@average_joe_mcc: Turns out that fluid dynamics can become pop math damn #math #navierstokes #ai #stem #LearnOnTikTok

JP :

mathematicians: "hey computer, tell me you solved this math problem" ai: "i solved it" mathematicians: "oh my God"

user2546335350827 :

I thought you were gonna explain the math not give a gossip girl recap wtf

yonderrr999 :

Hot take: OpenAI didnt solve it if they cant prove it to a peer-review-able mathematical standard

Sqordil :

Thank you, this is the only I’ve seen video that has provided the entire context and nuance of the situation

user9017787983156 :

A shameful handling of what should have been a monumental moment. This is not how science should work, and sends a poor message to academic users especially on OpenAI’s platform

Fafner :

"This academic paper is sponsored by OpenAI, your number one sorce for plagiarism." Also, they may have solved the problem, but did they ever understand it? Yeah, you got a few hundred thousand lines of code telling you the logic works, but that's about as useful as saying 42 is the answer to life, the universe and everything.

limealive :

These are the exact kinds of reasons why AI just needs to go man. Legit so disappointing :/

Eric Varela :

Oh this is so much messier than I first thought 🍿

user3757872997887 :

Oh man that's so good tea. What a crazy story of "we couldn't do this without capitalism, but also capitalism ruined it"

Gary Bussy :

OpenAI stole their work

Flores :

so did they actually solve it or not

Mike :

This explanation would be better if you explain that Navier-Stokes is a way to calculate fluid flow over time given perfect knowledge of it's starting state. The question was are their conditions where it "blows up" meaning you get infinities (which are physically impossible). Particularly what was sought was either proof that Navier-Stokes never blows up or proof that you sometimes get infinities. The stepping stone problem is using the Euler simplification of Navier-Stokes without friction (zero viscosity).

ifiknewhow :

okay that was such a good explanation of the situation thank you

Wyatt :

davit_math :

https://zenodo.org/records/22694089

sma :

this is so depressing imo

rainyzen_ :

finally, someone knows what they are talking about

HappyCrow :

Thank you for the excellent walk-through! One thing I find funny is that it has long been the wish that private companies take more interest in research mathematics, and now OpenAI did exactly that and spent millions on it in a very few days - except didn’t try to solve the full NS for the sake of mathematics, but rather to prove that their model was superior. My guess is that when the novelty of AI being good at math has died down, and no longer will be used as a benchmark of AI capability, then we will see much less of such efforts.

Andy Ma :

Do we trust Lean? I don't really know how Lean code works, but I feel like I've heard examples where Lean gets it wrong. For example, could it be that somewhere in the Lean code there's some kind of error or bug that was not known and that allows them to get the proof? In my mind, I think the jury is still out on this problem until the math community can verify its correctness. I'm just still a little skeptical because the analysis of Navier-Stokes is very delicate

Kimo :

what are these reddit names bro

Azura210 :

wasn’t it found that specific chats relating to specific pieces of information and code could be located in a foundation model ? for example, they found out recently that the open source Stable Diffusion foundation model had 1000s of links to real CSAM. so in theory they may have the capability to find out if those guys chats/code/research that they used ChatGPT with was used in OpenAI’s 88 hr process

michaelwright9749 :

it is c and d so its only example of a negation case also it had no tools generated so its a waste of time

michelle ⏾ :

as a physicist I’ve been keeping up with this whole thing and it just makes me sad