Announcement
New paper argues preference-based AI alignment has moral principles baked in
The authors argue that philosophical reflection can help resolve clashing principles, rather than relying on preferences alone.
TLDR
In “Between Ideal and Non-Ideal,” @xuanalogue and Henrik Kugelberg argue that preference-based AI alignment is not free of moral assumptions. They say principles already shape how people judge AI behavior, and ideal theory can help untangle cases involving conflicting principles or harms. Their broader question: what role should AI serve in a just society?
Combined views
1.6K
3 Sources, first seen ago
Comments