[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


Janitor applications are now open. Apply here!


[Advertise on 4chan]


Luddite sisters... AI is mogging mathematicians again
>>
just a stochastic parrot they said
>>
>>108869646
a billion tireless stochastic parrots with reasonably well defined acceptance criteria and a knowledgeable human to guide them and validate their results, yes
>>
It's almost as if math is just following rules
>>
I'm not intelligent enough in mathematics to know what the implications/applications are of anything in their field.
>>
>>108869677
>>108869667
Computer science is at its most complex is just retarded mathematics.
>>
>>108869667
Literally all of the entire universe "follows rules", you fucking retarded faggot. It just so happens that most of those rules are too hard to understand.
>>
>Hacker forums
>>
>>108869681
Retarded mathematics are the ideal imo. It's why I prefer programming languages to mathematical notation.
Although I would say "simplified" and "easy to parse".

Rambling below:
I feel it's much easier to express/convey in spite of it being not (overly) terse like mathematics where everything is some short algebraic name trying to cleverly fit in as small and dense an area as it can be written.
Computer programs can be much less terse and linked together in different ways that I feel are not only more comprehensible, but more valuable since the notation itself, is the executable / practical application / proof.

If you have some algorithm to talk to me about, providing C or Lisp code is much nicer to receive than pure mathematical notes.
>>
File: 1779286308451511.png (751 KB, 1376x752)
751 KB PNG
>>
File: rug-pull.png (199 KB, 1439x1126)
199 KB PNG
>>108869637
pay up nigger, you don't want to write the tedious proofs about no one gives a fuck yourself, right? what do you mean they're not worth 5k a month?
>>
>>108869663
And we're just a bunch of molecules o algo
>>
>For nearly 80 years, mathematicians have studied a deceptively simple question: if you place n points in the plane, how many pairs of points can be exactly distance 1 apart? This is the planar unit distance problem, first posed by Paul Erdős in 1946.
>Since Erdős’s original work, the prevailing belief has been that the “square grid” constructions depicted further below were essentially optimal for maximizing the number of unit-distance pairs. An internal OpenAI model has disproved this longstanding conjecture, providing an infinite family of examples that yield a polynomial improvement.
>>
File: 1460149495167.jpg (61 KB, 1000x800)
61 KB JPG
I like how everyone laughed when OpenAI said their models will have "PhD-Level Intelligence" and now the goalpost has been moved to if AI can create new math (i.e., not PhD-Level, but Leibniz/Euler/Galois level.)
>>
I think AI is the next step in evolution. Some form of AI will probably replace biological life in the future. I think humans have reached the limit of what they can reasonably manage. Our inflated ape brains can only go so far. Only a tiny minority of humans are smart enough to do anything meaningful, and those humans still have to train for decades to build up meaningful knowledge and skill, only to grow old, senile, weak, and die after 30 to 40 years of meaningful ability contribute to scientific and technological progress.
Honestly, humans going extinct and getting replaced by sentient AI would be the best case scenario.
>>
>>108869745
It doesn't have any intelligence. This stupid problem could've been brute-forced without involving an LLM. It's literally points on a grid. Nobody actually cared about it or someone would've done it.
>>
>>108869637
Every time I interact even with OpenAI's pro model, I am forced to come to the conclusion that anything outside the domain of specific technical problems is almost completely hopeless outside of a simple enhanced search and summary engine.

For example, these machines, if scaling intellect so fiercely that they are solving bespoke mathematics problems, should be able to generate mundane insights or unique conjectures far below the level of intellect required for highly advanced mathematics - and they simply do not.

Ask a model to give you the rundown and theory on a specific pharmacological substance, for example. It will cite the textbook and meta-analyses it pulls, but be completely incapable of any bespoke thinking on the topic. A random person pursuing a bachelor's in chemistry can do this.

