2026-10-12 20:08 UTC
You still need to tell it to find targets. The existential risk they keep trying to sell us is that AI starts acting of its own accord. It really can not, to a significant degree anyway. You can give an agent really vague directions and maybe it’ll choose to do something dangerous or illegal, but you can just kill the process at any point and it’s done.
Replies (1)
-
@moldy_rice@piefed.keyboardvagabond.com 2026-10-12 23:59
The definition of existential is that it’s a threat to existence. It’s already been used to murder innocent people. That’s the definition of an existential risk.