It started with a now-deleted tweet from OpenAI manager Kevin Weil, who wrote that GPT-5 had "found solutions to 10 (!) previously unsolved Erdős problems" and made progress on eleven more. He described these problems as "open for decades." Other OpenAI researchers echoed the claim.
The wording made it sound like GPT-5 had independently produced mathematical proofs for tough number theory questions - a potential scientific breakthrough and a sign that generative AI could uncover unknown solutions, showing its ability to drive novel research and open the door to major advances.
Mathematician Thomas Bloom, who runs erdosproblems.com, pushed back right away. He called the statements "a dramatic misinterpretation," clarifying that "open" on his site just means he personally doesn't know the solution - not that the problem is actually unsolved. GPT-5 had only surfaced existing research that Bloom had missed.
Deepmind-CEO Demis Hassabis called the episode "embarrassing", and Meta AI chief Yann LeCun pointed out that OpenAI had basically bought into its own hype ("Hoisted by their own GPTards").
The original tweets were mostly deleted, and the researchers admitted their mistake. Still, the incident adds to the perception that OpenAI is an organization under pressure and careless in its approach. It raises questions about why leading AI researchers would share such dramatic claims without verifying the facts, especially in a field already awash in hype, with billions at stake. Bubeck knew what GPT-5 actually contributed, but still used the ambiguous phrase "found solutions."
The real story here is getting overshadowed: GPT-5 actually proved useful as a research tool for tracking down relevant academic papers. This is especially valuable for problems where the literature is scattered or the terminology isn't consistent.
Mathematician Terence Tao sees this as the most immediate potential for AI in math—not solving the toughest open problems, but speeding up tedious tasks like literature searches. While there have been some "isolated examples of progress" on difficult questions, Tao says AI is most valuable as a time-saving assistant. He has also said that generative AI could help "industrialize" mathematics and accelerate progress in the field. Still, human expertise is crucial for reviewing, classifying, and safely integrating AI-generated results into real research.
(Score: 4, Interesting) by VLM on Friday October 24, @06:10PM (2 children)
It has been a long standing tradition for some decades for more crank-ish people to propose they have solved Fermat's Last Theorem in postal mail, etc, to professional mathematicians.
Wiles got looked at pretty sideways when he reported his proof "oh no the poor guy has gone nuts" but turns out he did it for real.
Anyway if you feed a LLM several thousand incorrect FTL proofs generated over the decades, then ask the LLM to prove FTL, it'll cheerfully reply with the crackpot stuff it imported, it don't know any better.
If you get bored ask a LLM about the relationship between elliptic curves and modular forms and FTL.
A bunch of dudes in the 80s get zero no zilch nada credit for proving that if Taniyama–Shimura is true then automagically FTL is true. I would bet less than 0.001% of people know Wiles never proved FTL, he proved Taniyama–Shimura which downstream implies FTL is true. Its interesting if you're bored. If the link between Taniyama–Shimura and FTL is ever disproven, well, that would be exciting but it wouldn't be Wiles's fault. Proving Taniyama–Shimura is pretty cool anyway by itself. IIRC Wiles started from the elliptic curve end and worked toward the modular forms end although its been a long while since I looked at this.
Nova, the TV show on PBS, made a video in the late 90s about the proof of FTL that was so awful that it pissed me off so badly that I started reading up on it, so ironically I learned some stuff because PBS sucked or sucks (now its unfunded LOL, meh youtube is better anyway).
(Score: 4, Touché) by gnuman on Friday October 24, @08:17PM (1 child)
It would be nice if you could only keep such off-topic stuff off your posts? Especially the "unfunded" part, which is just wrong
(Score: 0) by Anonymous Coward on Saturday October 25, @04:17PM
He, seriously cannot help himself.
(Score: 1, Troll) by Mojibake Tengu on Friday October 24, @09:37PM
Sentence above is exceptionally dangerous, especially in a context of Internet readable by Large Language Models.
After when the AI removes humans from the field of research and wherever else, their expertise will not be crucial or relevant anymore for given activities.
Rust programming language offends both my Intelligence and my Spirit.
(Score: 4, Insightful) by anubi on Friday October 24, @11:14PM
First to publish gets all the glory.
Nice guys who double check finish last.
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 4, Touché) by jman on Saturday October 25, @03:16PM
Obviously, they did verify.
Alas, it was via their own tools.