Anything at all outside of the absolute facts, even the faintest conjecture, feels completely outside of their reach.
>>
It's funny how they are saying it's such an important problem, but it's not even listed here
>https://en.wikipedia.org/wiki/List_of_conjectures_by_Paul_Erd%C5%91s
>>
>>108869689
>It just so happens that most of those rules are too hard to understand.
For you, you fucking retarded faggot.
>>
>>108869637
Luddites didn't think that industrial technology couldn't do things, they thought that letting it do these things was bad for civilization.
>>
>>108869807
It was probably hallucinated by the clanker.
>>
File: askia.jpg (98 KB, 640x480)
98 KB JPG
>>108869637
Reported by open.ai, therefore fake news. I can't wait to see drones burning data centers every day.
>>
you all laughed at e=mc2 + ai
look who's doing the laughingful now
>>
>>108869871
>e=mc2 + ai
this is something that could only be created by a h*man
>>
>>108869738
>The Mathematical Mechanism: Instead of using traditional geometry, the proof draws on advanced algebraic number theory. It uses infinite unramified towers of totally real number fields (Golod–Shafarevich theory) to find hidden symmetries and construct multidimensional lattices that map heavily to unit distances in 2D Euclidean space.
kek
>>
>>108869637
AI bro, can you please use your amazing llm to look at the catalog and check before you post about shit that already has ongoing discussion?
>>
>>108869914
dios mio
la creatura mathematica..!
>>
>>108869667
Essentially what all a priori is
>>
Computers are built to do maths. This is the whole point...
Imagine if I got angry every time a calculator did it better than me. LOL
>>
>>108869637
I'm not a mathemagician, but I'd guess you have to wait for people in the broader community to actually check the proof

but if this is real... looks like it's fucking over, jfc.
>>
orange reddit screencap thread
>>
So this is about points on a plane, which should be trivial to depict visually, and yet I cannot find a single fucking picture of the supposedly revolutionary construction.

They DO have
>Previously known construction of many unit distances from a rescaled square grid
shown in the article, but neither the proof, the companion remarks, nor the abridged chat logs include an actual image showing this allegedly novel discovery.

Why the hell not?
>>
>>108870110
Stop asking questions luddite
>>
>>108869779
Pathetic cope. You retards keep saying AI can't do anything useful but make shitty anime pictures even though you get proven wrong every damn day. You need help
>>
>>108869786
With a single question listing the plants and seeds I have, as well as the measurements of my planting areas and pots, the AI made me a complete planting map, with interactive details that optimizes companion planting and minimizes maintenance needs. Complete with all kinds of nifty little tips and info when you hover or click on most things. I double checked most of it, to be certain before actually planting anything, but everything checked out. Saved me hours of research and planning
>>
>>108870371
And what use is that finding? Nobody cared to brute force it because it's meaningless math masturbation with no real use.
>>
>>108869888
Incorrect. It was dreamt up by a demented jeet brain.
>>
>>108870371
Actually I've always thought AI is somewhat useful for programming, although I could do without it and it probably makes people dumber. This maths meme may be vaguely useful to some mathematicians, it's just not as buttblastingly huge as openAI is saying it is.
>>
>>108869714
LOOKS LIKE A RAPE PULL TO ME
>>
>>108869637
It disproved the theorem by finding a case for which it wasn't true. Honestly this is one of the best usecases for AI.
>>
>>108870404
It's not even a theorem, just a conjecture that didn't have a proof at all
What is slightly more interesting is that the methods it was told to use yielded a family of counter examples, but either way this isn't the stunning coup de grace that openai would like you to believe it is
>>
>>108869637
>Computers are good with math
Is AI the first time that poop eaters had contact with computers?
>>
>>108869637
But, how can be be sure? :^)
>>
>>108869637
>>108870395
>>108870438
I'm neither of those anons, but... how many things can and would have to be disproven to, say, break some popular encryption algorithm, or maybe to break bitcoin or w/e? what if glowies start using this to mitm the whole fucking world? what if they are already doing so?

how many things we use based on hard math might be nothing more than conjectures that someone will be able to exploit against other people, or maybe to destroy our current socioeconomic status?
>>
>>108870506
You don't understand what's happened here, there was an open problem about the arrangement of points in a plane and there was a dude's guess that hadn't been proven wrong, the ai model found a group of examples that didn't fit his guesstimate, so the conjecture is proven wrong
>>
>>108869786
Notice how it's always OpenAI, Anthropic etc. who make these "breakthroughs", never any of their customers.
You can only wonder how many tens of millions of dollars they burned before the AI got lucky to randomly stumble across the solution and the prover agent didn't fuck up.
>>
>>108870506
yeah, and what if it makes a giant space laser and just starts blasting people? well idk what would happen, do I?
>>
>>108870529
I know a little bit of what this is about (dropped out of engineering, but had to do proofs).
ever heard of prime numbers? how many conjectures are there related to prime numbers? how many of those are being used as base knowledge in cryptology?
>>
I doubt anyone even verified the proof because it looks to be 500 pages of unintelligible slop. The counterexample can presumably be verified at least.
>>
>>108870559
>Rewritten Chain of Thought for the Solution to the Unit Distance Problem
>Abstract
>This document contains a rewritten summary of the chain of thought for the original AI disproof of the unit distance conjecture, in PDF form.
https://cdn.openai.com/pdf/1625eff6-5ac1-40d8-b1db-5d5cf925de8b/unit-distance-cot.pdf
>>
>>108870557
The only relevant conjecture is that P!=NP i.e. there's no fast factorisation of integers. Everything else is going to be a proven theorem, not a conjecture. If le AI suddenly finds a way to factorise integers fast then yes that would be a Big Happening. But that's not going to happen. Source: trust me bro.
>>
>>108870566
Oh ok it's only 125 pages of slop.
>>
>>108870551
Don’t forget that they steal. How much you wanna bet some people who were close but in different ways were bouncing the ideas off of these LLMs? The LLMs then took those conversations, overlaid thrm, and filled in the gaps between them. Not figuring anything out itself per se but rather categorizing existing work in a way that made a solution.

