A kill switch may be effective for an agent driven attack or something. But I would think that what ever hypothetical national security endangering threat that was created would be something a deterministic program would handle, not an agent driven thing. So implementing a kill switch would have no effect on the hypothetical deterministic program that was created.
A kill switch may be effective for an agent driven attack or something. But I would think that what ever hypothetical national security endangering threat that was created would be something a deterministic program would handle, not an agent driven thing. So implementing a kill switch would have no effect on the hypothetical deterministic program that was created.
> A kill switch may be effective for an agent driven attack
No doubt. But is it effective against an agent-driven attack?
:)
> Anthropic has restricted access to its cyber tool Mythos on the grounds that it is too powerful to fall into the wrong hands.
Too late. It's already fallen in to Anthrophic's hands.