OPENAI'S NEW “CRITICGPT” MODEL IS TRAINED TO CRITICIZE GPT-4 OUTPUTS

Summary from the TLDR newsletter

CriticGPT is a new AI model from OpenAI designed to identify mistakes in code generated by ChatGPT. Based on the GPT-4 family of large language models, it analyzes code and points out potential errors, making it easier for humans to spot mistakes that might otherwise go unnoticed. CriticGPT was trained on a large number of inputs containing deliberately inserted mistakes. Its capabilities extend beyond just code review - it is able to identify errors made by ChatGPT in some other tasks as well. OpenAI plans to integrate CriticGPT-like models into its RLHF labeling pipeline to provide its trainers with AI assistance.