GPT-4 passes Turing test, researchers claim

Researchers claim that GPT-4, the AI behind ChatGPT, has passed the Turing test, which measures if a machine can mimic human conversation.

In the study, 500 people chatted with four respondents: one human, GPT-3.5, GPT-4, and an old AI program called ELIZA. Participants then guessed if they were talking to a human or an AI.

GPT-4 was mistaken for a human 54% of the time, while ELIZA fooled people only 22% of the time. GPT-3.5 scored 50%, and the human was correctly identified 67% of the time. This indicates GPT-4 can effectively mimic human conversation.

Experts warn that AI being seen as human can have social and economic impacts. They also say the Turing test might be too simple, as it doesn't account for the social and emotional aspects of human communication. The study highlights the progress of AI from simple programs like ELIZA to advanced models like GPT-4.