I've been looking into AI research and I've noticed a nontrivial amount of individuals either received a PhD in math or had advisors who received a PhD in math. Some names include (Ben Goertzel, Marvin Minsky, Jürgen Schmidhuber).
I can see how a formal (arguably more difficult) training in math would help tackle issues in AI, but it'd make more sense to do a PhD in AI if one wants to make a career studying AI, no?
In contrast, all the high impact papers for top machine learning conferences (KDD, ICML, NIPS) were mostly from CS labs, but I cannot say that this is the norm in other AI subareas.
Is it actually better to do a PhD in math and do research in a topic that converge with AI?