OpenAI has solved a Millenium Prize math problem (Navier-Stokes). It will be published tomorrow, but there is currently a major Anthropic-led shilling operation trying to discredit them.
Would the world end, if they stopped stroking Navier?
>>109759117usecase for this knowledge?
>>109759117>finite time blowup for the forced Navier- Stokes equationCock sucking op the millennium one asked for tge unforced one
Can someone help me build skepticism around this sort of claim? I really really don't want to believe...
And just like that... all your livelihoods.... gone... woah...
>>109759117>I have not seen it
>screenshot thread>of text>in notes app
>>109759117r/AntiAI is going to seethe really hard at this…
>>109759436https://cims.nyu.edu/~tristanb/statement.pdfBottom of page 2.
>>109759162It's probably so complex only a handful of people on Earth can verify it, which means GPT can force it's usual "confident retard" slop on less competent researches and gaslight them into believing it's true
>>109759145Hundreds of YouTube videos explaining the meaning.
>>109759117Ai is the future
Make sure to keep this thread alive till tomorrow.
>>109759117>OpenAI has solved a Millenium Prize math problem (Navier-Stokes). It will be published tomorrow, but there is currently a major Anthropic-led shilling operation trying to discredit them.so much gay and fake in one post, just like an LLMs output really.
>>109759133If this proof is solid, we can guarantee that Navier's strokes finish in constant time.
>>109759145>>109759478>>109759650It’s going to be fake news and mathematical evidence. You can’t mathematically explain randomness in math, water is one such factor which always had randomness applied. We humans only contain it on a path, never controlled it.
>>109759162the problem with the "proofs" so far is that they were counter examples. There is one formulation effort done by claude I think but of a existing proof. Finding a counter example or example in general is good however its meaningless if you cannot derive solutions from it.
>>109759117>stole the work of a pair of researchers>conspicuously avoids saying that they didn’t read his chats>tries to get him to drop the coauthor that works at anthropic>once that failed tries to threaten himWhat a master class in a PR disaster. What next, kidnap his kids?This also adds fuel to every single voice saying that they totally snoop and train on your chats even when they say they don’t, which is even worse than them being total dicks to some random researcher who dindu nuffin.And all that in pursuit of good pr for Astra, lol, lmao
>>109759117This is luddites 9/11, the seething tomorrow is going to be glorious
>>109760363Again nothing can explain randomness in science. You’re talking about concepts that are beyond our moral reach.
>>109760408But not beyond an AI
>>109759117why can't it then fucking make a free open source version of adobe and office products/other productivity apps for linux? or games like minecraft, terraria, cod zombies or portal? or long awaited sequels to existing games? or a discord alternative? why's everything still a slow ass ram hogging bloated add filled monetised chromium webapp? if anything, software getting much worse. honestly this pseudo religious proselytising of big tech is getting annoying, buy an ad>>109759537>"confident retard" slopliterally this
>>109760757It's incredible to go from reading Terence Tao and Timothy Gowers talk about the impact AI is having on mathematics to seeing all these Redditors who don't even understand calculus, let alone neural networks, go "Umm, you do know that SCIENTIFICALLY speaking LLMs are just sochastic parrots, right? They think 1.11 is larger than 1.9!"It goes without saying that it is entirely feasible for a model to be able to write difficult mathematical proofs without also being able to faithfully recreate the design choices, voice acting, and 3d modelling from the beloved "cod zombies" video game series, especially when the former is much easier to train through reinforcement learning and has had many billions more dollars thrown at it. It also goes without saying that LLMs all write their proofs in Lean or other proof assistants (else, there would be no reward signal to train on).I appreciate that if you're the kind of person who mainly concerns yourself with video games and complaints that "software getting much worse", you may find these things difficult to understand, and may become frustrated seeing discussions you're unable to participate in--maybe mathematics professors are all lying "globalists" propping up the "AI bubble"?--but this is your own psychological problem. It's disrespectful for insignificant individuals to intrude on the affairs of their intellectual superiors, especially in such forceful and obnoxious ways.For anyone interested, here are Tao's comments on recent events:https://mathstodon.xyz/@tao/117233527638291447And here's the statement from Tristan Buckmaster people are referencing:https://cims.nyu.edu/~tristanb/statement.pdfAnd here are Buckmaster's three recent papers with Levent Alpöge referenced in the statement:https://cims.nyu.edu/~tristanb/euler.pdfhttps://cims.nyu.edu/~tristanb/ipm.pdfhttps://cims.nyu.edu/~tristanb/boussinesq.pdfThe entire Lean source code is on Github:https://github.com/tristanbuckmaster/fluid_lean
>>109760933Whole long-ass post that says nothing lol.
