Meta Unveils Self-Taught Evaluator AI Model

• Meta has released a new AI model, the 'Self-Taught Evaluator', aiming to reduce human involvement in AI development.
• The model uses a 'chain of thought' technique, similar to OpenAI's o1 models, to make reliable judgments about other models' responses.
• The model was trained on entirely AI-generated data, eliminating human input at that stage.
• This development could lead to autonomous AI agents that learn from their own mistakes.
• Unlike other companies, Meta has made their models available for public use.

Source: Fast Company