x
The Projection Problem: Two Pitfalls in AI Safety Research — LessWrong