Corrigibility
9 pages tagged "Corrigibility"
Might an aligned superintelligence force people to change?
Is it possible to limit an AI's interactions with the Internet?
Could we program an AI to automatically shut down?
Why would we only get one chance to align a superintelligence?
Why can't we just turn the AI off if it starts to misbehave?
What is the Center for Human Compatible AI (CHAI)?
What is "Do what I mean"?
What is corrigibility?
What is "Constitutional AI"?