As is the potential for instantaneous updating of optimised procedures, which would take years to decades with humans. In the end I'm certain they won't be infallible, but just as with autonomous vehicles they're likely to be many orders of magnitude more reliable than humans.