OpenAI claims to have found a way to prevent artificial intelligence (AI) models from producing erroneous information known as hallucinations. The company has trained a model in solving complex mathematical problems using process supervision, which provides feedback for each step of a problem rather than just on the final outcome. OpenAI found that the process supervision method helped produce significantly better performances, and could prompt the AI models to follow a human process, increasing its interpretability. Nonetheless, OpenAI cautions that the method might not work outside of mathematical problems, but it is vital to extend this study to other domains to ensure better AI models.
OpenAI is a company that seeks to promote and develop safe and ethical AI.
ChatGPT is an AI language algorithm developed by OpenAI capable of generating responses to language prompts.
