Terence Tao a month ago:
"The AI companies have clearly... I'm sure they have secretly devoted much more time to the Riemann hypothesis than what they've disclosed. And they couldn't touch it. I mean something beyond current AI is needed. So maybe some combination of humans and AIs in 2050 or something will make further progress. Who knows?"
It seems bold for Tao to state that they couldn't touch it; Considering that after Anthropic announced Navier-Stokes and got all the negative press, largely driven by Terrence Tao, it doesn't surprise me that they didn't disclose anything afterwards.
I think a lot of his AI skepticism is just a self-fulfilling prophecy, and a misunderstanding that AI at the cutting-edge is a tool in the hands of the user. In other words, sure current AI is not going to solve the Riemann hypothesis until 2050, but what do you think the current state of AI would be if Terrance Tao was the one driving it?
Somewhat related: I dont think AI is coming for everyones jobs, and I will never understand the argument that it is, because I always have the ability to use the AI and add my "human ingenuity" to it. So why _wouldnt_ that be a better option than just raw AI?
On Anthropics side it's literally just levent. He isn't even doing this on an official capacity. OAI have teams of Mathematicians. This is like david vs goliath.
Direct link to the quote: https://youtube.com/watch?v=bAnwMdfcMVw&t=1035s
"I'm sure they have secretly devoted much more time to the {{matter}} than what they've disclosed. And they couldn't touch it. I mean something beyond current AI is needed. So maybe some combination of humans and AIs in 2050 or something will make further progress. Who knows?"
- All of humanity, unaware their pants have fallen around their ankles
In 24 years with a 24 year margin of error
> Number theorists are going to be busy the next decade and a half digesting this proof.
- someone else
So maybe not too far off.
[flagged]
>I mean something beyond current AI is needed.
Every time someone says 'current AI is not enough, we need new architectures', it turns out that current AI was enough and all we really needed was more scaling.