AI Safety Researchers Debate Motivated Cognition Standards

Richard Ngo expressed concern that AI labs exhibit strong motivated cognition favoring rapid development, making claims of genuine belief in safety positions unreliable under low evidentiary standards. Geoffrey Irving countered that describing such positions as real beliefs remains accurate even amid motivated reasoning, rejecting stricter thresholds as unnecessary for factual description. The exchange centers on how to characterize lab statements about AI risks and racing dynamics without downplaying underlying incentives.