@lzg@mastodon.social Choose the most outrageously outlandish interpretation of what they said when responding to them, and maybe you'll train them the way LLMs train their interlocutors.