Large language models like GPT-3 have impressive capabilities, yet their limitations in ethical decision-making underscore the risks to performance.
