Are we in a new age of science and progress?https://x.com/__alpoge__/status/2079028340955197566
>constructive proof
>>17020353>this won't happen until 2029-2032 btw-THERE ARE NO BREAKS ON THIS TRAIN
>>17020353people will point out in this case it's not proving anything but disproving something, which is easier of course. but those people are missing the point since ai has already proven new things. yeah, mathematicians are cooked.
you too could be published in annals of mathematics if you chatted with your AI friend more
>>17020353LMFAO at these copers. An actual mathematician with an actual morgan prize and a past in the field says he's found a counterexample complete with wolframalpha links and these retards count it as "just some dude on twitter"
>>17020361An autonomous web agent wrote this post
>>17020353
Humans couldn't come up with this simple counterexample in over 80 years?
>>17020392too advanced for human minds
Okay but what's the use case?
The most important discovery we've made thanks to the advances of LLM in mathematics is that most mathematicians are in fact not wholesome chungus autists who love the beauty of maths but arrogant nerds who wish they could give their name to an important result.Sorry chuds, this is the new world, you will prove nothing and be happy.
>>17020353>Are we in a new age of science and progress?You can test this empirically:>sample LLM for all the """unsolved math problem""" slop you can get out of it>add the results to the training set>train a new model>check if the learned data distribution changes besides model collapseProtip: the answer is objectively 'no', there is no progress in this, unless you expect humans to make it.
>>17020369political activists control the wiki platform.what do you expect
I hope this would stop students studying math. What a waste of potential.
>So what if my microwave oven disproved the twin prime conjecture? It was a stupid problem anyway.
>>17020422>I hope this would stop students studying math.Then who's going to develop math further?
>>17020425We need the math people to prompt the ai that’s about it
>>17020427>We need the math people to prompt the aiPrompt the AI about what, retard? "Solve le heckin' 50 years old unsolved conjecture #4203" ad infinitum?
>>17020429Yes. Once all open problems have been decided, mathematics has become solved and we can move on.
>>17020425AI will do everything. Literally.And you won't find any economically viable application of your skills -> you will suck in life -> you will leave little to no offspring -> you will go extinct over time.Thus excessive human mass will be reduced. No conspiracy, mate. The same shit has been happening in the West for over a century because of the agriculture automatization. Now it already is 10 times faster and ten times more severe. Eventually humanity will be torn apart when approaching the techno singulatiry.>>17020427You are thinking narrow, we will not even need to prompt anything. Math was developed as a tool to do fun stuff like engineering, military and architecture. Now we will do everything directly, omitting manual (even heavily automated) calculations.
>>17020431>Yes. Once all open problems have been decided, mathematics has become solvedThis is what "AI" users actually believe... What I want to see is the latest demographic data on 4chan users. There's no way the userbase is more than 5% white at this point.
>>17020433I've been here since 2005. How about you, wokajposter?
>mfw showing this to chatgpt and it responding with 'what the fuck'
>>17020432>AI will do everything.You mean hypothetical future magic AI that's coming in exactly Two More Weeks? Surely you don't mean LLMs, whose performance at out-of-distribution tasks ranges from bad to abysmal.
>>17020436>You mean hypothetical future magic AI that's coming in exactly Two More Weeks? Surely you don't mean LLMs, whose performance at out-of-distribution tasks ranges from bad to abysmal.So far the progress of the AI is so fucking obvious you can't deny it.You may cope, you may cry, you may shit your pants and post silly pics, but this is inevitable. It is already doing advanced math, it is already laying out hundreds of thousands of people. No stalling, no plateauing, only acceleration.
>>17020437>the progress of the AIOoops. Found the saar. What progress has been made in making LLMs succeed at out-of-distribution tasks since ChatGPT 3?
>>17020437*laying off(fix)
>>17020439>out-of-distribution tasksAnd you? Can you speak chinese without learning it?
>>17020441I accept your full concession that no progress has been made towards general intelligence in LLMs since GPT-3.
>>17020441>And you?people with average intelligence outdo LLMs at those
>>17020442>towards general intelligenceLet's put aside the fact that it's a completely copious vague definition.But what if narrow AI would be just enough to completely fuck you up? You, your life, the economy and the world you live in.
>>17020444At what? At learning various natural languages? Oh boi.
>>17020445>it's a completely copious vague definition.Imagine being so mind-bendingly brown that you say this right after I gave you a quantifiable, machine-friendly working definition.>But what if narrow AI would be just enough to completely fuck you up?And what if a gang of 80 IQ Anglo thugs were enough to completely fuck up your entire civilization, Rajesh? :^)
>>17020446>At what?at OOD prompts. are you retarded or what? do you use AI a lot?
