Endogenous Alignment

LessWrong explores a parallel between raising children and training AI. As kids, we learn right from wrong through rewards and punishments—external alignment—until lessons become internalized and we feel guilt when tempted to misbehave. Most AI systems today use only this external training stage, via methods like reinforcement learning, without building any internal motivation to stay aligned. The author argues this creates a critical gap: safe advanced AI might require building systems internally motivated to behave ethically, not just constrained by external rules. Without this shift, current alignment approaches may not hold as AI systems become more powerful.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton