Bartek Wlodarczyk

@kareem_carr Yeah, rebuilt using chain of agents with Lean that only by chance resolved 9 problems and failed with 300 others. Let's not forget AGI is coming next week.

I'm honestly kind of awestruck to be witnessing this. The methods of one of humanity's oldest intellectual traditions being rebuilt from the ground up right in front of our eyes.

@sumjitg Uh oh... what a shame for hypesters.

Just saw Demis Hassabis push back hard on all the recent Erdős problem hype. He basically said: sure, today’s AIs are knocking out some of these tough math problems, but that’s nowhere near real AGI. It doesn’t come close to the kind of raw creative invention someone like

@Polymarket LOL, no. They tried all 353 and only 9 were able to solve using Lean

JUST IN: Google DeepMind says its AI solved 9 open Erdős problems — considered among the hardest unsolved questions in mathematics.

@bandyci_drogowi @Rz_Policja Not sure what's the point of these videos.. Was returning from Swietokrzyskie to Wroclaw on 1-3rd May and 99% of dumb fucks disobey rules. Police are another lazy fucks. That's clearly the country made of shit and sticks. Always was..

Moim zdaniem kierowca wyczerpał wszystkie znamiona drogowego idioty: autostrada, BMW i brak mózgu. Myślę, że można pójść o krok dalej i powiedzieć, że ów kierowca BMW po prostu się "zmajtczakował". - napisał nam autor nagarnia.

Kolejny Majtczak. Potencjalny zabójca, do natychmiastowego wyeliminowania z dróg zanim kogoś zabije. PY12170

@clashreport Medium rare or well done?

IDF announces death of Staff Sergeant Noam Hamburger, 23, killed by Hezbollah.

@realannapaulina @IvankaTrump LOL sure thing

We’ve just learned of the horrifying plot to assassinate @IvankaTrump — a devoted wife, daughter, mother, & someone with no personal quarrel with the Iranian regime. She was targeted to bring pain to her father & to her husband. There is nothing more sick and demented than that…

@tech__unicorn Total bs

the elites don't want you to know this but we've already achieved AGI - the labs know - the government knows the reason you're still getting "impressive but not quite there" announcements is because nobody has figured out what happens to markets, jobs, and governments the day the world actually believes it so we're in a managed disclosure buying time Anthropic please!

@nicksortor Cuba to fuel more distraction from the island

🚨 NOW: President Trump is now BACK in Washington, DC after CANCELLING his weekend plans in New Jersey CBS is reporting intelligence officials have ALSO canceled Memorial Day plans, and are on standby. Something may be imminent 👀

@kimmonismus It won't in his lifetime. Bullshit crap. He shall be ashamed to fart such nonsense.

Demis says the Singularity may now be only a few years away, potentially set in motion by the arrival of true AGI. "Its being so transformative, it will be the most important technology ever"

@TruthTrumpPost 3? Why not 1.5?

NOW: Poland has officially received its first three F-35A "Husarz" stealth fighter jets at Łask Air Base 🇵🇱 This makes Poland the very first NATO nation on the alliance's eastern flank to operate these advanced fifth-generation fighters. Warsaw ordered a total of 32 F-35As in a massive $4.6 billion deal, with full delivery locked in by 2030. A massive, historic upgrade to secure Europe's eastern border against growing threats 👀 Huge win for Poland, NATO, and global security 🇺🇸🇪🇺

@mehdirhasan Not killed but sent to space 😈

The character based on him is killed in it.

@BenjaminPDixon Nonsense. Why you're talking things you don't understand?

I'm not sure what to say to people who don't realize our brains are literally producing the wetware equivalent of the next-predicted token with every word we say and compose.

@ai_sentience Ironically you failed intelligence test

If you don't realize the AI is smarter than you you have ironically failed an intelligence test

@thePartyPartyUS @AcerFur Uhmm... glad you ate all brains 🤣 keep prompting and let me know when you'll get cure for cancer.

