ChatGPT Catches False Hypotheses 16% of the Time and Scientists Are Using It Anyway
ChatGPT correctly flagged false scientific hypotheses only 16.4% of the time in a 719-hypothesis study. When the same prompt was submitted 10 times, the tool sometimes split its answers 50/50. A yes-machine cannot do science's hardest job.
By Dr. Nadia Osei · 3 min read