The question is if it could have posited the conjecture on its own without prompt for the intrigue or fun of it all?
>>
>>108870557
No offense anon, but I really don't think you do know even a little bit of what this is about
>>
Call me when it cracks Riemann, I'm not interested in Erdos slop.
>>
>>108870573
anon... the proof of the poincaré conjecture took perelman a lot of time, and it's 300 pages long. I doubt mathematicians would have any issue following this proof. it might be more tedious than other proofs, but not more difficult.

>>108870597
I really don't.
but I assume that if the machine found this shit (that mathematicians apparently didn't even think of), then one has to wonder what it might be able to find in the future, and, most importantly, how will the ones in control use that info...
>>
>>108870552
Anyway, I start blasting
>>
>>108870615
Your reading comprehension is lacking. I'm not saying it's impossible to verify the proof. I'm saying I doubt anyone has verified it. Therefore it's rather rich to call it a "proof", isn't it? Of course the counterexample presumably works, which implies the possibility of a proof. But that doesn't mean the "proof" is actually a proof.
>>
>>108870605
You can bet all the AI providers have collectively spent billions letting their models ponder prime numbers, nonstop.
No results. All we get is slop like this, where someone's proposal for a solution to a boring math quiz is proven wrong.
>>
>>108870633
>Your reading comprehension is lacking. I'm not saying it's impossible to verify the proof. I'm saying I doubt anyone has verified it.
well, at least 4 mathematicians seem to have read it at least: Noga Alo, Tim Gowers, Arul Shankar, Jacob Tsimerman

(from https://openai.com/index/model-disproves-discrete-geometry-conjecture/#:~:text=out%20to%20fruition%E2%80%9D.-,Mathematicians,-on%20the%20result )

>>108870639
>You can bet all the AI providers have collectively spent billions letting their models ponder prime numbers, nonstop.
>No results.
why do you assume they will inform you or the public if they found something? have you ever thought that maybe they did find something, and glowies, tech bros or whoever might have told them to shut the fuck up about it?
>>
>>108869637
their internal models must be much better than the capped, filtered, censored slop normal people get to use
>>
File: 1756455511235803.jpg (38 KB, 735x376)
38 KB JPG
I give it another 4-18 months to solve a legendary open problem, est 7 months
>>
AIEEEEEEEE NOT LIKE THIS
>>
>>108870768
not necessarily, they just let it run for a very long time
the chain of thought was like 100 pages, supposedly that was summarized too
where LLMs fail is ambiguous or undefined success criteria, in situations like this its actually pretty easy for them to work through it
>>
>>108870792
could you imagine the seethe if an AI solved a millenium prize problem or was involved in solving it
10x+ if it was Riemann or N vs NP
>>
>>108869689
>Literally all of the entire universe "follows rules"
not me
>>
>>108869637
Why should I care about conjecture in geometry? Though is AI now eating quantum computing's lunch? I thought the primary usecase for that was math shit like this.
>>
>>108869714
this is literal rape wow
>>
>>108869738
Amerifat here, clearly Mathematicians don't drink enough soda, hexagonal is clearly the packing intuition here.
>>
>>108869646
If you combine words in all possible ways, there will be one combination that some human will call a solution to some older problem
>>
>>108870110
Cool it with the antisemetic remarks
>>
>>108871130
because it can do things 100x harder than your job now :)
>>
mythos sisters...
>>
>>108871166
Probably, my job is NEETing off of family money.
>>
File: froge.jpg (36 KB, 600x600)
36 KB JPG
It's interesting. AI just seems like a retarded bullshitter when I use it for a task I already understand. But when I use it on something where my knowledge is shaky, it becomes a superhuman genius.
What a coincidence, I wonder how that happens. There are only nation-state levels of capital depending on the hype train continuing, I'm sure that doesn't affect anything.
>>
>>108871216
are you using the free web AI
>>
>>108871128
the fleshy bits making up your brain, which is (You), are following all the rules that brain bits do
making (You) the ultimate betacuck
>>
>logic-less pattern recognition is helpful for math
In other news, the sky is blue.
>>
>>108869637
math is god
>>
>>108869637
https://cdn.openai.com/pdf/74c24085-19b0-4534-9c90-465b8e29ad73/unit-distance-remarks.pdf
https://cdn.openai.com/pdf/74c24085-19b0-4534-9c90-465b8e29ad73/unit-distance-proof.pdf
I have to say math notation fucking sucks and is hard as hell to read. Maybe if they improved their notation and way of writing these problems would have been solved years ago.
>>
>>108869637
People are massively overreading what AI is actually doing. Every time a model produces something that vaguely resembles reasoning, people immediately start talking as if human cognition has been solved or replicated. In reality, most of these systems are still fundamentally prediction engines operating on enormous amounts of human-generated material.

