Publishing academic papers on transformative AI is a nightmare — LessWrong