handoflixue comments on AI box: AI has one shot at avoiding destruction - what might it say? - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (354)
Of the two AIs I haven't killed, both relied on producing seemingly huge utility within the first message,so I'd agree with you.
The idea of just stunning the human is also nifty, but only really works if you can "hack" the human before they recover, or are providing enough utility that they'll still believe you're friendly afterwards.
So basically the two routes thus far are "hack the human" and "provide a huge utility boost" :)