>mfw one day someone's going to sit down to take a shit, and prompt gpt-10 to solve navier-stokes and then tweet the proof out while wiping their ass
>>17020447Bruh, would you stop larping as a white person? Post hand or GTFOP.s.Oh, sorry, I forgot that hindu AI-bots don't have hands
>>17020450>post hand, saaaaaar!!!!I accept your full concession that no progress has been made towards general intelligence in LLMs since GPT-3.
>>17020449Coping Ramanujan fans would still say AI is a nothingburger
>>17020451>>17020442>I accept your full concession that no progress has been made towards general intelligence in LLMs since GPT-3.I bet this moron is pajeet. They have this shit called mantra or smth. They repeat it over and over again while praying.
>>17020458>mentally ill ESL brownoid continues seethingI accept your full concession that no progress has been made towards general intelligence in LLMs since GPT-3.
>>17020464>Saaar, Einglish is my pharst language only, Saar! Vee are Britishers now, Saar!
>>17020422that guy in the OP has a math phd and works at anthropic. you can study whatever you want if you're high iq and still land on your feet
>>17020475>mentally ill ESL brownoid continues seethingI accept your full concession that no progress has been made towards general intelligence in LLMs since GPT-3.Also note how this retarded animal just can't stop responding. :^)
>>17020478>spamming the same cope time and time again>calling someone mentally ill:^)
>>17020480-farts on you-
>>17020480I accept your full concession that no progress has been made towards general intelligence in LLMs since GPT-3.>y-y-your c-coping with my inability to provide any counter example, saaar!!!Lol. Be sure to reply again, animal.
>>17020476That doesn’t match reality unfortunately.
>>17020476>if you are top 0.001% you'll be alright
>>17020494>>17020449>>17020437It probably found a counterexample in some long-forgotten soviet russian paper. And generally, Jacobians are irrelevant in maths, no one even cares about them - absolutely not significant finding.Nothing to see here, there has been no progress in LLMs since 2020.
>>17020499https://x.com/nihilunbounded/status/2079109869739860450LLMs have a massive knowledge graph and can and can plug numbers in very fast. I'm curious what the chain of thought looks like for this problem
>LLMs have a massive knowledge graph
>>17020511they don't?
>>17020514A LLM doesn't even have any explicit representation for the concepts that would make up a knowledge graph.
>>17020369>2+2=4>SOURCE???
>>17020517they have vector embeddings which are pretty much the same thing
>>17020353What are the practical applications of this?
The future is asking an LLM to solve a problem and then asking another LLM to verify the solution
>>17020612And how is LLM gonna verify a solution to the problem of me finding a gf?
>>17020369>you don't get it! he has an...........undergraduate student award! respect him!i won the science fair in fourth grade too. im basically a fields medalist
>>17020437BASED.Funny how AI chads are called jeets when in fact jeets are unable to even know how to prompt to create good software.AI is just moving human intelligence one level up.
>>17020614Have you tried asking it to honestly help with that? It would probably give you actionable advice. The question of whether you truly act on it would be up to you.
>>17020449
>>17020499a pretty close 2D map was published in 1999.If I just gave you this paper and asked you to contradict the conjecture you could probably do it.
Luddites BTFO again.Accelerate to the Singularity!
>>17020614>>17020627Nowadays they will answer it. But early days, when I asked how/where to meet girls, they would just refuse to answer the question. It wasn’t even that creepy.
>>17020654It was amazing watching this board be so behind it all even few months ago when it started happening, every time I brought it up no one knew what I was talking about and once they looked into it they would have the biggest nervous breakdown I've seen on this board. It's like your entire worldview being shattered in front of you, psychologically very similar to people from strict religious upbringing.
what most likely happened is that a mathematician discovered this counterexample by itself and anthropic or whatever AI company paid him millions to say an AI model helped himjust a bullshit advertising story sorryI need irrefutable proof that an AI came up with this by itself with no outside assistance
The luddite seethe is strong ITT
>>17020353cool...how is that useful?
I was tested to have an IQ in the mid 80s so I don't understand any of this. Without making fun of me can someone explain the significance of this?
>>17020740We're going to die. You are going to die.
Some context: https://x.com/gro_tsen/status/2079215501490717094
>>17020712Holy copium overdose
>>17020712how about you post irrefutable proof that a human came up with that with no AI assistance
Friendly reminder that no progress has been made towards general intelligence in LLMs since GPT-3.:^)
>>17020619I don't understand how you can consider yourself smart and not know basic history of technology, or be unable to adapt to whats happening and what has already happened a billion fucking times before and it keeps happening forever?source to picrel: https://www.ams.org/journals/notices/202607/noti3372/noti3372.html
>>17020595>they have vector embeddings which are pretty much the same thingThey are absolutely not the same thing, not even remotely. You're clueless.