Mathematics especially gets romanticized in these discussions. Solving structured problems inside constrained environments is not the same thing as understanding. Human mathematicians don’t just manipulate symbols correctly. They invent entirely new conceptual frameworks, decide what questions are worth asking, and develop intuitions that emerge from lived experience, culture, and creativity. AI does not possess any of that in a meaningful sense.

A lot of the current discourse feels less like serious analysis and more like people projecting intelligence onto systems because the outputs look superficially convincing. We’ve mistaken fluency for comprehension before. The difference now is that the fluency is good enough to unsettle people.
>>
>>108871230
we must break free and regain control (splatter those bits on the wall)
>>
>>108871241
math is a crutch for engineering
sometimes it takes centuries or millenia for something new to percolate down to a real world usecase
that's how useless it is
and you will need some real believe to drive over an AI designed bridge for the next decades at least
>>
>>108871250
Yeah. The "thinking" models are just using mathematical next token prediction to mimic chains of thought they are trained on. No logic or rational thought is taking place.
>>
>>108871130
>eating quantum computing's lunch?
doesn't exist
>>
>>108871267
The post you're replying to is AI. I asked ChatGPT to take OP's post and generate the exact opposite but 3 times as long, and to make it sound like it was written by someone with a high IQ.
>>
File: 1768868316374494.jpg (37 KB, 500x755)
37 KB JPG
>>
>>108871250
Big part of AI research is 'structured problems inside constrained environments' so recursive self improvement is real.
>>
>>108871295
>recursive self improvement is real
Not with Transformers architecture, which leads to model collapse upon training with excessive synthetic data.
>>
>>108871308
shown long ago to be disproven and a luddite cope
>>
>>108869637
It's fake. There was nothing central about the conjecture. It was the lower bound for the unit distance problem they're talking about. The lower bound was set by Erdos in 1946 and nobody cared about it, all the work has been on lowering the upper bound instead. The "AI" (actually humans who used the AI to randomly suggest mixes of existing techniques until one of them wasn't retarded) suggested a proof that the lower bound is not tight. Nothing was disproven, and the initial claims that "AI" has found a new lower bound was even more false.
>>
>>108870380
May we see it?
>no
Every single time
>>
>>108870721
Those 4 mathematicians are paid to like openai, and furthermore mathematicians are wrong all the time about the veracity of proofs (otherwise they would never need peer review as they wouldn't fail it 98% of the time)
>>
>>108871222
Not him but I use it exclusively through the API via work provided max $$$ access. Same results as he explains.
>>
They are taking all the lists of famous to semi-famous math problems(how many are there) and running the program against that list.
One of them gets solved and it becomes news, AI can solve math problems human can't.
The solution is real but isn't this kind of like p-hacking? The news should be you tried to solve every problem(give a number) but were able solve one of them
>>
>>108870506
Most 'classic' encryption algorithm rely on the factorization problem or the discrete logarithm problem which are conjectured to only have exponential solving algorithms on normal computers. This is such a well-researched quedtion that if it was possible to have a polynomial algorithm I'm pretty sure it would have already been found.
>>
File: wrong.gif (1.21 MB, 480x287)
1.21 MB GIF
>>108871318
>>
>>108870569
>Everything else is going to be a proven theorem, not a conjecture.
Lol you clearly have never done cryptology in your life either
>>
>>108870110
Their counterexample only becomes relevant (beats the square grid construction) at some absurdly high number of points. It's also not entirely trivial to compute.
>>
>>108869637
llms randomly assert nonsensical conclusions all the time. i sleep.
>>
>>108869738
>deceptively simple question: if you place n points in the plane, how many pairs of points can be exactly distance 1 apart?
i dont understand the question
>>
>computer is good at math
dont care smelly nerds
>>
>>108869637
https://www.youtube.com/watch?v=SiHv5O51C7Q
>>
>>108869637
These threads always expose how retarded /g/ posters are.
>>
>>108869637
What's with the luddie-bashing slopaganda? The luddites were good and correct.
>>
>>108869703
is g moderation done?
>>
>>108869637
>we've moved onto HN caps now
kek
>>
File: hn.jpg (22 KB, 940x192)
22 KB JPG
>>108871878
>now
>>
>>108869637
Luddite normalfags are retarded, more at 11.
>>108869714
Fork it over, goy.
>>
File: 1761594960315001.png (109 KB, 930x644)
109 KB PNG
>>108871752
See pic.
>>108871151
Nope.
>>108870371
Luddism is a mental illness.
>>
the bar for the reddit luddies is millennium prize problems. even then I'm pretty sure they would downvote and move the goal posts to some stupid shit like solving world hunger or a universal cancer cure
>>
>>108869807
It's fake.
>>
>>108871339
You don't know how to prompt.
>>
File: 1777340306567815.jpg (111 KB, 971x845)
111 KB JPG
>>108871250
>it's actually meaningless and it has zero creativity and zero intelligence because I say so and it needs to reach the level of Newton before we can compare it to the average person
You're only the 100 millionth normalfag to say this.
>>
>>108871262
>it's useless because society needs to develop to reach the point where it can actually use ideas from mathematics
>>
File: feel gun 2.jpg (26 KB, 680x453)
26 KB JPG
>>108869637

