Meta has announced the release of a new AI model called “Self-Taught Evaluator” that aims to reduce human involvement in AI development.

This tool, first introduced in an article in August, uses the “thought chain” technique, which mirrors the approach of OpenAI’s o1 models, to improve the reliability of AI judgments.

The “thought chain” method, which involves simplifying complex problems into smaller, logical steps, is said to have shown promise in increasing the accuracy of AI responses, particularly in complicated areas such as science, coding and mathematics.

Meta researchers have taken a significant step by training the evaluator model exclusively using AI-generated data, bypassing the need for human input at this stage of development.

Using AI to evaluate other AI models is intended to provide insight into the future of creating autonomous AI agents capable of learning from their own mistakes.

These self-improving models are intended to eliminate the current need for Reinforcement Learning from Human Feedback (RLHF), a process that is both costly and inefficient.

RLHF requires human annotators with specialized skills to label data and confirm the accuracy of complex mathematical and written solutions.

Metaresearcher Jason Weston: “We hope that as AI becomes more superhuman, it will get better at checking its work, so that it will actually be better than the average human.”

He emphasized the importance of self-study and self-evaluation to achieve unprecedented levels of AI competence.

Recently, Meta collaborated with Hollywood company Blumhouse, known for producing horror films, to test their generative AI video model Movie Gen.

“Meta Introduces New AI Self-Assessment Model” was originally created and published by Verdict, a brand owned by GlobalData.

The information on this website has been included in good faith for general information purposes only. It does not constitute advice on which you should rely and we make no representation, warranty or guarantee, express or implied, as to its accuracy or completeness. You must obtain professional or specialist advice before taking, or refraining from, any action on the basis of the content on our website.