In our @NatureMedicine article, we discuss how GPT-5 marks a leap in #LLM performance — yet also highlight the risk of the "illusion of understanding": when the model remains statistically driven, and is not truly reasoning: nature.com/articles/s4159…
#ai #research @StanfordDeptMed
#ai #research @StanfordDeptMed