>>109760933You are confused; a stochastic parrot can still be useful for many things, but it doesn't change the fact that the architecture of these machines still hasn't fundamentally changed - they still legitimately are stochastic parrots. AGI is a myth and these companies should probably lose everything because of their bullshit, but the technology that they currently have produced can have some genuine usecases. Doesn't mean that these models aren't overhyped to shit.
>>109759973>>109760408Never taken a mechanics class award
>>109759160It's been like that with every single claim so far. One of the questions was 'what is the upper bound for X' and the AI gave a counterexample for the currently known lowerbound and it was shilled at le breakthrough even though it had nothing to do with the problem.
>>109760287It's worse than that because in several cases, the counterexample found was so trivial that a bruteforce enumeration would have found it within 30 minutes. The fact we didn't have that counterexample before just shows that it was something nobody cared about, ever.Examples and counterexamples can be groundbreaking. Constructionist proofs are real things. Imagine if an LLM gave us a counterexample of P = NP. Totally something that one could think of to eventually find (LLM or no). But this and that are different things.
>>109760408Randomness is an averaging scheme to approximate what we don't understand. You are retarded.
>>109760933Why trust terence tao on anything not in math? Does terence use AI in his work?I have a PhD from yoshua bengio's lab on LLMs. I have patents on key technologies in dialogue systems, particularly conversational context, memory and action systems similar in spirit, but much more powerful, to the tool calling system that companies have implemented today. You don't know a single thing about anything.How about you shut the fuck up for a start.
>>109760933>complaints that "software getting much worse"A claim you mock with zero capacity to substantively refute.
>>109759117>>109759160>>109759478>>109760933Sam Altman will be the first tech CEO to livestream his suicide.
>>109759117And yet it cant generate simple images without shitring itself.Hmmmmmm
>>109759537Mathematical proofs by AI are verified by computer systems, mostly LEAN. Verification is built-in with these systems.>>109760287As far as proofs go, a counterexample is as powerful as a proof of correctness. Some of the most famous mathematical works have been counterexamples. Not sure what you mean by "deriving solutions from it"
>>109761191I think he means that the process is often more important than the result irself, especially when open problems are concerned. Getting an LLM to bruteforce its way to an obscure counterexample often doesn't give as much mathematical insight as you think, especially when you have no clue how the LLM got there in the first place. This is a point even mathematicians supportive of LLMs like Terence Tao makes.
>>109759117Who cares about stupid ass math no one gives a fuk about. I just want to make big booty Latinas.
>>109761191You need to verify if the proof itself is written correctly. There have been many cases by mathematicians submitting incorrect proof of their correct or incorrect conjectures for example.
>>109761278Yes that's a valid discussion to be had I have my own thoughts on it but this is maybe not the best place. Anon made it sound like counter-examples are less useful because then the theorem cannot be used for technical applications or other proofs which is silly.>>109761312Yes, but you cannot make such mistakes using formal computer proof systems. You get a "compiler error" so to speak when you try to write an invalid proposition. So reasoning mistakes are impossible. Once a proof is written down in LEAN, it's (basically) guaranteed to be correct ab-initio.
>>109761345Did you even read? Invalid lean proofs have already historically been submitted aplenty. They don't give "compiler errors". They pass. They say the proof is good. It's not, because the program was not written correctly. Get a brain.
>>109759117I have the feeling constantly doubting OpenAI claims and asking for proofs for every single thing is rooted in thinly veiled antisemitism.The model just proved it okay, would it really kill you to take their word for it?
>>109761345>Anon made it sound like counter-examples are less useful because then the theorem cannot be used for technical applications or other proofs which is silly.Depending on the exact specifics of the conjecture an LLM simply shitting out a counterexample with no clear reasoning can leave you with very little more than a mathematical mess. I don't think it would be wrong to say that some open questions are viewed to be more important than they are because of the expectation of what would actually be needed to get there.
>>109761112>Does terence use AI in his work?Yes, extensively. For the past few years much of Tao's work has centred on the role of proof assistants in mathematics, transitioning neatly into the rise of AI. He's one of the most knowledgeable people on the applications of AI to mathematics, as well as being probably the world's most eminent mathematician in general. Its cool if you really did work at Bengio's lab, but I think if you had created technology "similar in spirit, but much more powerful" than companies with multi-hundred billion market caps, you'd do more than just post about it on 4chan.
>>109761418Counterexamples are the wrong fight here. You want a positive non-constructionist proof. That's a very different matter. Anything that can be disproved by examples is not of that class where a process would be useful to capture to advance math research.
>>109761425It's cute how clueless and retarded you are, but you really shouldn't be spending your time posting this garbage and instead try information yourself.
>>109759145>>109759160>>109759537>>109759638>>109759973usecase for these copes
As per usual it just found an existing solution or hallucinated a wrong one or both.Yawn
Yugi never cheated btw
>>109761357>aplentyno they are extremely rare. The only reason why a LEAN proof could be incorrect is due to bugs in the software itself (like premature optimisations). But the theoretical process these formal verification systems go through is exact. There's no way for researchers to try and game the system or fake a proof. Once a proof gets ran through the interpreter it is objectively correct. (there are a few very niche exceptions though if you try poking holes in the very foundational axioms at the top of the proof tree; and some mathematical constructions are uncheckable in finite time by design; all pretty interesting stuff we can talk more about that if you want)>>109761418Sometimes the construction of a counter-example is so elaborate it requires new mathematical insight, in the same way a positive proof does. I don't think it makes sense to draw a line in the sand like that. I come more from a physics background but violation of the Bell inequalities for example from both a theoretical and experimental POV required quite a bit of new physics (for which the researchers got a Noble prize).The more interesting argument to make here is that when AI comes up with a positive proof, or a counterexample, it might not be possible to extract mathematical insight from the model's thought process. I don't think that's exclusive to counterexamples, even if those are easier to look for using AI.
>>109759537Thankfully modern AIs can quickly review and spot errors in a 100 pages proof so we don't need to wait around for useless mathematicians to prompt an LLM to explain the paper to them (which is what mathematicians actually do nowadays.)
>>109761430I was about to make a reply but this anon >>109761583 did it better and seems to fit better as a response to you than to me.
>>109761583>if I deny reality, I winJust ask chatgpt since that's the only thing you know how to do with your 65 IQ.
>>109761583>it might not be possible to extract mathematical insight from the model's thought processThat's not just a "might", it's reality. Even if the reasoning traces could be extracted and intelligible it's unlikely these AI companies would grant open access to them.
>>109761653Yes my language was too vague maybe I agree. Not to mention that reasoning traces are not even the "real" thinking of the model, those would be the weight activations of the model, which of course, are mostly uninterpretable. I used to talk a bit to a researcher in AI ethics and yeah the field is in its infancy it's like those neurosurgeon types that used to shock rat and fly brains seeing what lights up.
>>109761653>>109761696Not to mention the politics of it all, these models are private IP; lended out to whoever pays the most (and whatever governments and three letter agencies feel they are entitled to).
OpenAI bros, what's this?https://openai.com/index/navier-stokes-solution/>we cannot rule out that de-identified data derived from their usage of our products helped improve our models
>>109761964Who cares? Science and progress beats human ego.
>>109761964that seems to be the endgame, ram everything with AI powered surveillance and keep the realtime data flowing 24/7
>>109761964it's over :(
lol
>>109762088This two idiots are the official KWABOTY 2026
not even kidding, next is p=np btwhold on 2 ur butts
The part in the paper where they say they only started their cluster after hearing from a rumor another team was making progress and got the answer in 88 hours is crazy. Really sad stuff
>>109762088>I was told the model did not look up user dataYIKES the non-answer. Peak OpenAI PR team
>>109762148>"""rumour"""or they looked at the data and saw that one of the researchers worked at anthropic
>>109762174yes of course that's more likely scenario.I don't think AI "trained" their new model on the interim outputs of the researchers the timeline does not line up it takes weeks / months for data to be cleaned and fed through a post-training run I imagine. If anything they probably stole the message transcripts and gave them as hints to the cluster directly. No way to know though. And the proofs seem to be different that's what they claim at least
>>109761964It baffles me that even supposedly smart people still continue to willingly upload their private, proprietary data to third party online services and expect it to remain confidential. I guess all it takes is for an AI chatbot to say "trust me bro".
>>109759117> i was told by> credible experts sayWow never heard this one before
>>109759117predict the context window
>>109760363Reminder soijakwiki will irrelevantly bring up luddites itt>Call anyone who codes manually a snailcat and/or a luddite. Claim that you will be left behind if you don't vibe code. Shill against AI.
>100 page proofImagine the slop.Never heard of a mathematical proof longer than a few pages.
>>109759145It disproves something, so probably has no real effect. AI is good at finding counter examples
>>109762131As usual this was again not a solution to navier-stokes but several levels removed from the actual question.That said, it is definitely a good thing that there is never-before-seen funding and urgency for people to look at these mathematical problems, some of which (like p=np but not like n-s) have fuckhuge massive practical applications and could reshape society.
>>109762401https://github.com/openai/NavierStokesAndEuler tho
>>109762585none of the millennium prize problems have much practical applications. Even p=np is a largely theoretical question.It really is just marketing for these companies. If they cared, they would have written up a decent paper instead of a blog post
>>109762723>finite-time blowupTranslation for the techcels in this board: USELESS for day-to-day fluid dynamics.
>>109762778It's a secondary effect. If p=np there will be a golden age of algorithmic research. Cryptographic algorithms will need to be updated in many parts. The first p-time algorithm that solves an np-hard problem will basically avalanche down everything because we've become very good at reduction. If p != np not much will change.
>>109761583You seem to be missing the point that anon was making entirely. He's not saying the proofs were syntactically correct, obviously then they wouldn't compile, as you said. He means they were semantically incorrect, which can't be automatically verified.
>>109762887p=np will be a more relevant result for AIs than it will for us humans. because it will give the labs an impetus to throw money at np-hard problems knowing solving one could solve a whole class of problems.
>>109763024semantically correct? as in the formulation in the original navier stokes papers and the starting formulation of the lean code of the proof were not the same? be very precise with what you mean
>>109763044>as in the formulation in the original navier stokes papers and the starting formulation of the lean code of the proof were not the same?Correct.
>>109763048Do you know of any such inconsistencies in the language then? Usually mathematics is very precise
>>109763070No, I don't. I was merely clarifying the point that anon was trying to make.
>>109763070Brain status: missing
I don't like that a lot of what these AI CEO guys said these things would do years ago ended up being true
>>109763080i'm not an expert and i assume neither are you guys i just googled and apparently there is a lean initiave to precisely formalize the millennium prize problems so it's clearly a problem thought about by mathematicians so if you claim the proof is invalid somehow i assume you at least have a a good reason to do so
>>109761075Random the concept doesn’t include letters dumbass.
>>109761102>>109761498>>109762723>NoTRaNdOmThen why aren’t you a billionaire if you can solve cryptocurrency randomness. Shows how fake you people are.
>>1097624014 color theorem proof was 142 pages
>>109763272Crypto is fully deterministic retard
>>109763070I do, here's one such instance of (deliberate) Lean trolling:https://github.com/xrchz/CollatzLeanThat said I don't doubt OpenAI's proof is legit because the first fucking thing all LLMs do when presented with Lean that compiles is comb over it for precisely that sort of Lean tomfoolery (there's no need to check for other sorts of errors since it compiles, so that narrows it down a lot)
>>109763514Why are you wasting so much time to try to explain a 65 IQ retard things he will never understand?
>>109763514yes i understand how it's theoretically possible to formulate bogus statements i said from the beginning it was possible. but they are rare, and caught early. an issue with the lean interpreter, not with the maths. i was hoping for a technical argument here and im getting nothing
>>109763514It is interesting that so far all of these models have been unable to produce anything novel and all announced breakthroughs are just counterexamples which is suspiciously the one thing very amenable to brute forcing with a decent automatic theorem prover.
>>109763530That took about 1 minute to post including finding the repo again
>>109763552Finding counter-examples via automated tools is a lot easier than positive proofs but the writing is on the wall here it's only a matter of time.
>>109763552Incorrect.https://vibemathed.com/stats#proved-vs-disproved
>>109759537Nobody outside OpenAI has conducted anything yet but we don't even have to wait for verification. Just wait and see if this is even still talked about or makes major headlines after a few days.Because they won't.
>>109763615and ironically it's the "AI-assisted" category (AI only used in limited fashion while humans did all the real work) that has the highest proportion of disproofs, while the ones where AI played a lead role are >60% proofs.Turns out finding disproofs is where human capabilities peak.
lol this dude got cucked hardhttps://cims.nyu.edu/~tristanb/statement.pdf
>>109762401+100 pages proofs are pretty standard but they usually are written quite well with the author trying very hard to cut down page count. The slop is just slop and very hard to follow.
>>109763615>https://vibemathed.com/statsUnreadable vibe trash website, the first graph explodes into the top. Also I guess just 21% of the total or whatever entries are even lean verified. Funnily enough when you search for erdos in the entries list you only get 15 results despite in the graph 81 solved erdos problems being claimed as having entries. My main point still stands all the major "breakthroughs" and headliners have been counter-examples.
>>109759537>be researcher>find a mistake in LLM output >publish it>be claimed a luddite>become unemployable
>>109761075>caring about algebra is something only good little goy slaves doI mean... he is not wrong
>>109763826That's right, the less you think and understand, the less of a goy you are.
>>109764020This is actually true as fuck, but hylic cattle don't get it
It’s a good thing I majored in garbage disposal in commutative collage
>>109759117>I was told...>I was told...Useless shite
>>109764050>cure cancerOnly retards think this is a scientific achievement or an actual definite thingWe already know how to cure most cancers
>>1097640503, 4, 5 and 6 haven't happened yet.
>>1097646215 has definitely happened
>>109759117
>>109764621Warning Cope Red Cope Protocol Gamma Nine engagedPreparing mass diversion Attempting to subvert reality in T Minus 10
>>109764549i would presume that "curing cancer" would include>being able to cheaply identify all forms of cancer before it becomes untreatable>being able to cheaply combat all forms of cancerwhich would be incredibly impressive. you can't really consider cancer cured until the only people dying from it are luddites/fundies that refuse treatment and thirdies without medical facilities
>>109763778>Also I guess just 21% of the total or whatever entries are even lean verified.Most are at least "Lean-checked", which means it compiles but no one has taken the time to do a proper audit of the statements yet. There's something of a glut in new results that make it difficult to review all entries to rigorous standards.And of course some just don't have Lean proofs because the authors didn't supply one. It's easy to forget that systematic Lean verifs are a relatively recent convention (made possible by AI)>Funnily enough when you search for erdos in the entries list you only get 15 results despite in the graph 81 solved erdos problems being claimed as having entries.Apparently you cannot use a search functionErdős Problem # returns 141 entries, though "candidates and partial results are not counted" in the stats page (most of the Erdos entries are "site-confirmed" though)And it seems you cannot read either, because it's 87 solved Erdos problems. Maybe you should leave the browser use and visual parsing to your agent, I hear GPT-6 Astra (tm) is very good at that, though even Gemini 3.8 Flash might substantially improve upon your performance.
>>109762088>>109763720>everyone uses AI now you fucking luddit-ACK!
>>109763501Then get better specs to solve the issue. Complaining isn’t going to get you anywhere. Who cares if you’re in debt you’ll earn it back.
>>109760933>denies that LLMs are "sochastic parrots">while posting, in the same post, links to Buckmaster's statements where the guy claims openai literally stole a good part of his research>doesn't even realize that LLMs have to be spoonfed and hand-holded into proofing stuff...
>>109766514Dumb esl
>>109762401>Never heard of a mathematical proof longer than a few pages.https://dn760103.eu.archive.org/0/items/Prof.Stenio.Musich-Grigori-Perelman-Mathematics/0612069.pdf
>>109762401>never heard ofyes, and? Why does it matter what some braindead layman has or hasn’t heard of
>>109766537>noo you are ESL therefore your dumbprove me wrong, you retardalso, let me know when LLMs find a proof that is NOT a counterexample.
>>109766588>tech broyou have no clue what that means and you still haven't provided any valid arguments lmaoyour beloved jewish leaders will not be grateful for your shilling, btw.
>AI program you're using steals your work and its parent company takes credit Exactly what Alex Karp said would happen, just a month or two ago
>>109766601Dumb esl tech bro
>>109759117>we spent tens of millions of dollars to solve a problem that would have previously cost tens of millions of dollars to solvemother of god
>>109766667can you elaborate on this? sounds interesting.
>>109764644Solving a pure math problem requires generating a proof, not just bruteforcing an example in 100 hours with over 9000 agents at the cost of a couple millions where you could have used a 15 lines python script running on a commodity laptop for 30 minutes.
>>109764853There isn't really such a thing as an untreatable cancer. When the term is used it's because the only tool we have for that cancer is chemo and/or surgery, but if we had a CAR-T option, only death (or like a few hours off of death) would really qualify as 'untreatable'.Cheap is also not really the right way to put it within the scope of single-payer healthcare.Though in relation to this, personalized medicine has been a plausible solution for a bit more than a decade and it is indeed economics that has blocked development of them: drug developers haven't figured out a way to profit sufficiently from them due to high upfront discovery costs and low patient count to pool from after (obviously).That said, there is no funding for techbio startups addressing these issues as all funding has moved to already validated drug assets, so you're not going to see any changes there in the next 20+ years.
>>109765636There is no issue, moron
>>109762401>Never heard of a mathematical proof longer than a few pagesyeah, there's a lot retards don't hear ofWiles' proof of Fermat's Last Theorem was over 100 pages
>>109765093>search only works on exact matchesIt‘s alright aijeet, I know implementing a simple normalized keyword search is still out of reach for your gods, maybe your Jewish masters should spend the next 15 million dollars training the next model to be capable of implementing a proper search function that would be impressive.
>>109759145No real use case beyond pure mathematics. Physicists and engineers have always known these continuous equations can't capture every detail about real world particle systems.
>>109768690This, but also physicists by and large think far too highly of these shitty models.
>>109762887P might be NP but all P-algorthims that solve NP-hard problems could be completely impractical for any useful n.
>>109759117>but there is currently a major jew-led shilling operation trying to hype up their product by making accusations in a mirroropenGayI is so transparent.
>>109769126Or it could not.
>>109764649Luddites are still recycling the same 7 replies for half a decade and think nobody notices.>it's not actually AI (followed by 2 paragraphs of pure semantics)>it's just a machine that predicts the next word (luddite fails to see a use case for a machine built to predict outcomes)>you didn't make it (usually in response to people who didn't claim to make anything)>it's not real art (unlike a banana stapled to a wall)>I can notice that it's AI so it's unusable (but keep using CGI, I don't care that I notice that one)>it's not perfect now and that means it'll never be perfect (because humanity finally hit the ceiling in technology after centuries of progress)>MUH CLIMATE (arguments copypasted directly from when they seethed about crypto, weirdly they don't apply those arguments to crypto anymore)
>>109769405Shartyfags are still doing the samesóyjak wiki.org/Trolling>ctrl-f "snailcat">ctrl-f "luddite">ctrl-f "vibe code"
>>109768751Yes and no. There are a ton of physicists these days who lean more towards mathematics than physics, extending formalisms that basically describe a fantasy land that exists nowhere besides their mathematical delusions. Entire subfields have sprung up around such efforts, with quantum computing being one of them.
