AI research organization OpenAI has introduced CriticGPT, a new GPT-4-based model designed to identify and correct errors in code generated by ChatGPT. 🤖💻 CriticGPT is trained to generate criticism that points out inaccuracies in ChatGPT responses. It is one of the GPT-4 series models that improve the usability and interoperability of ChatGPT.

As large AI models like ChatGPT increase in intelligence and behavior, their accuracy improves and their errors become more subtle. This creates problems for AI trainers who must identify inaccuracies during RLHF comparisons. OpenAI emphasizes that the CriticGPT model aims to solve this problem by providing a means to evaluate the results of advanced AI systems.

OpenAI also announced a partnership with Oracle Cloud Infrastructure to improve its capabilities. Through this collaboration, it will use OCI Supercluster to train and infer its next AI models.