OpenAI proposes a new technology PVG to solve the "black box" problem

PANews reported on July 18 that according to Economic Observer, OpenAI's official website said today that with the use of ChatGPT in the fields of finance, law, and marketing, the "black box" situation will occur because its accuracy cannot be verified. In order to solve this problem, OpenAI proposed a new training framework Prover-Verifier Games (abbreviated as "PVG"), for example, using small models such as GPT-3 to verify and supervise the output of the GPT-4 large model, thereby improving the output accuracy and controllability. Recent news shows that OpenAI is making the capabilities of artificial intelligence (AI) large models closer to human-like reasoning levels.