DFNaiff comments on AI misalignment risk from GPT-like systems?