4 ms·
It's not a question of being edgy[0], it's that we don't have the faintest idea what it even means for an AI to be aligned with our values. Closest we have now
by ben_w 4y ago
It's not a question of being edgy[0], it's that we don't have the faintest idea what it even means for an AI to be aligned with our values.
Closest we have now is following the intuitive ideas that generally work for most humans — giving positive/negative feedback for particularly good/bad answers, and asking nicely in plain language — but last I checked, the people who have made it their mission to seriously consider what it means from first principles consider this to be so unsophisticated they call it "pre-paradigmatic"[1].
[0] unless you're worried about minors getting access and learning what the swear words mean, which is a thing
[1] which was a bit over a month ago, so practically lifetime in AI: https://www.planned-obsolescence.org/disagreement-in-alignment/ https://www.planned-obsolescence.org/disagreement-in-alignme...