zlacker

>>nl+(OP)
I know the problem the AI solved wasn't that hard but one should consider math toast within the next 10-15 years.

Software engineering is more subtle since you actually care about the performance and semantics of a program. Unless you get really good at program specification, it would be difficult to fully automate.

But with math the only thing you care about is whether the moves you made were right.

>>vatsac+7E2
> one should consider math toast within the next 10-15 years.

I keep a list of ridiculously failed predictions, and this goes into it.

Can I ask you to be more concrete? What does "math is toast" mean to you?

>>emil-l+mG2
In the sense that in 15 years my bet is that an AI system can solve my thesis problem for under 100$

https://ems.press/journals/jems/articles/14298293

>>vatsac+AQ5
Math is toast if a computerized system can solve a problem you personally failed to solve during your PhD?

Computers have proved stuff people couldn't since at least 1976 when Appel and Haken proved the 4-color theorem.

>>emil-l+xj6
The paper I linked is the problem I solved for my thesis, which was published in JEMS (upper echelon journal). I spent 8 hours a day for one and a half years working on it. 75% of mathematicians don't have a theorem that good, but the Terry Taos of the world get a theorem like that every year.

My claim is that in 15 years an AI system will be able to prove it from first principles for under 100 dollars. That would render us normal mathematicians toast.