Many open mathematical problems of form "There are no numbers satisfying the following criteria..." were resolved when computers became fast enough to perform arbitrary-precision arithmetic at decent speed. Some folks got excited at the time - but it was soon obvious that 99% of mathematics was not usefully accessible to number-crunching approaches. Will modern AI's similarly prove poor at cracking a large percentage of hard math problems?
How much is this situation about AI, and how much is it about throwing AI-bubble $billions at math problems? Lead times would be longer, but you might see incredible progress in astronomy, physics, or a number of other fields, if the richest people in the world were throwing blank-check budgets at building space telescopes, or atom smashers, or whatever.
Two points:
Many open mathematical problems of form "There are no numbers satisfying the following criteria..." were resolved when computers became fast enough to perform arbitrary-precision arithmetic at decent speed. Some folks got excited at the time - but it was soon obvious that 99% of mathematics was not usefully accessible to number-crunching approaches. Will modern AI's similarly prove poor at cracking a large percentage of hard math problems?
How much is this situation about AI, and how much is it about throwing AI-bubble $billions at math problems? Lead times would be longer, but you might see incredible progress in astronomy, physics, or a number of other fields, if the richest people in the world were throwing blank-check budgets at building space telescopes, or atom smashers, or whatever.