Hiring engineers and researchers to help align GPT-3 — LessWrong