Of course. "tried to" being key words in the comment. If he had the help of Claude at the time, how much more dangerous would his bumbling have been?
A real nuclear engineer with the knowledge he needed would also have said "no, don't do that and I won't help you." We are programming the knowledge into the ai agent. Giving ai a little discretion makes sense too.
A real nuclear engineer with the knowledge he needed would also have said "no, don't do that and I won't help you." We are programming the knowledge into the ai agent. Giving ai a little discretion makes sense too.