Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

OpenAI claims gold-medal performance at IMO 2025

twitter.com

Post reply on HN
Print view View original

651–660 of 737 posts

Re: OpenAI claims gold-medal performance at IMO 2025

#651

Post by taneq »

krackers wrote: post #589
torginus wrote: post #580

Earlier quoted context omitted.

In transformers generating each token takes the same amount of time, regardless of how much meaning it carries. By cutting out the filler from the text, you get a huge speedup.

Except generating more tokens also effectively extends the computational power beyond the depth of the circuit, which is why chain of thought works in the first place. Even sampling only dummy tokens that don't convey anything still provides more computational power.

Dummy tokens work for humans too. “Shhh I need to think!”

Re: OpenAI claims gold-medal performance at IMO 2025

#652

Post by ksec »

asadotzler wrote: post #339
ksec wrote: post #87

I am neither an optimist nor a pessimist for AI. I would likely be called both by the opposite parties. But the fact that AI / LLM is still rapidly improving is impressive in itself and worth celebrating for. Is it perfect, AGI, ASI? No. Is it useless? Absolutely not. I am just happy the prize is so big for AI that there are enough money involve to push for all the hardware advancement. Foundry, Packaging, Interconne…

But unlike the trillion dollars invested in the broadband internet build out between 1998 and 2008, when this 10 year trillion dollar bubble pops, we won't be left with an enduring and useful piece of infrastructure adding a trillion dollars to the global economy annually.

>we won't be left with an enduring and useful piece of infrastructure adding a trillion dollars to the global economy annually.

Nearly all colleagues I know working inside a very large non-tech organisation are using Copilot for part of their work in the past 12 months. I have never seen tech adoption this quick for normal every day consumer. Not PC, Not Internet, Not Smartphone.

I actually had discussions with parents about our kids using ChartGPT. Every single one of them at school are using it. Honestly I didn't like it but they were actually the one who got used to it first and I quote "Who still uses Google?". That was when I learn there will be a tectonic shift in tech.

Does it actually add productivity? may be. Is it worth the trillion dollar investment? I have no idea. But are we going back? As someone who knows a lot about consumer behaviour I will say that is a definite no.

Note to myself. This feels another iPhone moment again. Except this time around lots of the tech people are skeptic of it, but consumer are adopting faster. When iPhone launch a lot of tech people knew it will be the future. But consumer took some time. Even MKBHD acknowledge his first Smartphone was in the iPhone 4s era.

Re: OpenAI claims gold-medal performance at IMO 2025

#653

Post by justinhj »

taneq wrote: post #651
krackers wrote: post #589

Earlier quoted context omitted.

Except generating more tokens also effectively extends the computational power beyond the depth of the circuit, which is why chain of thought works in the first place. Even sampling only dummy tokens that don't convey anything still provides more computational power.

Dummy tokens work for humans too. “Shhh I need to think!”

"Let me be clear..."

Re: OpenAI claims gold-medal performance at IMO 2025

#654

Post by wanderlust123 »

SiempreViernes wrote: post #531
UltraSane wrote: post #497

Earlier quoted context omitted.

Humans who excel at IMO questions are also "fine tuned" on them in the sense that they practice them for hundreds of hours

Sure, but nobody is using their IMO score to prove they are superintelligent and pulling it off in wider groups.

I’m pretty sure that a high score in imo is a sign of high intelligence.

Re: OpenAI claims gold-medal performance at IMO 2025

#655

Post by kevinventullo »

futureshock wrote: post #334
demirbey05 wrote: post #277

Google also joined IMO, and got gold prize. https://x.com/natolambert/status/1946569475396120653 OAI announced early, probably we will hear announcement from Google soon.

Google’s AlphaProof, which got a silver last year, has been using a neural symbolic approach. This gold from OpenAI was pure LLM. We’ll have to see what Google announces, but the LLM approach is interesting because it will likely generalize to all kinds of reasoning problems, not just mathematical proofs.

I’m much more excited about the formalized approach, as LLM’s are susceptible to making things up. With formalization, we can be mathematically certain that a proof is correct. This could plausibly lead to machines surpassing humans in all areas of math. With a “pure English” approach, you still need a human to verify correctness.

Re: OpenAI claims gold-medal performance at IMO 2025

#656

Post by kelipso »

zug_zug wrote: post #597
causal wrote: post #118

Earlier quoted context omitted.

Where did you get this? Don't see it on the 2025 problem set and now I wanna see if I have the right answer

I asked chatGPT. However it's saying that's 2022 problem 5, however that seems to be clearly wrong... Moreover I can't find that problem anywhere so I don't know if it's a hallucination or something from it's training set that isn't on the internet....

Butlerian jihad can’t come soon enough lol

Re: OpenAI claims gold-medal performance at IMO 2025

#657

Post by sebmellen »

SiempreViernes wrote: post #531
UltraSane wrote: post #497

Earlier quoted context omitted.

Humans who excel at IMO questions are also "fine tuned" on them in the sense that they practice them for hundreds of hours

Sure, but nobody is using their IMO score to prove they are superintelligent and pulling it off in wider groups.

I’ve seen IMO rank used to justify more than one $100m+ seed round.

Re: OpenAI claims gold-medal performance at IMO 2025

#659

Post by queenkjuul »

miguelacevedo wrote: post #271
mikert89 wrote: post #133

The cynicism/denial on HN about AI is exhausting. Half the comments are some weird form of explaining away the ever increasing performance of these models I've been reading this website for probably 15 years, its never been this bad. many threads are completely unreadable, all the actual educated takes are on X, its almost like there was a talent drain

Basically this. Not sure why people here love to doubt AI progress as it clearly makes strides

I don't doubt the genuine progress in the field (from like, a research perspective) but my experience with commercial LLM products comes absolutely nowhere close to the hype.

It's reasonable to be suspicious of self aggrandizing claims from giant companies hyping a product, and it's hard not to be cynical when every forced AI interaction (be it Google search or my corporate managers or whatever) makes my day worse.

Re: OpenAI claims gold-medal performance at IMO 2025

#660

Post by gsf_emergency_2 »

famouswaffles wrote: post #491

From Noam Brown https://x.com/polynoamial/status/1946478258968531288 "When you work at a frontier lab, you usually know where frontier capabilities are months before anyone else. But this result is brand new, using recently developed techniques. It was a surprise even to many researchers at OpenAI. Today, everyone gets to see where the frontier is." and "This was a small team effort led by @alexwei_ . He took a resea…

"frontier" seems to be "zapad" for OpenAI

They have a parallel effort to corner Ramanujan called https://epoch.ai/frontiermath/tier-4

(& Problem 6, combinatorics, the one class of problems not yet fallen to AI?)

The hope for humanity is that of the big names associated to FrontierMath (starkly opposite to oAI proper) Daniel is the one youngish nonexsoviet guy :)

Post reply on HN