@bartek_wl @AcerFur the AI is solving Erdos problems, and you aren't classifying that as intelligence. Whatever you're running on, it's not intelligence either.

I think we can finally put the “LLMs are just stochastic parrots” ideology to rest now

@thePartyPartyUS @AcerFur LOL, you basically have no experience with anything AI, apparently. Read again, there is nothing emotional with my responses.

@AcerFur @bartek_wl you're arguing facts and Bartek is arguing his emotional response. There's no words you can use to convince people like that. They'll come around in time, but only when they're confronted with direct experience in ways their emotional guards can't block.

@davidmbarnett Recent Google I/O was about curing cancer with LLMs and next week OAI will show how to cure all diseases with prompting their Erdos model... oh... wait...

Hi, AI people, have you cured cancer yet? Or are you too busy making art and writing books?

@emollick It's disproof and no new math was created.

Its funny how much the whole "strawberry" thing, which turned out to be o1-preview, was dismissed as overhyped at launch when it is clear in retrospect that it was way underhyped. A direct line from models unable to do basic math to solving unresolved math problems in 18 months.

@AcerFur You're clearly confusing things. Highly guided search over human-designed abstractions as evidence of broadly creative mathematical reasoning conflates sophisticated optimization with the invention of fundamentally new ideas.

@bartek_wl This is a gross oversimplification of the process. There is ingenuity required in realising to set the problem up in the search space it looked through. This was the actual insightful contribution.

@AcerFur Result concerns a specific combinatorial construction problem where a computer search over algebraic number theory structures is a natural fit. It doesn't straightforwardly generalize to problems requiring conceptual leaps or genuinely new mathematical frameworks.

@bartek_wl Not sure how that’s the case when it resolved something many mathematicians failed to and so that solution data would have never existed for it to have been trained on, ergo it’s not just “parroting” its training data since it’s pushing beyond it.

@CesareGArdito @ValerioCapraro No they don't, dude... really, go back to basics.

@bartek_wl @ValerioCapraro There are LLM-produced original new ideas in front of the world to see, I understand that your priors do not allow it but, for the love of God, update!

AI has not solved mathematics. Let me explain why. In the last months, there has been growing excitement about AI in mathematics. Yesterday the excitement reached a new peak, after OpenAI announced that an internal model had produced a counterexample to Erdős’s unit distance conjecture, an 80-year-old conjecture in discrete geometry. Many AI commentators are now jumping on the chairs, shouting: “Math is cooked”. Their argument is simple: mathematical statements have a true/false classifier, so mathematics is “easy” for AI. Unfortunately, this argument is wrong. Let me say it clearly: people saying this have no idea how mathematical research works. Small premise. I have a PhD in pure mathematics. I have published research papers in mathematics and a book. I no longer work as a professional mathematician, but I do have a fairly advanced understanding of how mathematical research works. Mathematical research is not just “solving problems”. Broadly speaking, mathematical research can be divided into problem solving and theory building. Problem solving concerns questions that are already well defined. There is a statement. There is, at least in principle, a proof or a counterexample. This is where AI is going to be really powerful. But theory building is different. Theory building is not about finding the answer inside an already existing conceptual space. It is about building the conceptual space itself. It means inventing objects, analogies, and structures, that did not previously exist. There is no simple true/false classifier for that. No LLM could have invented scheme theory before Grothendieck by simply recombining the mathematical literature of the time, because the relevant conceptual world did not yet exist. And, crucially, many contemporary hard problems of the first type require theory building. Probably we cannot solve the Riemann Hypothesis not because we are missing one clever trick, but because the necessary theory does not yet exist. So no, mathematics is not cooked. What is much more likely is a shift. AI will become extremely powerful at helping mathematicians solve local problems. But the deepest part of mathematics, choosing the right questions, inventing the right language, and creating the conceptual architecture within which future problems become solvable, will still require human intuition. I think that what’s going to happen is that mathematicians will be able to develop theories more quickly with the help of AI. Mathematicians are not cooked. Quite the opposite. Exciting times ahead. * In the image, legendary mathematician Alexander Grothendieck