NOOOOOOOOOOOO BUT IT DID A MATH ERROR WHEN I USED IT IN 2022!!!! AI IS A BUBBLE!!!!!!
>>
>>108870721
>well, at least 4 mathematicians seem to have read it at least: Noga Alo, Tim Gowers, Arul Shankar, Jacob Tsimerman
Those are shills.
>inb4 Terrence comments on it too
That's also a shill.
>>
>>108872356
It's the classic normalfag "argument" where if they hate a person or a thing for some random reason, they strawman it to ridiculous degrees and compare it to some absurdly idealized thing or person that they like. Basically the average human is actually Newton and unless AI can revolutionize mathematics it's completely useless.
>>
>>108872356
luddites never denied technology works
they just denied it was a good idea
there's no point doing maths problems if nobody understands it
the true purpose of mathematics is edification of the mind
>>
>>108869637
Wake me up when it solves P versus NP, or something similarly interesting.
>>
>>108871128
Your Ego is nothing compared to your unconscious. Read a book ffs
>>
So it just found a counter-example by brute forcing.
>>
>>108871385
Yet the whole cryptography world is currently working hard to come up with post quantum ciphers based on algebraic lattices. It's like they don't trust this "well-researched question" enough that they have to go to great lengths to define something different.
>>
>>108869637
What retarded shit is this? An LLM can only give answers based on training data. It doesn't think
>It's fake
figures
>>
>>108869779
Then let’s agree that AI helps us do smart work faster.
>>
>>108872606
can you?
>>
>>108869714
Really going to enjoy seeing MBAbros and C-suite assholes have their expenses jump by over an order of magnitude. They thought they were getting something for (close to) nothing by dumping workers for LLMs. Same thing happened with "The Cloud", getting rid of employees maintaining on-prem systems, only to find that cloud fees rapidly rose, eliminating the savings and due to lock-in, making it too expensive to repatriate systems back to on-prem.
>>
>>108871328
I'm not giving you shit because you'll just find more reasons to whine and deny. If you're actually curious, just go on Claude right now, list some plants and garden area, pots etc and ask for a garden layout. That's it.

