Hurried Announcement: Alessandro Masullo GRH

2346 points1059 commentsa day ago
Aurornis

The manuscript claims he has been working on this problem for 12 years. I can only imagine the feelings that must have hit him when he saw the Anthropic announcement. Though if he believed he had solved it, he must have known it was only a matter of time until one of the AI companies found it, too.

A hurried announcement with the draft published within hours is the only way to claim a tie to the discovery. Time will tell if either holds up under further analysis, but at least he has shown his hand to the world with a timestamp on it now.

show comments
prometheus1992

Is it possible that brian conrad somehow uploaded the proof to anthropic to cross check something and then the anthropic guys got the proof?

- it doesn't have to be the full proof - even if portions were uploaded, that should also be enough or at least calls for a thorough check on ants side. the human proof has been in the works for 12 years, who knows when something was uploaded somewhere or some info just stayed on a google drive which was later sold.

show comments
kyykky

While ago, there used to be a world where a person, with large enough brain capacity, could earn a living simply by being smart. That is no more the case. Earning your living by intellectual endeavour will require an access to a machine which is not controlled by you, not controlled by the government representing you, but rather a group of very rich individuals who may just decide to bar your access. Perhaps you attempted something they do not approve, perhaps it was just a mistake. In any case, it is hard to claim that you still have control.

show comments
Kotlopou

This looks crazy? 940 pages and the very first section 1.1.1 is empty. I can see why the author wanted to polish it. Let's just hope somebody clears this up somehow.

But at the same time Brian Conrad is presumably not a crank (per Wikipedia he worked with Wiles on FLT), but at the same time it seems insane to try and keep a 940-page result for a year hidden from public view in the current mathematical climate.

Somewhere around this summer, the world stopped being comprehensible...

show comments
rlt

Mathematicians should start regularly posting hashes of their in-progress work with verifiable timestamps to at least claim partial credit when AI solves the problem they've been working on.

show comments
pbkompasz

900 pages vs 250,000 pages Compression is intelligence

show comments
Anon84

There seems to already be a counter example that invalidates this proof

https://philarchive.org/rec/NIEIOM

show comments
dimator

What a gut wrenching experience this must be, to work on something for 12 years and have it snatched out of your fingertips. And on top of that that something is one of the most important open questions in mathematics. Absolutely heartbreaking for that guy.

show comments
jc_811

Hey can someone familiar here explain the why the GRH is such a big deal to mathematics? What will this allow us to unlock? Will it have any effect on the general population?

show comments
smokel

For those who don't eat math for breakfast: "GRH" is the Generalized Riemann Hypothesis [1].

[1] https://en.wikipedia.org/wiki/Generalized_Riemann_hypothesis

show comments
JoshTriplett

A timely announcement like this makes me wonder: was the time just "right" for this, in mathematics? What proofs would have come out without AI, in the next few years?

Or, alternatively, did people start working harder in some fashion, even on creative problems that require deep insights, because they felt it was their last chance?

Or, alternatively, are people just (sensibly) starting to post all their work-in-progress results to establish priority (on either the results or approaches used)?

show comments
caidan

So many people in this thread are sure that they have economic value when their strength, brains, and knowledge are worthless. This is going to be a century of rude awakenings for all!

show comments
dcolkitt

Even a short audit by Astra shows multiple, easy to verify errors across the manuscript.

Maybe it can be repaired with time, but sorry to say this is not anywhere close to a completed proof

show comments
brap

I’m no mathematician, and I’m genuinely curious: how do these things usually work? Are other mathematicians now expected to sit and read through literally a thousand pages of condensed math to review this? And somehow have enough focus to catch subtle mistakes?

show comments
morsch

So, the work was done without AI, okay. But how likely is it that his colleagues, who were reviewing and advising him, used Claude to do so? Maybe even a year or two ago when they maybe didn't realize that this would end up in the training set?

show comments
MinimalAction

Very conflicted on how to take news like this. Should I be happy and delirious for the science we objectively get to explore regardless of who does it or should I be scared for the commoditization of knowledge that makes someone like me with average talent sidelined? Is this the only way to claim originality against a machine that was trained over all (stolen) data of humanity?

gr_norm

By the linked paper, this has been in the works for 12 years.

show comments
demibabs

https://philpapers.org/archive/NIEIOM.pdf

This write up is claiming his proposed proof is wrong.

show comments
tensegrist

i love algebraic number theory and learned a little bit about it in a past life

it's very cool that there are multiple ways to arrive at this result because that means more things will be understood about the integers as a result, and so if i have a vested interest here it's that. i want some cool mathematics to come out of it and i want to learn and understand some of it, not having had as much contact with mathematics as i'd like in years

many people also hoped for the proof of the riemann hypothesis to involve the creation of some insane new technology like "F_1 geometry" [0]

i am far from being an expert and only have enough background to appreciate the overall directions each of these proofs takes. with that said:

the human proof at a high level looks to be in the spirit of hilbert–pólya [1], connes, etc: doing spectral theory with prime numbers ("primes are like atoms in a gas", "primes are like geodesics on a 3-manifold", etc) and expressing a sum in two ways, one as an L-function, one as a sum over the spectrum of an operator, and showing that everything on the right side can be accounted for by zeros on the line so nothing remains. masullo makes a big deal in the preface of the fact that his method cannot possibly work in characteristic p

the alpöge–claude one (which i understand even less, having read remark.pdf and discussed the lean output with some llms) involves several pieces of very hard machinery being applied to translate the way stepanov proved rh over curves in characteristic p to the characteristic zero (reals, complex numbers) case, which is hard because a lot of tools you have vanish entirely. this is very much in the tradition of the langlands program

[0]: https://en.wikipedia.org/wiki/Field_with_one_element

[1]: https://en.wikipedia.org/wiki/Hilbert%E2%80%93P%C3%B3lya_con...

_1

> S(G) the space of Schwartz-Bruhat functions on a Hausdorff locally compact abelian group G.

I'm done.

show comments
segmondy

I was happy when Kasparov lost to Deep Blue, it was special, we can make these damn machines compete.

I was sad when Lee Sa-Dol lost to AlphaGo, realization that one day the machines will just beat us via sheer brute force.

I'm crying now when someone in a month could beat 12 years of effort due to AI.

show comments
Timwi

I find it curious how neither the title nor any of the section headers in the table of contents, mention GRH. They also don't say “Riemann zeta function”, or even “prime” (except in the title). Is this because a sufficiently clued-in mathematician can tell from the titles that that's what it's about? I realize that I'm not in the target audience but seems weird.

softwaredoug

Do human mathematicians need to publish more incrementally? Seems many are hoping they’ll save up their findings for one big bang finding?

IMO academia could benefit from lower key, more tech blog style discussions of their work, and get away from this attempt to only publish the final result.

show comments
zone411

There are errors in it that may or may not be fixable. It would be sad if they didn't use AI to look for holes because they were paranoid that AI companies would steal their work.

show comments
yolo3000

Will any of these math discoveries lead to more concrete discoveries in other fields, or are they simply 'puzzles'?

show comments
buzzy_hacker

Given this proof is orders of magnitude shorter, seems like we could have AI translate it into Lean without much of an issue to see if it checks out.

show comments
troyvit

> It is all human, with no AI. The hope was to bring Alessandro to Stanford for a year to polish the writing and then announce it after running it by enough people. Now it seems best to put out the current draft.

It's a little like the Patch Gap, but for math proofs.

show comments
xpct

Is it common for mathematicians to keep a polished version of their work? I'm surprised by how quickly they are able to put work out after it's been contested, and this is 940 pages of well-formatted text, from the looks of it.

show comments
howunfortunate

This feels like a "Move 37*" moment that will go down in the history of AI.

A poetic John Henry style duel between man and machine. Potentially one of the last...

show comments
samayashar

I strongly believe that the reason AI is able to find solutions to these millennial math problems (or any tough problem for that matter) is because of its compute capacity. Our brain is limited in the number of ways that it can think or correlate at a single point in time.

If we can link up 20 different theorems at a time, AI can probably do 2*10^(unknown). If someone finds a way to map a mathematician’s IQ to Neural Net activation for a frontier model and then have a go at these problems, a mathematician would win.

bobmarleybiceps

yikes. I hope this is mostly correct and he gets credit!! :-(

show comments
opendomain

I can’t help but this that AI copied the original work at least in part.

