Comment by seanmcau

It would/will be extremely irresponsible to put non-deterministic and fallible models in charge of weapons. We are not close to having solved the problem of ensuring AI pursues good outcomes