Oxford researchers have seemingly found a solution for AI hallucinations. Semantic entropy can determine the difference in the meanings of the outputs generated, making it easier to identify false information presented as facts.