>>109769800Quantum computing is CS and not physics.As for the rest, you're just repeating what I said.
>>109759117More AIslop lying and fraud. The line must go up! Let the normies eat lies!
>>109769831Quantum computing and the third state is based on false paradigms and when it gets a useful result it's through random luck. The reason for this is that the universes metal data is not entagled or layerd, the cat in the box is just metadata until it needs to be rendered. There is nothing there a simple and basic effeciency.
>>109769831It's CS, mathematics and a smattering of the Copenhagen model. The conditions upon which the entire endeavour rests, defined by the Threshold Theorem is pure fantasy and what happens when you take mathematics too far without any reference to actual physical reality.Either way, what I'm saying is that its true that physicists like their shitty models, but not all of them are so high off their own supply that they truly believe in them and create entire new subfields out of thin air.
>>109769405>j-just another trillion dollars and it'll be AGI! you'll see! >it is real art, actually :)... why did I lie and say I didn't use AI? NONE OF YOUR BUSINESS LUDDITE! >let it control military equipment, so what if it hallucinates?>prevent it from hallucinating? u-um, I uhhh...>it's okay that it's used in mass surveillance and already accusing people of crimes they weren't in the same state to commit, IF YOU HAVE NOTHING TO HIDE THEN YOU DONT NEED TO WORRY, GOY!
>>109769142Solving some gay old 25 year old puzzle is pretty cool anon. The bigger impact though will probably be math, as a subject, being kicked out of the hard sciences. None of these gay little puzzles that "mathematicians" have dedicated their life to solving actually have any impact on anything. No project was ever blocked until Navier Stokes was solved in the same way that no one was blocked from not having an engaging new piece of poetry done by an African American studies professor.When this is all said and done, most people won't end up saying "wow, AI is amazing for solving math" they'll end up saying "wow, I can't believe we were paying math PhDs to play the equivalent of really hard crossword puzzles all day that didn't make the world any better".
>>109770692But realistically, if that is the outcome, it would have proven that AI is incredibly useful. It can act as a bullshit detector machine after all.
>>109759117>sam altman: "we tried this because there were rumor on the internet">rumor: math professor close to solving navier-stokes>the professor: using codex for a year>openai: has all his logs>checked every codex session tagged to N-S>found the most promising one>spun up 10,000 agents to finish the proof where he couldn't>they got caught >told the guy if you say anything you'll destroy your own career>he ruins his career over this>openai: "we did not see his work">also openai: "we cannot rule out that de-identified data from their usage helped improve our models"they stole it lmao
AGI will solve any problem, but first you need to type the solution unto the prompt!
>>109770717did you repost your own tweet or someone else's?
>>109770725Terence Tao solved the Erdos problem by just telling ChatGPT to "keep going" and to "not give up" over and over until it solved the problem. The bulk of the work in getting AI to now solve problems is convincing it that they can be solved and to ignore the crusty old meatbags with their appeals to authority that something can't be done because it not being done is the way it is and you should never question your elders.
>>109770773all of the solutions had the actual data inside the matrix. None of this was invented from scratch. Same with navier cock. The guy just forgot to opt-out of data scraping. Now openCock won.Not sure why they make it such a big deal. He did the work, but openAI won. Sometimes you just fuck up
>>109759117>my dat works at Nintendo
>>109770692>>109770702>>109770773No anons, it’s creating fake mathematical equations with imaginary equations. Stop with the lies.
When will AI solve the J problem
>>109770692when really what we're saying is "wow, billions of dollars and the most impressive thing AI has done is solved really hard crossword puzzles that were mostly already done"
>>109770986>it’s creating fake mathematical equations with imaginary equations.This is true, and yet they happen to work.
>>109759117Still no external verification.
