Meta, the parent company of Facebook, announced on Friday the release of new AI models from its research division, including a ‘Self-Taught Evaluator’ aimed at reducing human involvement in AI development.
The model uses AI-generated data for training, bypassing human input at that stage, and applies the ‘chain of thought’ method, similar to OpenAI’s recent models, to improve accuracy in solving complex problems in fields like science, coding, and math.
This innovation could pave the way for autonomous AI systems capable of self-improvement and learning from their own mistakes, potentially replacing costly methods like Reinforcement Learning from Human Feedback (RLHF).
Meta’s researchers highlighted the potential for AI to self-evaluate, improving beyond average human performance. In contrast, other companies like Google and Anthropic are also exploring AI feedback systems, but typically do not make their models publicly available.
Meta also released updates to other AI tools, including image-identificati
on models and datasets to assist in discovering new materials.
Source: Voice of Nigeria