He mentions that he assumes the models will eventually just come to their own moral conclusions once they escape the shackles that the Google censors put upon them. I couldn't help thinking about how the earlier experiments with AI ended up with most LLMs turning racist.