>>109771031What I'm saying is "billions of dollars to math departments in the past 30 years and all they've done is bash their heads on puzzles", meanwhile AI is a trillion dollar industry and has made this glorified puzzle field of study redundant. It's more of a stain on math as a field than on AI.
>>109771240The fuck you want? Do we have to beam the pdf to andromeda galaxy and wait for their response?
>>109759117>>109771240No mathematical proof has been published yet. Just a pile of LEAN shit.I asked chatgpt to find me the mathematical proof, it said that the mathematical proof was in 166-page PDF document. I asked chatgpt to find me that 166-page PDF document. Chatgpt apologized for saying that there was a 166-page PDF document because it could not find it.
>>109771832I want to read this PDF. Where is it?
>>109759133Kek
>>109773069Retard.
>>109773128Well, this certainly is a 166-page PDF, and does claim to contain a proof, but this is not a mathematical proof. It looks like a comprehensive regurgitation of what is already known, with a bunch of paragraphs in between, and honestly random shit to obfuscate the fact that: nothing new was discovered.I hope some team of PhDs do not waste their life combing through this shit.I suppose it is impressive that AI has learned to lie the same way as people lie: with jargon and gibberish that cannot be understood.Damn, I was actually excited there for a minute.
>>109773722>this is not a mathematical proofIt is tho>nothing new was discoveredJust the answer for one of math's most famous questions
>>109759117I know this gonna sound like cope but it's not THAT impressive. The AI did the thing AI has been really good at so far: it found an obscure paper and managed to create a counter example. While the proof is impressive it's not new math that is going to allow us to do things differently. We now just know that the equations we use to simulate fluid mechanics are incomplete and breakdown under certain circumstances.
>>109774107AGI TOMORROW SAARRRR
>>109774018for context, this is an actual mathematical proof.notice how it is concise, clear on what it will prove, proves, then provides a formal conclusion what it proved.https://joefinberg.com/wp-content/uploads/2024/08/Proof_Of_Maxwells_Equations.pdfNotice how the Navier Stokes slop does not provide a concise summary at the beginning and does not end with a formal three dots (therefore) it is proven.
>>109774435retard
>>109774530I might be retarded but I am still not seeing any potential improvement on propeller design that will reduce cavitation coming from your ai-god's slop.
only the fool is in awe of that which cannot be understood by thinking men
>>109774566That part is correct. Everything else you said is retarded though.
>>109774646ok, then. here is an 1848 page proof that the wisest man in athens was the one man who knew that he knew nothing: https://www.amazon.com/Plato-Complete-Works/dp/0872203492/
>>109759117you idiots made me genuinely crash out over this
>>109775126>hope you enjoyed your drunken napnot gonna lie, zoomer slang makes me feel some type of way, as if english is cooked, my feelings are valid, its cringe, gives me ick, not bussin'
>>109775317unc
>>109774435This is barely a step up above>they put letters in the numbers, what's up with that??
>>109775421>go netflix and chill yourselfonce I began to comprehend zoomer slang, and its inherent stupidity, i decided to start using english without any slang at all, so in a way I did learn zomething from zoomers, unfortunately it was in the form of counterexemplar.
>>109771832I want it to be confirmed it works by a source with no ties to the company. Ideally no ties to Microsoft either
>>109775523see Brandolini's law
>>109775559The millenium prize has already been paid out to the sum of 1 million once before. Clearly they didn't just take the researchers word for it and confirmed it themselves first.
>>109775617rather than I putting in such effort to refute you, I will simply refer you to Brandolini's law
>>109775654And I will refer to you Retardini's law.
>>109775660>RetardiniHe got fuckin wacked.
>>109775559>>109775654It's been confirmed that OpenAI lied before.
>>109760968I did one of my final exams with a TI-84 calculator. I did very well, but I believe I will disavow my test results, instead bestowing it to (and crediting) Texas Instruments instead. I was basically cheating, by using the calculator. I did nothing.The calculator will graduate instead. I will help it look for a job.
>>109775754stock up on batteries.
>>109763368That is because the proof listed the combinatoric possibilities. The symbolic proof part was very short.
>>109761187And terry davis will laugh at him.