Maybe he had some files in the cloud. Or worse - was listening in.

Is there a way to find out if the human proof was at copied in any way?

show comments
rubikscube09
skeledrew

The race between humans and machines just turned up some more. But some things are inevitable, like the pull of gravity causing things to fall. And the machines eventually winning all races.

show comments
2716057

Anyway, it's time to put Tom Lehrer's Lobachevsky on my playlist.

lf88

woah! If this proof is correct it's likely much more valuable than a giant lean file!

dieselgate

> If any message at all this work is meant to send to the mathematician who feels discouraged by the pace and tone of the current climate, it is this: we are still here. We are very much alive. We are not going anywhere.

Dang...

uptrails

I sincerely hope this is right because I see inherent value in math as a human endeavor. Proving GRH, which we already believed to be true, has no conceivable economic implications: it doesn't improve the productivity of factories, it doesn't let you ship new products, it doesn't allow us to reach the stars. The result matters a lot less than the process of getting there.

Because of this, I find it repulsive that we use this endeavor as a benchmark for AI labs, where they throw tens of millions of dollars just to steamroll a small but prestigious academic field. The damage is already done, but I'm still gonna root for team human. I want us to carve some areas of human activity that LLMs are not allowed to take over. Math, along with art and writing, seems like reasonable candidates.

show comments
elffjs

“If any message at all this work is meant to send to the mathematician who feels discouraged by the pace and tone of the current climate, it is this: we are still here. We are very much alive. We are not going anywhere.”

bilater

I feel for the guy, but what frustrates me is that a lot of people (including, I would guess, probably the majority of the people on this platform) will take this as some sort of an excuse that AI is not getting better. The new cope will be that it somehow cheated and found his solution. The 4 minute mile effect is real.

Supersaiyan_IV

Does this mean that as Alessandro Massulo started using AI's to speed up this process, his findings were leaked through training - causing Claude to find the result for someone else?

show comments
theK

So this is some sort of reaction to the Anthropic Generalized Riehman Hypothesis discovery?

https://news.ycombinator.com/item?id=50574968

eviks

Is there a chance the situation will force math scientists to become more collaborative, publish their work earlier even if unfinished to spread ideas better instead is locking themselves for many years in solitary confinement to try to take all the credit, but lose all the benefits of distributed intelligence?

show comments
earth2mars

I have a feeling this news is waited until Anthropic goes public. Same goes with the Blackhole news. I am not disputing the achievement (I have no credentials for it), but wondering timing of it to hype the stock price

show comments
vismit2000

But what is the Riemann zeta function? Visualizing analytic continuation (3blue1brown): https://www.youtube.com/watch?v=sD0NjbwqlYw

ChaseRensberger

man this is all just so insane

markasoftware

@AcerFur (former openai math employee on twitter) claims there are issues with this proof: https://x.com/AcerFur/status/2123853069599547786

Maybe repairable, but either way, alpoge/anthropic certainly has the claim to the first proof

show comments
origini

Credit to were credit is due, nonethless crazy times that just in such a short time millenium problems have been solved.

vismit2000
dudeinjapan

Masullo is classical and Alpöge is jazz.

In Masullo, primes, zeros, and infinity are different manifestations of an underlying topological structure. Primes are periodic orbits, Archimedean places act as attracting loci, and zeros appear as spectral information. The critical line at 1/2 emerges naturally from the way Fourier analysis connects the additive and multiplicative structures of the number system. This fits into the RH tradition of Weil, Connes, Deninger, etc.

In Alpöge, primes and infinity are tied together by an underlying constraint, which translates into information about the zeros. There's a balance between primes and infinity: you iterate theta lifts m times to generate divisibility at finite primes; the Archimedean contribution makes those lifts expensive, and it turns out that m ~ √X is the optimal way to balance the two. This is alien-tier math: he's effectively manufacturing an analogue of higher-order multiplicity for integers, using iterated theta lifts as a substitute for differentiation.

