LLMs factor in unrelated data when recommending health-related therapies Scientists uncover nonclinical info in individual messages — like typos, excess white Room, and colorful language — lowers the precision of the AI design. Examine whole story → Far more information on MIT Infor