Brain-Computer Interfaces and AI Alignment — LessWrong