Davidmanheim comments on Language Models are a Potentially Safe Path to Human-Level AGI