The Easy Goal Inference Problem Is Still Hard | AI Safety Fundamentals | Podwise