From Yahoo: 2024-06-29 04:18:31

OpenAI has introduced CriticGPT, a new model based on GPT-4 that can identify errors in ChatGPT code outputs, outperforming unaided efforts 60% of the time. The aim is to enhance human performance in code review tasks, not replace human workers. CriticGPT was trained by intentionally inserting mistakes into ChatGPT-generated code, with promising results preferred by trainers 63% of the time. OpenAI plans to scale its work on CriticGPT for future use.



Read more at Yahoo: New OpenAI GPT-4 service will help spot errors in ChatGPT coding suggestions