So does that mean he thinks we'll eventually see the deployment of fully autonomous weapons in combat? I think it's very hard to imagine a world where you physically take the capacity out of the hands of rogue regimes... The technology is so ubiquitous that a reasonably competent programmer could build a crude autonomous weapon in their garage. The idea of putting some kind of nonproliferation regime in place that actually keeps the underlying technology out of the hands of people -- it just seems really naive and not very realistic. I think in that kind of world, you have to anticipate that there are, at a minimum, going to be uses by terrorists and rogue regimes. I think it's more of an open question whether we cross the threshold into a world where nation-states are using them on a large scale.
And if so, I think it's worth asking, what do we mean by"them"? What degree of autonomy? There are automated defensive systems that I would characterize as human-supervised autonomous weapons -- where a human is on the loop and supervising its operation -- in use by at least 30 countries today. They've been in use for decades and really seem to have not brought about the robopocalypse or anything. I'm not sure that those [systems] are particularly problematic. In fact, one could see them as being even more beneficial and valuable in an age when things like robot swarming and cooperative autonomy become more possible.