Aren’t there some pretty easy ways to eliminate these potential problems?

It might look like there are straightforward ways to eliminate the problems of unaligned superintelligence, but so far all of them turn out to have hidden difficulties. There are many open problems identified by the research community for which a solution would be required for success in alignment.