[Intro to brain-like-AGI safety] 3. Two subsystems: Learning & Steering — LessWrong