The Long (Self-)Correction

Researcher Wei Dai has published an essay on the AI Alignment Forum proposing the Long Self-Correction as an alternative to AI Pause and Long Reflection. Rather than pausing development because we haven't reflected enough on the risks, Dai argues humanity itself is fundamentally unprepared to build or oversee powerful AIs—not for lack of thought, but because we're flawed in ways reflection alone can't fix. He catalogs multiple vulnerabilities: no workable moral framework to guide us, incompetence at long-term strategic thinking, false confidence in our wisdom despite overwhelming evidence to the contrary, and how status concerns quietly corrupt our reasoning about humanity's biggest decisions. Dai's pitch: humans need a sustained, uncertain process of moral and philosophical self-improvement before we're ready to build powerful technologies. His hope rests on the fact that humanity has historically made progress on these deep flaws—suggesting that if we preserve the conditions for that progress and prevent permanent derailment, we might eventually be ready to reshape the world according to our actual values.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton