DeepMind’s New AI Looked At 100,000,000 Examples!

📝 The results are available here:

📝 My paper on simulations that look almost like reality is available for free here:

Or this is the orig. Nature Physics link with clickable citations:

🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
Alex Balfanz, Alex Haro, B Shang, Benji Rabhan, Gaston Ingaramo, Gordon Child, John Le, Kyle Davis, Loyal Alchemist, Lukas Biewald, Martin, Michael Albrecht, Michael Tedder, Owen Skarpness, Richard Sundvall, Taras Bobrovytsky, Ted Johnson, Thomas Krcmar, Tybie Fitzhugh, Ueli Gallizzi.

[archer-sh]

This will aid mathematicians in working on multiple approaches to a single problem which will save a lot of time. Very excited for the next 10 years for science.

[kellymoses]

LLMs combined with theorem provers is really exciting because you can tell when they are "hallucinating".

[CodexPermutatio]

Most people have no idea of ​​the implications of this technology.

This is the beginning of something very big.

[Jimmy-pn]

This is definitely what AI/LLM should be made for.
To be used as a tool to speed us along or to teach us new ways of thinking.

[David-ct]

How crazy that this is already "old news".

[waarschijn]

3:16 introducing Timothy Gowers as "a gold medalist" is pretty funny. He is one of the top mathematicians alive.

[DreckbobBratpfanne]

This is the first version of this more general system. Remember AlphaFold? It's second version was a totally different beast. Or AlphaZero? It's second version was 1000x stronger and more general than AlphaGo.
What will this things #2 be like?

[BotanicalBasil]

In two more papers? The paper won't be written by a human. Two more papers after that, it won't be read by a human.

[UnderDogTIGER]

Two more Papers down the line they will give solutions a human can’t understand. Maybe it will look wrong to us but it’s right.

[AdvantestInc]

The leap from struggling with math to solving complex olympiad problems with AlphaGeometry 2 is remarkable. It’s fascinating to see AI crossing new thresholds in mathematical reasoning!

[treicy]

riemann hypothesis will be solved by a computer

[shadowdragon]

I believe that logic models are the key to AGI because math and logic is the basis to understanding reality.

[KearnuPhoenix]

Reminds me of Tony Stark... "JARVIS: Invent timetravel" *leans back and sips coffee*

[QEDAGI]

Weave, oh, weave... That's what you said. Almost tripped trying to rewind to hear that clearly.

[nmadcre]

"Mathematics is the language of nature" - Pi Movie
(hence this performance will unlock a ton of AI features not directly linked to mathematics ?)

[ethanlewis]

One person on this topic said there was a human-assist involved for every single question, and that really makes it a LOT less impressive if true and really does put an important asterisk on this accomplishment!

[alinasri]

It's not only mathematics, it's the mechanism to solve all problems of humanity! That's because all problems are essentially problems of logical reasoning. The AGI will not come from GPT5 or scaling law or whatever, but from the successor of AlphaGeometry and AlphaProof!

[TheMidnightMinute]

Continual improvement of this tool will allow us to approach math concepts which humans have yet to comprehend. If math is the language of the universe, then this could be the key.

[soen]

fields medal 15k dollars? looks like math cutting edge science well respected in society.

[bzikarius]

— Hey, medbot, did you reanimate this man?
— Yes, 93% reanimated!
— So he is dead, right?
— Nooo. He is alive. Only his brain dead. As I said 93% of success! Success! Success!