LessWrong contributor Elias Schmied argues that people working on long-term problems—particularly AI safety—should try less hard. His logic: deep uncertainty about whether your efforts help or hurt, combined with intense goal-directed effort, creates a dangerous trap. Push hard enough toward any goal, and motivated reasoning takes over: you rationalize your choices and lose the mental space to question them. The risk is especially acute in AI safety work, where interventions could backfire—accelerating capabilities, increasing geopolitical tension, or creating adversarial relationships with future AI systems. The argument isn't for laziness; it's for maintaining enough intellectual humility and cognitive slack to course-correct when your assumptions prove wrong.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton