1. Researchers aim to address AI hallucination issues caused by words with multiple meanings. 2. They propose using semantic entropy, a measure of confusion due to ambiguous words, to assess the likelihood of AI hallucinations. 3. This approach could help improve the accuracy and reliability of AI systems.