Meta Platforms has launched a series of new AI products, including a model that can verify the accuracy of other AI models, thus reducing human involvement in the development process.
Meta’s New AI Solutions
Meta has introduced new AI models following its August report that details the use of a 'chain of thought' technique similar to that used by OpenAI. These models can evaluate the accuracy of responses to complex problems by breaking them down into simpler logical steps.
Benefits of Self-Evaluating AI
Meta aims to tackle inaccuracies faced by AI models like ChatGPT. Researchers have hinted at building autonomous systems capable of learning from their mistakes. Researcher Jason Weston emphasized the importance of self-taught and self-evaluating models in reaching super-human AI levels.
Towards Autonomous AI
Researchers suggest that self-improving models could eliminate the costly process known as Reinforcement Learning from Human Feedback. Other companies like Google and Anthropic are researching AI feedback concepts, though they often do not release their models for public use. The use of AI to check AI may be crucial in developing autonomous AI applications.
Meta is moving forward with creating autonomous AI models capable of operating without human intervention, potentially leading to self-correcting models that learn and improve independently.