you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 16 Nov 2026
26 points (100.0% liked)
TechTakes
2796 readers
310 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 3 years ago
MODERATORS
Anthropic pulls out the big guns to juice it's stock price right before the lockup ends (right?):
https://www.scientificamerican.com/article/anthropic-mathematician-claims-to-prove-riemann-hypothesis-most-famous-problem-in-all-of-math/
I would love a number-theoretic perspective. From a category-theoretic perspective, this is just not very surprising, and since I don't trust Lean, it's not clear whether there's anything to learn.
In short, we already expected RH to be true. The reason is that RH is a global statement, about all of number theory and the natural numbers, but we already proved several local RH statements for very wide classes of number-like systems; it would be weird if RH failed globally because it would mean that we can't glue all of the local cases together into a single global theorem. The interested reader probably should start with Siegel zeroes, which we don't think exist; my current understanding is that there can only be at most one Siegel zero and it would have already been revealed numerically by distorting the distribution of primes.
I have nothing to say about the mathematics, but my god the remarks are a disaster.
I love my robot girlfriend.
If this isn't just AI psychosis, maybe he's trying to appeal to mathematicians who are angry about the destruction of the values and ideals of mathematical discovery? It's not working.
Edit: never mind, the more I read it, the worse it gets. It really feels like Anthropic is a cult where they believe they are divinely ordained to deliver salvation to everyone. Or not, maybe AI will kill us all. Even OpenAI's announcements tried to be professional.
This feels so much like watching friends and colleagues crash out with a substance abuse disorder.
One wonders just how familiar this person is with the stories of Erdos' amphetamine habit
I continue to think that Anthropic is over-estimating the buying public's interest in obscure mathematical theorems[1]
[1] obscure in a general sense, not a nerd sense
OpenAI already got Navier-Stokes and dumped hundreds of slop proofs onto Github. At this point I believe that mathematical theorems are already priced in, even for the most mold-poisoned boosters.
One thing that is different about this result is that the prompter actually has serious expertise in math and had a major role in guiding the overall direction of the proof.
That definitely changes the situation tbh
Yeah think I'm not going to draw any conclusions over this until more people (who are better at math than I am) have a had a chance to go over the proof.
That might take a little while, given that the Lean proof is over 1GB. The prompter didn't release a natural language proof because it ended up being more than 250,000 pages. Of course, there are many sections that are wasteful and can be simplified.
If people are finding sections that are wasteful or unnecessary in less than a day after the damn thing was shat out into the world, that doesn't speak well of the originator.
They’re currently in the process of verifying the Lean code right now
No idea how long itll take for the 250K pages to drop though
One uses the first computer to generate the Lean code, then one uses the second computer to see if the Lean code compiles, and then one uses a third computer to generate a statement that the Lean code says what it is supposed to, and then someone in Memphis gets coal-miner's lung.
See additional information that I didn't know already arrived ;).
(Thanks, I had no idea).