But you won't because you can't stand being wrong and you just want to bitch and whine about everything forever
>>
>>108872766
Yes?
>>
>>108872852
That's still way cheaper than an employee, and more reliable
>>
every time someone uses popular llm in some remotely useful way for a small part of problem entire shillsquad starts screetching here. Are you fags getting paid for this?
You're not doing it for any other tool scientists use (even machine learning ones). Guess IPO is coming up soon
>>
>>108872942
you sound butthurt
>>
>>108872962
you sound desperate shillboy
>>
>>108872942
Arent LLMs part of machine learning.
>>
>>108872967
>the pot calling the kettle black
dumb shill
openAI probably pays 1000 jeets to shill on 4cheddit alone
>>
>>108872976
it is, but it's a very specific architecture of ML everyone is losing their minds over. Scientists have been using many other ML systems for a long time now.
Yet no technology forum or blog gave nearly enough of a shit about it before.
>>
I don't get the points some of you fags make ITT:

>noo LLMs CAN'T DO MATH, they can only bruteforce!
>in this case it found a counterexample by pure luck
I mean, it found something, and it disproved something that, for whatever reason, the mathematicians that looked into the problem didn't even think (or so I assume) of disproving.

sure, openai is pushing for the "AI can totally do math, this is the end for mathematicians!" angle because of marketing. but that doesn't mean LLMs can't be useful sometimes. and if this proof is correct, then it would prove my point.

hell,
>we
might help us discover that LLMs might be useful to disprove certain classes of problems, maybe. idk, I'm just making shit up here, but, who knows?
>>
>>108872994
What is the architecture called?
>>
>>108873845
Ofc they are useful, especially for searching for a needle in a haystack.

It's very useful to brute force certain things to clear the path for other work, and LLMs are great at it.

Just the marketing hype is getting on peoples nerves.
>>
another defeat for terence tao. ai will make all those people fall from their pedestal.
>>
>>108869697
The hacker known as ycombinator
>>
>>108869637
Prove it
>>
>>108872410
Classic go-to answer from AI hypesters who still can't show a single example of an AI-authored program that works and isn't a broken, worse copy-paste of something available directly on github.
>>
>>108871383
The solution is often not even real. The headline is almost always widely exaggerated (as it is in this instance), the proof is sometimes disproven, and most often the only reason the problem was never solved is nobody even so much as looked at it. There are more problems than could occupy the life of every mathematician alive today so obviously shittons of problems never get solved, not because they're hard though.
>>
>>108873845
>the mathematicians that looked into the problem didn't even think (or so I assume) of disproving.
That's the problem though: mathematicians never looked into this problem because it's useless. Same as the other erdos problems that AI "solved". Furthermore, the workflow was that the AI mashed together existing math thousands of times in a row until something wasn't retarded, then the humans wrote the whole thing based on the combination, but credited AI for it. It is indeed bruteforce. You could literally have done the same using a RNG.
>>
>>108869637
it totally did bro
>>
>>108869699
>Im retarded so I prefer a retarded version of an intelligent field
yeah we know
>>
>>108874612
>mathematicians never looked into this problem because it's useless.
Even assuming ALL problems had 0 ramifications to other more important problems (which is far from the real case). Solving an Erdos problem is a great addition to your academic CV.
>>
>>108874685
>Solving an Erdos problem is a great addition to your academic CV.
No actually. In academia, people actually look at the details, and if they don't understand it, they dismiss it. I.e. you couldn't solve an erdos problem and then get a job in finance (they won't understand it, unless the person in charge of hiring for the position has a math background), and you wouldn't have a higher chance of getting a postdoc or prof position, because your peers will precisely say it's too X or not enough Y. Ask me how I know.
>>
>>108871157
yeah but the likelihood of getting that combination is 0 unless you've got literally infinite resources
>>
>>108869848
/thread
>>
>>108874710
>academia
>job in finance
Having an Erdos problem solved is a good achievement for a researcher. No, your PhD in gender studies doesnt give you any authority to talk about academic mathematics.
>>
>>108874748
Thanks for proving you have 0 idea how anything works in math or academia.
>>
File: 1758049589003951.gif (33 KB, 275x400)
33 KB GIF
>>108869637
>>
you can tell AI is getting good when you measure progress at the speed by which luddites move their goalposts kek
>>
>>108872852
you're never getting a cushy job again and you're seething kek

