Meta releases AI model that can check other AI models' work

Meta has released new AI models, including a "Self-Taught Evaluator," which aims to reduce human involvement in AI development.

The model, introduced in an August paper, uses the "chain of thought" technique to break down complex problems and improve response accuracy.

Trained entirely on AI-generated data, it offers a potential path toward autonomous AI agents capable of self-improvement without human input. This could eliminate the costly process of Reinforcement Learning from Human Feedback.

Meta also released updates to its image-identification model and tools for speeding up LLM response generation and discovering new inorganic materials.