Masullo may offer the deeper explanation of why RH is true; Alpöge may give mathematics a more powerful tool for discovering new truths.

show comments
lcprk

If true, this could be the last great intellectual achievement in mathematics produced solely by human effort.

Vale!

laszlojamf

is there any hint here that anthropic used the same techniques, like with the navier-stokes equation? Feels awful to have to ask, because Levent seems like a genuinely just talented and interested new kind of mathematician, but it could have been done without his knowledge.

reactordev

So, I like to think of myself as math brained. I keep up with most papers but my first reaction reading this is oh my god…

Oh my god!

It’s beautiful.

choeger

So what does that imply? Whatever someone is currently working on using a computer connected to the Internet is used for training LLMs?

show comments
temp2026nov

Is not this an opportunity to provide someone a locked service? A third-party cloud service who host the model and guarantee that nothing leaves the service. Of course trust and hack are the risks.

yewenjie

Wow, I wasn't expecting this for sure. On the same year a human mathematician attempting full GRH?!

show comments
narmiouh

This is wonderful, go humanity!

taimurshasan

I'm not a math expert nor that great at math but I think RSA is still safe

zajio1am

Do they have a lean proof, or just human-readable version?

mmmaantu

Unfortunately the proof seems to be invalid: https://philarchive.org/rec/NIEIOM

show comments
Igrom

I hope that taking the time to produce a legible write-up instead of releasing the proof up front doesn't bite him back. It would be great if, looking back in history, all contributions to solving the hypothesis were attributed fairly; but that's largely up to mathematicians themselves!

I wonder about the cause and effect. Is this only that this hurried announcement is released in response to Alpöge's publication today, or could Alpöge also have known about another mathematician whose work on proving GRH was close to completion — and felt compelled to release first? Especially given the statement in this draft that "[t]he math is complete".

tgtweak

931 pages... that is pretty crazy.

Madmallard

Anything you do novel or interesting that might earn a lot of money NOT as a registered enterprise (and thus having privacy restrictions on model training in place with your licenses) with but utilizing frontier model services enters into a dangerous game where they are incentivized to steal and publish or otherwise market that business before you.

We need local models NOW.

We need distillation NOW.

blinkbat

Cannot imagine doing anything for any kind of accolade. Truly weird

gizajob

Jeez, you wait ages for a proof of the Riemann hypothesis and then two come along at once. Typical.

show comments
waldrews

A proof by a human? How good could it possibly be? A human can only laboriously recombine ideas from the literature and occasionally inject randomness into the search path. Humans hallucinate so often, report results overconfidently, and often stubbornly refuse to accept counterarguments. For a truly original, creative solution, you still need an AI. Sure, the humans are learning and evolving generation after generation, and can sometimes work effectively in swarms, but ultimately the architectural constraints of being trained only on a finite corpus of three-dimensional data, optimized for survival in the next moment without a complete coherent world model, mean that they're a dead end and will never achieve true intelligence.

show comments
A_D_E_P_T

If this is real, and the results on all sides check out, the author made a HUGE mistake by not publishing a few months ago.

Other mathematicians should take heed. If you've got any ten-year-running projects, take them out of the shadows and into the light, or an Anthropic or OpenAI employee is going to solve your problem when they decide to bother with it and have enough spare compute at hand. And then you'll be scooped, with, at best, a "runner up" trophy.

I'd also add that the time is upon us where hobbyists can solve outstanding problems and prove theorems. The short-lived age of the professional mathematician may be genuinely over.

show comments
ion_trapper

So every big AI math result will now have its “me too”s tagging along claiming they were “almost there” and a crowd of people insinuating that the work was stolen. If I was Anthropic or OpenAI, I would just give the whole math proof thing a rest and focus on things that demonstrably produce value. If you need to prove a math theorem or two to get there, then do it as needed, but not for the sake of just proving theorems. It’s too much drama for very little reward. Fields that are heavily prestige based instead of value based always descend into these meaningless debates about primality of work. After Riemann, there’s not much besides PvNP that would make as much noise anyway.

show comments
QuesnayJr

Yay, humanity is back in the race! (If the proof checks out.)