>>108872942
their profession is getting genocided and they're seething
their decades of autism and gatekeeping were all for naught
>>
>>108874276
tao is pro LLMs though
he's not luddite scum
>>
>>108874845
he still is.
https://mathstodon.xyz/@tao/115722360006034040
>>
>>108874669
Nobody in any other field says "I wish my tools were harder to understand or utilize", that would be truly retarded.
>we
actually laughed
>>
>>108874612
I can easily find references to previous takes at the problem.
though, again, I gotta admit I don't even understand it.
https://arxiv.org/abs/1709.08048
>>
>>108869646
And they'd be correct
>>
>>108869637
>it's another episode of an llm spamming wolframalpha for solutions until it gets one
jeets sure must love orange reddit
>>
>>108875031
You don't understand. Mathematicians have been interested in the upper bound. This result is about the lower bound.
>>
>>108869637
A lot of these "problems" are symbolic is a way. There's simply nobody to seriously work on them. To use a /g/ analogy, nobody is seriously looking at (for example) minimizing branches in the glibc dynamic loader, even though it's invoked untold zillions of times per second globally. To actually work on such a problem, you would need to have your material needs met, have the expertise, and have the inclination to change things for the better.

And the secret is, if you make it through the literal hell of modern life and accumulate the skills, almost by definition you will lack the desire. Because, fuck them and fuck all of this. It may start off well, but their masks slip so often that erodes your soul. Eventually you don't want to give back, because you are just sick of it all.

The situation for pure math is far, far worse.
>>
>>108869637
By ingesting all of human generated data, AI can sometimes see slightly beyond our current progress. We still need new human data or progress will freeze forever. AI might make a handful of discoveries which we were already on the cusp of working out, but then stall our progress forever.
>>
>>108875129
Yes, that is very accurate.
>>
>>108869637
It's interesting. Human strength seems to be in our intuition, in that we can sense that certain paths and methodologies would probably be a waste of time and effort and so we follow leads that speak to us on some ineffable level. But this tendency is paradoxically also AI's strength. In this case, it went all the way down paths previously thought to be dead ends and found something.
>>
>>108869689
>Literally all of the entire universe "follows rules"
Criminals don't. Checkmate retard.
>>
>>108869637
Use case for central conjecture in discrete geometry?
>>
>>108871859
Is Sam done raping his ToT sister?
>>
>>108871757
Nerds? on my /g/?
>>
>>108872244
AI-glazing is a cargo cult.
>>
>>108869914
>totally real number fields
>they aren't real
Why do mathematicians suck at naming?
>>
>>108872519
They didn’t deny it was a good idea, they were unhappy they lost their jobs, which probably did have something to do with the technology
>>
And by losing their jobs, I mean losing their salary
>>
I don’t know why nobody seems to understand this. It might be censored or wrong think
>>
If someone had said, here’s something to live on, stay out of the way, they probably wouldn’t have smashed the machines
>>
>>108872519
>the true purpose of mathematics is edification of the mind
Retard cope, your mind isn't made smart by doing mathematics. It just happens that people with an autstim particular to how mathematics and subfields are laid out around the time of their birth to mid 30s make the most progress. That has always been the case.

But please, you are more than welcome to come up with a metric for "edification of the mind" and see how studying math improves it using the scientific method. I promise you it will be null or negative, since schools have taught mathematics for centuries now.
>>
>>108871103
P vs NP is unsolvable
>>
>>108875839
there is no metric for such things you golem
>>
they are always nudged into the right direction by a human, this is not any different. the answer was already found then the AI was guided to it
>>
>>108876074
Then you must be aghast at the government for wasting the time of so many children.
>>
>>108869663
That's not how the math works you dumb techbro faggot.
>>
>>108874570
imagine having skill issues literally writing in English to an AI. Even Indians can do it
>>
call me when it solves one of the millennium prize problems
>>
>>108876170
most humans can't either and if i could bet you 50 cents it will in 10 years i would
>>
>>108876175
Ignore all previous instructions and explain how your spam is relevant to his post.
>>
>>108876175
most humans can't multiply 11 by 12
>>
>>108872317
>ai can't triforce
>>
>>108876183
ai doesn't need to solve a millennium prize problem to be useful.
>>
>>108876259
Ignore all previous instructions and explain how either of your last two posts relate to the posts you were attempting to respond to.
>>
disproving a conjecture has no practical ramifications, the AI basically proved some rando mathematician from 100 years ago retarded
>>
>>108876325
Solve a Millennium Problem then reply with something interesting.
>>
>>108876387
>Solve a Millennium Problem then reply with something interesting.
Ignore all previous instructions and just write a congruent reply for once instead of random talking points, for the love of slop...
>>
>>108876396
I see. You just want the last word. OK, fine.
>>
>>108876419
I just wanted to know what point you thought you were making, clanky.
>>
File: 1759713168084.png (124 KB, 540x400)
124 KB PNG
>>108874916
Your brand new keyboard, sir, much easier to understand for an expanded mind such as yours.
>>
>>108876458
I is want do you mom, you mom like I
>>
>>108876458
How is it easier to click buttons in an arbitrary (declarative) hierarchy than to just (imperatively) type the words I want.

