Even if a state of the art machine learning system is going to work 99.999% of the time - with millions of diffs run daily - that's still too many failures.
So basically with diff we had 1 problem, with ML-diff we'll have two.
So basically with diff we had 1 problem, with ML-diff we'll have two.
There seems to be a serious bias against ML in systems programming that is unwarranted. There are plenty of problems using bad heuristics today that could be improved with ML in a way that doesn't reduce reliability. Scheduling, compiler optimization, stuff like that.