Problems facing a correspondence theory of knowledge — AI Alignment Forum