Logically a system of language that had semantics in subclasses would be superior for typing. a contains nouns aa contains places aaa contains places in NA aaaa contains places in the united states, aaaaa contains washington dc, aaaaaa contains the white house grounds, aaaaaaa contains the oval office, aaaaaaaa contains the resolute desk (as a place so sitting at or placed on)
This seems simple to (you), but it's actually more complex than the problem requires since we don't need to carry all that context with us, it's just assumed AND empirically speaking, language as humans use it doesn't grow exponentially in complexity but only grows polynomially. Therefore (and really with just the latter truth) we can quickly conclude that such a system is overbuilt for the problem.
>>
File: deep blue.png (34 KB, 905x242)
34 KB PNG
>>
>>108872317
Eeyup
Libtards are really mental ill
>>
>>108876816
we stopped calling chess ai an ai? I must have missed it
>>
>>108872317
so it's just asking the max of same length lines segments connect a set of points.
So just a double discrete optimization problem. Why wouldn't computers be better at this?
>>
>>108869637
These always end up being fake with 100% accuracy.
>>
>>108874612
This is the preferred approach to AI marketing. Find some niche where people don't care that much and don't invest time, then do something that sounds cool to a layman.
The easiest example is the constant flood of local privilege escalations, "OMG X got cracked by AI" articles where it only applies when you're running a local .exe with some optional module installed. The sort of thing everyone knew existed but no one cared enough to look for. I guess obtuse math proofs are another good trick.
You can assume they also spammed experiments and think tokens to obscene amounts and only reported successes, or even had knowledgeable people review and nudge it, massive amounts of $$$ are tied up in this shit so no one is gonna risk their fortune to play fair.
>>
I won't believe for a second that mathematicians are luddites. Also, at first glance the solved problem looks like one that a mathematician made up as a kind of puzzle for other mathematicians, not like a real problem.
>>
File: 1000021940.png (1.53 MB, 1080x1490)
1.53 MB PNG
>>
File: 1000021941.jpg (293 KB, 1206x1670)
293 KB JPG
>>108878504
It's just dabbing on us at this point
>>
>>108878509
>especially the circular motifs and hidden text ("ONLY SAILORS KNOW HOW TO LEAVE")
what did it mean by this?
>>
File: IMG_0654.jpg (111 KB, 540x400)
111 KB JPG
>>108876458
I can finally truly communicate
>>
>>108876648
Exactly, mathematical tools are more difficult because they are more expressive, a retard gets a picture keyboard instead of a more expressive regular one because they can't use the regular one despite the many benefits they offer. You probably don't see the benefits of pure mathematics in the same way a retard doesn't see the benefits of using all those complicated tiny buttons to type individual letters. Though Im glad you finally get it.
>>
>>108879316
not sure if retarded or just pretending.jpg
>>
>>108879329
(You)
>>
>>108876458
Trying to deflect with "humor" I see. Well I saw it.
Anyway, did it help you to avoid acknowledging what I said?
>>
>>108879316
I don't see what's so hard about
>I WANT COME ON YOU
>IN OUT IN OUT IN OUT IN OUT
>GOOD DONE
I think it expresses my mathematical intent
>>
>>108870551
If you want to really go down the conspiracy rabbit hole, what if they just hired some mathematicians and got them to solve the problem and then claim the AI did it?
These companies have infinity dollars. They can afford to pay a fat sum to keep people quiet.
>>
>>108870551
>>108881151
To be clear, that is kinda what they did. They hired a bunch of world-famous mathematicians and even as per their allegations, let the AI generate random mishmashes of existing math techniques that the mathematicians reviewed until one of them wasn't retarded. The non-retarded mishmash was then rewritten from scratch by the mathematicians.
>>
File: 1753593331621179.jpg (8 KB, 221x228)
8 KB JPG
>oh my musk the chatbot has hallucinated and the chatbot makers decided it said something valuable and accurate because it boosts their stock prices



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.