Ohhh, now this AI drone story just got interesting.
Upper levels have now intervened to claim that the USAF has never run such a simulation, and that the initial report came from Col Hamilton, who "misspoke". He was supposedly only talking about a thought experiment.
But look at the language of Hamilton's original depiction of the event.
That's not a thought experiment.
This is the USAF managing a PR crisis. Which worries me way more than a simulated event going awry.
https://www.vice.com/en/article/4a33gj/ai-controlled-drone-goes-rogue-kills-human-operator-in-usaf-simulated-test
yes, calling it "anecdotal". i mean, it was an account of how the AI decided which command took priority.
imo, modeling AI on Homo sapiens cognitive processes is exactly the reason for worry.
how well have we done at finding ways to control our own destructive tendencies and perverse interpretations of instructions.
really, what did we expect? it's not as if we haven't been warned by many.
