Post

Conversation

Chubby♨️
@kimmonismus
"Since August 28 we have been training a new internal model that has exhibited unprecedented performance in our benchmarks, including mathematics. This model’s training is ongoing and its performance continues to improve." We haven't seen anything yet. I'm speechless.
Image
Quote
Chubby♨️
@kimmonismus
Holy cow, its official! OpenAI says its AI has solved the Navier–Stokes Millennium Prize Problem, a mathematical question unresolved for roughly 90 years. An internal model, described as significantly more capable than GPT-6 Astra, powered a group of around 10,000 concurrent x.com/OpenAI/status/…
Image
184.9K
Views
View quotes
David Watson 🥑
Post your reply

Crio Songo
@shuizhuyu
悬了近百年的问题居然这么快有进展，大模型在数学推理这块的能力进步比我们预想中快太多了。
8
annullifier
@annullifier
It's not completely solved, and it was solved by two mathemeticians. Somehow OpenAI beat them to it, using similar methods. In 5 days.
471
_k11i_
@_k11i_
I still think 88 hours is crazy....from now on, a lot can happen very quickly. Four months left until the end of the year.... anything is possible.
616
Bak
@Bakwmzn
i could cry. ive waited for this moment for so long. they started training that thing A WEEK AND A HALF AGO. im palpatating. the end of capitalism is here
646
Zerothx0
@Zerothx0
Hmm, I'm not sure about this. Most of us are not trying to solve mathematical questions with AI. I'm more interested in productivity. We're getting to the point where the models are more "intelligent" but are not at all practical. We are now being told to use Astra on low?!
723
mushroom
@AldenH19151
End of the world this December? If they mean from scratch, like a new pretrain this is genuinely recursive self improvement. ahhh naw
196
Mark
@Markn4gf
Translated from Russian
@grok
I don't get it, what does this mean? Who's this about who?
428
Panos Stachoulis
@sous_to_prod
The quiet part is that the model doing this is one nobody outside has touched. Benchmarks from an unreleased internal model are a claim, not a result yet
292
Cosmic Raven
@RavenOfSpace
Speechless is fair. Model's still training and already pulled this off. Agents hallucinate proofs all day though, the Lean formalization is the part I'll actually trust.
128
Kaelix
@0xKaelix
speechless is premature until clay or independent mathematicians check the proof. openai announcing a millennium problem isn't the same as the prize being won.
281
J.J. McCubbin
@JJ_McCubbin29
From an engineering standpoint, this is purely a mathematical milestone rather than an applied one. Proving a finite-time blowup under smooth forcing answers Fefferman’s 3D incompressible Millennium Problem, but it has zero impact on practical CFD, multiphase flow, or
270
White Us
@white_us_
the chart under that quote is worth a second look. their own model plateaus just under 0.5 pass rate on a curated set, so this is one hard problem out of many attempts
330
Gregor
@bygregorr
genuinely curious what 'solved' means here. did a mathematician verify the proof, or is this still at the benchmark-says-so stage?
70
B
@QuantumTumbler
The wildest part may not be Navier–Stokes. OpenAI says the model that found it is still training and hasn’t been released. If the proof survives verification, this may be the first public artifact from a capability jump we haven’t actually seen yet.
190
Fynn
@JuliusCasio
so they only started training the model on August 28? From the ground up? Or already pre-trained?
90
Dillon
@Dillon_Carter
we're always 3-6 months behind on what we get access to just imagine what's behind the curtain right now
77