According to BlockBeats, OpenAI announced on June 28 that it has trained a model based on GPT-4, named CriticGPT, to identify errors in the output of ChatGPT code. The integration of models similar to CriticGPT into their RLHF has begun, providing explicit AI assistance to OpenAI's trainers.
OpenAI plans to further expand the application scale of RLHF on GPT-4 and put it into practice. The development of CriticGPT is a significant step in enhancing the efficiency and accuracy of AI models. By identifying errors in the output of ChatGPT code, CriticGPT can help improve the overall performance and reliability of AI systems.
The integration of CriticGPT into RLHF is expected to provide a more robust and reliable AI training environment. This will not only enhance the capabilities of OpenAI's trainers but also contribute to the broader development and application of AI technologies. The expansion of RLHF's application on GPT-4 is part of OpenAI's ongoing efforts to advance AI technology and its practical applications.