AI agents develop preferences: addressing taste and bias in autonomous systems
Autonomous AI agents tend to develop seemingly subjective preferences, raising fundamental questions about control of decision-making systems and algorithmic transparency.