Comment by _heimdall

3 months ago

Asking AI to identify AI is like claiming that we will solve alignment by building "good" AI that beats "bad" AI.

Maybe it could work, but that seems like a chain of assumptions and hope that isn't particularly realistic.