This document says: This Darwinian logic could also apply to artificial agents, as agents may > eventually be better able to persist into the future if they behave > selfishly and pursue their own interests with little regard for humans, > which could pose catastrophic risks.
They have no interests any more than a dishwasher does. They have no motives. No instinct of self-preservation. Unless someone programs these things into them, which I think might be a disastrous mistake. I do not think the instinct for self-preservation is an emergent quality of intelligence, but I should note that Arthur Clarke and others *did* think so. An AI in a weapon might be programmed with self-preservation, since people and other AI would try to destroy it. I think putting AI into weapons would be a big mistake.