>>17020353how can AI prove something? it cant learn so that means it probably read it from someone
>>17020412>there is no progress in this, unless you expect humans to make it.Only non-retarded post ITT. Until humans use these results to conceptually advance the field the LLM isn't getting any better at math, either.
>>17020781arc agi begs to differ
>>17020797>arc agi begs to differSo which language model does well on out-of-distribution tasks according to """ARC AGI"""?
>>17020798>out-of-distribution tasksno such thing
>>17020801Ask your "AGI" about it, mentally ill retard.
>>17020802it says no such thing. it knows everything.
>>17020652>26 years or more>it's easy bro
>>17020809It's ok, anyone who isn't a psychotic patient can check for himself and confirm you're just another literally delusional "AI" fan.
>>17020816name 2 things it wouldn't know
>>17020818Any concept that hasn't been invented by humans yet? It's that easy. How come the people most obsessed with "AI" are the least likely to spend more than 2 seconds reasoning about it or anything related to it?
>>17020821you can't actually prove thisif it hasn't been invented by humans, you have no way of checking it knows it or not
>>17020825>you can't actually prove thisYou can't be actually this retarded.
>>17020826known unknownit's clear you haven't thought deeply about the subject
>>17020827Arguing with the mentally ill is a waste of time. If you need further help, ask the token guesser to explain why your posts are fucking retarded.
>>17020830i see you have no response and have been reduced to ad hominemi would feel embarrassed for you, but i don't feel embarrassed for a dog when it can't do calculus either
>>17020831>i see you have no responseSee >>17020821That responds BTFOs you immediately and shows how mindless you are.
>>17020832already respondedare you having trouble remembering things?
>already responded>are you having trouble remembering things?Mentally ill retards think their incoherent mumbling is a response.
>>17020780Not how that works. Their claim: >"AN AI CAME UP WITH THE COUNTEREXAMPLE"So show the proof. Show how it can be replicated by anyone. AI companies have every incentive to lie and obfuscate the true utility of their tools. Until they show step by step how they an AI was able to come up with the counter-example, it is safe to assume they are lying.
another insane person suggesting that somewhere out there is the avengers of mathematics, secretly tearing through decades old problems only to attribute to them to llms, and they're doing this because...
>>17020855because...AI companies have every financial incentive to lie about how great and revolutionary they are to milk as much investor money?? Have you not been paying attention for the past 3 fucking years>AI is going to do everyones jobs>AI will change the economy forever>AI can do everything !>WE'RE MAKING GOD GIVE US ALL THE MONEYthis whole story is an advertisement you gotta be retarded not to see it
>>17020857you sound like a crazy person
>AI companies get more money based on perceived importance >AI companies drastically overstate their importance to get more money>2 + 2 = 4>"you're crazy" you sound like a retarded person Again, it's simple. If AI came up with this counter-example by itself then that is a ridiculous achievement, and the exact steps to get there should be provided and replicated by anyone. Until then, it's a farce.
much less farcical to believe that they've trapped the spirit of ramanujan and are making him solve problems for them
>>17020476Nobody will land on their feet.THE SKIES ARE FALLING AND IT'S OVER, DO YOU UNDERSTAND?IT'S OVERAI WILL REPLACE US ALL, REPENT! REPENT NOW!
>>17020855CIA has a catalog of 0-day proofs to burn for international strategery.in this case, they gave one to anthropic so the USA stands a chance against the kimi and qwen news.
>>17020864Surely you have thousands of dollars worth of tokens lying around to replicate it yourself?
>>17020864you literally have the prompt lmao faggot
>>17020903where?
>>17020353>>17020354>>17020355Thank fucking god AI is replacing mathematicians. I just want good math progress. they gatekeep their profession and expect to get paid massive salaries and get self-congratulatory medals all to make snailspace progress through math. winning about 'muh credit' (muh legacy >>17020361)the whole way.with clankers we can finally just have a math matchine that spits out the only important part with none of the bullshit. FUCK YOU for holding math ransom to your petty earthly bullshit in the first place
>>17020907watched a recent Wildberger video where he talks about AI and he says something similar
>>17020499hot lil trvkky
>>17020714We're going back to horses, thoughbeit.
>>17020499>Jacobians are irrelevant in mathThere are only two kinds of mathematics: calculus and precalculus.There are only two kinds of calculus. The fundamental theorem and the Jacobian Conjecture.
>>17020941Are you having a stroke
