I used to believe exactly this, and I think this is mostly correct. There are exceptions though. The one that made me reconsider this proposition was Andrew Ng's early work. "New Fangled" AI to precisely control a helicopter! [1] Can't really accept mistakes or imprecision in the outcome. A lot of it was RL work [2]
[1]: https://www.youtube.com/watch?v=M-QUkgk3HyE&ab_channel=Stanf...
[2]: https://papers.nips.cc/paper/2003/hash/b427426b8acd2c2e53827...