It's insane though that he and Brian Conrad thought he had a year to polish the paper without getting scooped by AI. He should have put it out whatever he had on October 7th.

nathanfig

I really want the human proof to turn out to be the more correct and elegant proof.

That said, if you published a GRH proof at any point in the past century, someone would have claimed they were just about to solve it. Those tiny remaining gaps often prove to be the most insurmountable ones.

Still rooting for the meat mathematician, though.

efields

When was the last time math had a day that could be scripted into a prestige TV show?

show comments
kidel001

I do like that this one is not hundreds of thousands of pages long. As a non-mathematician this seems ... at least falsifiable in finite time.

reneberlin

And still i hear people saying: "But it only predicts the next character ..."

tornikeo

this is just sad, man

short audit by Astra shows several easy to check errors across the manuscript

human brains aren't cut out for math

genxy

I am morbidly amused that the frontier labs are publicly mauling mathematicians to show how capable their models are.

d0liver

A human proof would be worth much more than one written by AI. It's weird that we're willing to throw a trillion dollars at computers doing math but not at humans doing math. I guess the issue with funding humans is ownership.

xg15

Do we need an on-call rotation for math emergencies now?

hn8726

Will we soon learn that this human-made proof somehow made it to Anthropic pipeline and so their LLM later regurgitated (or at least worked off of) it?

show comments
cximenez

As an academic, can I just say that I for one am happy that, irrespective of the pockets of social mayhem caused by AI rollouts, the acceleration of peer review and publication timelines in academia is a true benefit. I hope more scholars feel compelled to get their ideas out faster now that they might be scooped, like, tomorrow by the next frontier AI model. And I doubly hope that the humanities will speed along quicker, too, not just STEM.

weatherlite

His proof doesnt pass lean, machine won. We lost.

serhei

What the labs have not succeeded in is making mathematics more polite or less stressful.

bonplan23

Thankfully the internet is fast, and we can already cool the (counter-)excitement: https://news.ycombinator.com/item?id=50580084

show comments
v3ai

Oh so everyone just had a proof of RH laying around and didn't want to tell us? Lol hope he did solve if fr fr though, best of luck

xendo

The real question is if Alessandro or Brian used Claude to polish the proof.

fguerraz

I mean it’s great for the field (pun intended) but we knew GHR to be right, it was assumed to be true, so it’s not like it’s going to give us superpowers or anything.

Still an absolute huge achievement, I’m just not sure that on its own it’s very interesting _unless_ new techniques were invented that can be reused to develop new theories.

I feel that it’s what we need the most: new conjectures, not proofs, even if they’re nice to have.

broast

Fun to still be in the stage where humanity is neck and neck with AI

217

We can just say it's proved by Massulo-Claude

Kuyawa

I love it, the days of hiding research in order to get an advantage is over. If you want recognition publish as fast as you can

Welcome to ship fast, break stuff, fix later way of doing things in programming

moezd

I mean, at this rate I will also have a proof of some unsolved problem somewhere. Let me check the old boxes in the attic.

Jokes aside, are we seeing mathematicians FOMO fr?

DeathArrow

Time to learn plumbing.

richard_chase

I wonder if P will turn out to equal NP. I can wait a week.

cubefox

The fact that this is heavily upvoted, completely uncritically, despite the fact that various different frontier LLMs all agree that the proposed proof is invalid because of major holes and errors, and the fact that people who have pointed this out are downvoted, [flagged] and [dead] -- all this speaks very badly about the critical judgement of the HN community. This is shameful group think unconcerned with the truth.

m3kw9

What does Terrance Tao have to say about all this? He's awfully quiet.

show comments
sudohalt

It doesn't matter how "principled" someone is, when it comes to over a trillion dollars no one is "principled".

It is very obvious Anthropic is trying to stabilize their stock. They will stop at nothing to ease concerns of investors, and it is apparent they probably stole results.

IncreasePosts

Is the method for this similar to Anthropic's or independent?

show comments
rvz

Now there are two proposals to the GRH. One assisted with AI and one that is all human with ZERO AI.

Both should be put to the test and all of them should be rigorously reviewed, just like FLT and the poincare conjecture. This one is 900 pages rather than 250K pages.

