Comment by IanCal
5 hours ago
His point is twofold: that the process of solving the problems leads to more than just solving the problem in front of you but other interesting things (he has an example of going on a hike to a waterfall and all the other things you might spot over in the distance or nearby on the way, which you’d miss if you were able to jump straight there), and also lots of the simpler open problems are ones early researchers learn on (this is akin to the “if we automate junior engineers how does anyone learn to be a senior?”).
Ability to jump straight to waterfall doesn't prevent you from taking another hike looking specifically for those other things.
If solving any problem is possible with ai, the job of mathematician is finding new problems, not trying to do what a machine can do better.
As for training, most of it is already done by solving known problems, and no one was suggesting to forget analytical geometry so that solving school geometry problems becomes cutting edge research.