Safety2d ago
Safety
Critique of AI Alignment Basics
A new LessWrong essay argues that the AI safety community neglects fundamental alignment basics. The author uses a fictional persona to highlight the gap between theoretical safety and actual model behavior. This critique suggests current alignment strategies fail to address basic cognitive biases. Practitioners must reconsider foundational assumptions to avoid catastrophic failure.