A new study shows that GPT-4 reliably wins debates against its human counterparts in one-on-one conversations—and the technology gets even more persuasive when it knows your age, job, and political leanings. Researchers at EPFL in Switzerland, Princeton University, and the Fondazione Bruno Kessler in Italy paired 900 study participants with either a human debate partner or OpenAI’s GPT-4, a large language model (LLM) that, by design, produces mostly text responses to human prompts. In some cases...