What problem is AI actually solving?

For me, it deals delightfully with meaningless problems that just involve technical work, that is, grunt work that I hate to do: especially coding in Matlab, but also things like re-ordering a bibliography according to author name rather than number, and adding brief summaries if need, etc. It’s like having a slightly dumb but totally diligent assistant.

I recently read an article about AI and mathematics - which is a hot topic now, because it has solved some thorny mathematical problems. It can, of course, do the grunt work a lot faster than people can. But there’s another aspect.

A human mathematician may be an expert in his own field and in a couple of other fields, but a good AI machine will have knowledge of lots of other fields as well. Some math problems have been solved by borrowing something from a completely unrelated field. If I remember aright, that is how Wiles solved the Fermat problem. That’s its advantage over even the best mathematicians, who can’t possibly master every single field of mathematics.

Its disadvantage, of course, is that people really are more creative, and will be finding and thinking about new problems. If it hasn’t been discussed in the past, AI isn’t going to know about it.