You can see how you cannot assume anything yet as both are still claims. It starts to not make sense to do so with the top post in this thread [0] apply appeal to a singular authority:

"This is Levent Alpöge/Alessandro Masullo. If he says he has Riemann, well then, he has Riemann."

Rather than that, instead we should wait for multiple authorities to review both of them and the one that satisfies the GRH and is verified as the correct solution is the one that we go with.

[0] https://news.ycombinator.com/item?id=50574968

show comments
airtnp

Given the infamous history of mistakes made by human mathematicians, 90% this is just another wrong paper.

LMFAO for people thinking Anthropic stealing the idea.

TZubiri

non-zero possibility that someone used the AI to review the proof, and that was used to train a later model. Depends on whether the proofs are similar or not.

mi_lk

fuck it. I'm betting on human

jeffrallen

This is not how math, nor really any human endeavor, should be done.

This is a crime against the freedom and beauty of the human brain.

I weep for the sparkling future.

logicallee

has anyone been able to build the lean files for it?

wateralien

Yay humans

perching_aix

The tacit intractability of the gigabyte+ of Lean Alpöge submitted is pointless to deny. It's self-evidently unworkable. It is also very evidently going to be optimized massively. But people who think Claude simply found a way to cheese 4 independent checkers (https://math.mit.edu/~drew/grh/), or a way to misconstrue the task completely, are well into the territory of anti-ai psychosis tier wishful thinking.

Meanwhile, you chuck this proudly-human prose into the most bottom of the barrel model, and it points out blatant omissions and flaws by the dozen, with autoformalizations all having stalled out because of them. The former of which you don't even need to be a mathematician to confirm and appreciate. This is after decades of flawed human proofs being claimed left and right, by cranks and respected academics alike. Come on.

If it's repairable, great. But the optics of this are just sad, and will be profoundly so if the repairs don't work out.

bitwize

It's human-generated slop that's already exhibited several errors. I don't trust it a bit.

poppafuze

At least he'll be able to option the movie rights for one of those almost-real-tech-bro flicks the industry lives to crank out. ...Or the most poignant Veritasium episode ever.

twotwoone

Hmm, guys, do you think this work was stolen by the Anthropics directly or indirectly?

It seems quite likely to me; after all, every piece of text sent to an LLM ends up being used for training...

dom96

So the clear implication is that this proof leaked into Anthropic's training data set, who then used the models which were trained on it to independently create a proof? From memory, previous such mathematical discoveries also came with similar implications.

Has someone confirmed whether those previous discoveries were due to such leaks? If so, it does really seem to portray GenAI as a giant auto complete engine, rather than an intelligence that can deduce novel solutions to tricky problems.

show comments
alabhyajindal

I don’t like it when people release work they say they’ve been on for a long time, right after someone else publishes the result. I don’t care whether the other person used AI. That’s going to continue and we’d better get used to it.

It’s annoying because if you’d been working on this all that time, you could have made some effort to tell the wider community and get at least some recognition. Posting after someone else has solved it is like claiming you had the winning numbers once the draw is already public.

show comments
fhwm

Wait until people realize what ancient Axial Age technology all LLMS/AI operate on regarding 'matrix' math. That's really going to grind alot of people's gears. There's decades of all sorts of people's work as the basis for how LLMs function, they are a big database of people's work over the ages with connect the dot Axial Age Matrix Connection methodologies. Most people and their work will go unmentioned as it always has in history. A very powerful technology that it doesn't seem most at the forefront of it are being honest with what it is. BTW, There's correspondence across scales as large as Galaxies. Might want to look up what ancient societies did stuff like that from that era. $$$$, alot of fools rushing in.

cantalopes

Now this really proves "we don't train on your chat history" is bs for me

show comments
brindidrip

In this age no one will remember your name anyways.

johnwheeler

Nothing more than ego. If you claim to seek the truth, seek the truth. Don't fret if someone finds it before you. I don't care how hard you've worked. The truth is the truth.

show comments
hhh

i was here

twotwoone

It seems that even this piece of evidence was stolen by Anthropic and processed through AI.

It's even a shame that we never got AI, only the theft of scientific papers and insane deception in the stock market.