[jjcon]

> ML (AI) is, at least right now, fancy pattern matching. Nothing more.

I mean, every problem can be boiled down to some sort of 'fancy pattern matching', the question is really how fancy/sophisticated the solver and how large the problem space the problem. I'm not sure why AI couldn't be helpful here even if the convergence of the solver/problem space are still many years off.

[bawolff]

That's basically equivalent to saying, by the church-turring thesis computers can solve any solvable problem, therefore it can probably solve the problem at hand.

Which is technically true, but as a pragmatic matter doesn't really tell us much about if, when, or how the problem will be solved.

[jjcon]

Exactly - I’m not claiming a specific timeframe for AI to be helpful in this area - just pointing out that the claim that it is, ‘just fancy pattern matching’ isn’t limiting to its utility and that in theory it should be able to contribute here.

[taneq]

s/computers/engineers/g and your post still holds true. Predicting when we’ll solve unsolved problems is just hard.