While the killer robots definitely seem scary, what’s more scary to me is the possibility of an AI that can learn indefinitely.
A killer robot on it’s own could kill hundreds of people. A unbounded AI as I described above has the potential to do anything at a scale never seen before. It could mass produce killer robots and improve the killer robot design flaws etc. It can read everything on the Internet and make connections – things we’d never even contemplated. This could be amazing for mankind if we could in some way control such a process, but I think we’d ultimately be fcuked and the best approach would be to avoid this completely.
There are a few people on here who are a bit blasé about this as I guess it’s difficult to imagine or comprehend, but I believe if one of these things gets loose it would be very difficult to stop (cut the power everywhere all over the world??).
Just because something is not possible now doesn’t mean it won’t be in a few years. A human brain is just a chemical bath that’s immensely complicated, it’s not magical or anything … it can be replicated for sure, and probably will be in our lifetimes (immediately before we are all killed by our robot overlords).
TL;DR: I think Elon Musk is right FWIW.