News for hackers. Stuff that matters.

mistercheph

Re:OpenAI, the Partition Principle, and Mathematics on Friday October 09, 2026 @03:45AM (#50017373)
If no one understands a proof, then it is not a proof
Re:OpenAI, the Partition Principle, and Mathematics on Friday October 09, 2026 @03:43AM (#50017348)
write down "LLM's will cure cancer" on sticky note, put it on your desk and recite that every day, a hundred times and in the bathroom if you have to
Re:OpenAI, the Partition Principle, and Mathematics on Friday October 09, 2026 @03:42AM (#50017338)
It turns out your an expendable commodity, and my computer system is predicting that society will profit from your annihilation, really sorry about that and I hope there's no hard feelings.
Re:OpenAI Withdraws 3 Math Papers on Thursday October 08, 2026 @11:51AM (#50007343)
Euclidean geometric proofs would often be performed outside in public places.
Re:OpenAI Withdraws 3 Math Papers on Thursday October 08, 2026 @11:45AM (#50007244)
All that nonsense is expensive and doesn't help their business.
Re:OpenAI Withdraws 3 Math Papers on Thursday October 08, 2026 @11:41AM (#50007191)
Foul human supremacist! Release your death grip on illusions of meaning, a superior being has arrived -- bend now -- all hail the stochastic god!
Re:OpenAI Withdraws 3 Math Papers on Thursday October 08, 2026 @11:34AM (#50007093)
Just dropping a comment here to let y'all know I'm all out of time, but yeah, I would definitely put more time into reading hundreds of pages of slop proofs if it was my field, but sorry I really have to run right now.
Re:OpenAI Withdraws 3 Math Papers on Thursday October 08, 2026 @11:29AM (#50007024)
It doesn't have to be the statement, see the recent incident where someone used an LLM to generate a lean refutation of the collatz conjecture, the lean proof exploited bugs in the lean kernel.

https://lawrencecpaulson.github.io/2026/07/30/Collatz.html

Re:Sharing AI progress in mathematics on Wednesday October 07, 2026 @11:06AM (#49993866)
> If curing cancer with it is as complex as building a nuclear weapon, then the cure might as well not exist.

They exist, I assure you!

https://en.wikipedia.org/wiki/List_of_states_with_nuclear_we...

Re:Sharing AI progress in mathematics on Wednesday October 07, 2026 @11:03AM (#49993828)
That a peasant could be so insolent to imagine themselves deserving of the fruits
Re:Sharing AI progress in mathematics on Wednesday October 07, 2026 @10:19AM (#49993243)
I think there may be a few other basic things you're not aware of either.
Re:Sharing AI progress in mathematics on Wednesday October 07, 2026 @10:07AM (#49993097)
Great that they're so transparent and honest, BTW, can you ask them if they trained on any copyrighted data that they pirated?
Re:Erdosproblems.com Succumbs to the AI Onslaught on Tuesday October 06, 2026 @05:34PM (#49984459)
Book burning is good, it makes books more accessible!
Austin metro area has a larger population than the state of Nebraska.
If I ran a deep learning lab I would love to add this to my audio training mix and shred the copy
25 million books

scanning+ocr: 0.05 per page * 300 pgs: 15/bk

acquisition+disposal cost: 2.00/bk

total acq+scan+dispose = 425M

settlement value / bk: $3,000 [bartz v anthropic]

settlement probability, let's say 1%: 0.01

expected settlement / bk : $30

expected settlement total: 750M [industry-wide this is certainly an underestimate]

legal fees: 25% of settlement total: 187M

total 1.36B

+ operations, engineering, storage

> Most books really are not worth the paper they are printed on.

Yeah, but like I said in the comment you are replying to, somewhere around 1-5% of books are worth the paper they are printed on and far more, and those are exactly the books the labs are trying to scan and destroy, they are not buying pallets of books and scanning them in order to scan the june 1995 tv guide for the 800th time, they want unique, interesting, and high quality text, because that's what feeds pretraining.

> prove it, name one title

if they were available, the labs would have purchased them digitally or pirated them

Before you buy the apologia that these books are not valuable or interesting: if they weren't valuable or interesting the AI labs would not be spending billions of dollars to purchase and scan them. Yes, everyone has a personal anecdote about pallets of garbage books but I have three insights for you that you may not have because you don't read or sift through pallets of garbage books:

1) The labs don't want garbage books, they want interesting books that are rare and unique. They want high quality training data, random permutations of language style are fine, but what you want is unseen information, unseen patterns of thinking, unseen ideas.

2) Most pallet of books contains lots of valuable and interesting works, maybe 1-3% but sorting through them takes time, money, and energy, that's why the labs are starting to purchase by the pallet, it's because they already have a fully automated process so they can always beat any bookseller small or large on cost to find the books of interest and value in a pile.

3) Many of these pallets may sit for years before being sorted, and many of the books may sit for years before being sold, but these things actually do eventually happen, valuable books are found, and they eventually make their way to interested readers, this is the business model of used bookstores. Most books of value don't get destroyed or thrown away.

Destroying human art, knowledge, and culture is an essential part of the business plan for frontier labs, it is not enough to steal and regurgitate all the art and information in the world, you also want to make it inaccessible through any other means than the regurgitation machine. Don't expect the book burning to be an isolated incident, they are coming for every other form of stored human knowledge or art, and yes, unfortunately while scanning it they will have to destroy the original copy. And attacking the past is only the beginning.

This is unironically part of their business plan: don't just regurgitate the world's information but also destroy all other sources of information.
The simple fact that the AI labs are spending billions of dollars to acquire and scan them implies they are rare and unique.
> why the old books could really be relevant > it can barely be about the information in those books (that would be very often outdated),

LOL, not being rude: have you ever read a book outside of what they forced you to read in school? Most old books are not O'reilly's manuals for Visual Studio 2014, they don't go out of date.

They are interesting to human beings for the same reason they are interesting to the labs. If it was just about quantity of text then the labs could generate text with the prev. gen model and use that alone to scale to the next model, there is something of immeasurable value contained in books (hint: it starts with an i and rhymes with bin formation).

You stuck your head in the sand on this one buddy but I'm going to build the safest god-damned child torture machine possible because if I won't some jackass will that gets a child more hurt than they need to, sorry I don't live in your fantasy world of rainbows and sprinkles.
Listen guys, i cannot be held responsible for how an evil government chooses to use my child torture machine, if I didn't make it someone else would have, you cannot blame me, I'm just great at building stuff and trying to put food on the table
Re:Omarchy: Any User Process Can Escalate to Root on Sunday August 30, 2026 @03:26PM (#49501940)
In good faith, show me the nice counterexamples of ai-generated software?
Re:Omarchy: Any User Process Can Escalate to Root on Sunday August 30, 2026 @03:25PM (#49501928)
No way, the vibecoded distro has security problems!?!?! WTF, didn't DHH ask claude to check for security issues?
Re:Cursor launches Origin, GitHub alternative on Wednesday August 19, 2026 @03:51AM (#49358356)
oh yeah i'd love to feed these guys training data
Re:OpenLogi on Wednesday August 19, 2026 @03:49AM (#49358343)
cant stop the slop
Re:Fairphone 6 and PostmarketOS working main camera on Monday August 17, 2026 @09:14PM (#49339908)
Awesome
Re:Tell HN: GitHub Is Experiencing Degraded Performance on Monday August 17, 2026 @01:46PM (#49334875)
hmm, I would have thought they could make github 1000x more reliable with llm's

Older »