ChatGPT: these are not hallucinations – they’re fabrications and falsifications
Robin A. Emsley
Abstract
The artificial intelligence (AI) system, Chat Generative Pre-trained Transformer (ChatGPT), is considered a promising, even revolutionary tool and its widespread use in health care education, research, and practice is predicted to be inevitable 1 . Like so many others I was keen to test the capabilities of ChatGPT as an aid to scientific writing. An opportunity arose with a study I was planning on an existing dataset (structural MRI brain changes associated with antipsychotic treatment). After registering with OpenAI online, I provided basic information on the dataset and planned study and requested suggestions for the study methodology. These were promptly provided as broad stroke proposals about identifying a research question, hypotheses, suitable outcomes etc. The interaction felt eerily interpersonal and the chatbot’s demeanour was cordial and helpful, even eager. My questions relating to overall methodology produced mostly sensible suggestions, although largely predictable and somewhat mundane. I had difficulty in getting beyond the general responses that were provided for the statistical analysis plan. Specific questions were deflected with generalisations and recommendations to consult a statistician. At that stage I was unimpressed, but there were no concerns.