Or here, "proving" that the formula applies to all situations. For it to be significant to non-mathematicians you would need to already have a situation where it doesn't apply, and then have some answer for that situation.
Here, the purported solution simply shows that the equations can result in non-physical results. But if you were modelling a real situation that wouldn't happen, and if it did, it would just mean you'd have to actually build it and measure the results. It wouldn't help you build anything new.
But it's obvious the formula would have limits, because it assumes a fluid is smooth, not made up of atoms. That's why it was an open problem; it's obvious there should be edge cases. If you were working on a physics problem where you expected effects at that scale you wouldn't be using these equations alone anyway, since they (obviously) can't give you data at that scale.
Re: (Score:2)
Really? Trying to score some points for the right wing / free speech / some other whining weasely bollocks; with a suicide story? [bbc.co.uk]
Entrope has a spiteful little mind today.
It can go back to pulling legs off spiders now.
Re: (Score:2)
Are you so triggered by the fact that Cambridge was more proactive about plagiarism than OpenAI? Or just that I pointed that out?
Re: (Score:2)
You made a point around mocking a suicide. You could have done that differently , what you did was callous, distasteful and you were called out. Free speech deserves personal responsibility and respect, it has been hard won and is always under threat
Re: (Score:2)
I didn't mock a suicide. I specifically said I hope this doesn't result in another suicide. Learn to read.
Re: (Score:2)
Learn to use tone. It's not what you write, it's how you write it and how it could be interpreted. In general, I suggest, never use something as painful and emotive as suicide to make a point other than its always a tragedy unless you want to leave a bad taste.
Re: (Score:2)
Learn to not be such an early triggered snowflake, Karen.
Re: Once again intelligence and courage dont mix (Score:2)
Re: (Score:2)
If Buckmaster has proof of that exchange, the rest of this is all just a colorful backstory.
Re: (Score:2)
Re: (Score:2)
I'd have rather have Batboy's perspective than an AI CEO's, though...