The trade-offs of AI alignment work at major labs

A commenter argues that major labs offer resources for AI alignment research, but aren’t the only places to contribute. They say labs provide compute and the latest models to accelerate alignment research, plus somewhat misaligned models to study. They also cite Richard Ngo’s view of labs as a “pressure cooker” where moving fast dominates deep, creative thinking. Praising Goodfire, Redwood and METR, the commenter closes with a call to “come solve alignment.”