Tackling hallucinations: MIT spinout teaches AI to admit when it’s clueless

A new venture, spun out of MIT, is developing artificial intelligence systems designed to recognize and admit when they lack sufficient information. This innovation aims to combat the growing problem of AI hallucinations, where models generate incorrect or fabricated information, often presented with high confidence. The technology seeks to make AI more reliable by enabling it to express uncertainty, similar to how a person might acknowledge they don't know an answer.

This development is crucial for increasing the trustworthiness and safety of AI systems as they become more integrated into critical decision-making processes and information retrieval.