Famous math problems in machine learning and AI ?

[Stephen Tashi]

"famous" math problems in machine learning and AI ?

Are there any "famous"(i.e. crucial) unsolved mathematical problems in the field of machine learning or the general study of artificial intelligence?

There is interest in programming computers to solve problems by artificial intelligence and there are certainly mathematical results published in such studies, but I can't think of any practical goal in AI that hinges on solving a crucial mathematical question.

 

[.Scott]

There are no "famous" ones - not that I can thing of.

In computer programming, it is commonly more important to have a pragmatic solution than an mathematically ideal solution.

It would, of course, be of great interest to determine exactly what algorithms the human brain uses for almost anything - since what a human does is often the success criteria for whether a robot is responding adequately.

With machine learning, there is a problem with trying to replace "design" with a broadly adaptive algorithm. The problem is that you don't really know what the computer is learning. There is actually a "famous" story that goes along with this problem described here:
https://neil.fraser.name/writing/tank/

I can't determine if the story is real or not. I first heard it in the late 1970's, so the "1980's" part is wrong. Also, I know that the DoD was contracting out for this specific type of automated photo-interpretation solution (that is, finding tanks in a forest) - so it's very credible.