How to Build Safe AI (Without Making the AI Safe)

Interesting. Even tho I think there is another path that exists but it’s being ignored. Accuracy cannot be expected to result from generative and probabilistic machines especially when they programmed to retain attention instead of accuracy.