February 21, 2023 – A new artificial intelligence (AI) chatbot has the potential to transform the future of medical diagnosis, according to a February 13 op-ed in STAT co-authored by Andrew Beam, an assistant professor of epidemiology at Harvard T.H. Chan School of Public Health.
Launched in November 2022, ChatGPT is an AI model that predicts the next word in any given sentence. Trained on one of the largest databases of text from the internet, the popular chatbot has been used for a wide range of applications, from correcting computer code to writing original essays.
Beam and his colleagues previously tested website symptom checkers and smartphone apps that aim to help people self-diagnose their health concerns. They found that the tools were significantly less accurate in providing diagnoses compared to physicians. But when the researchers later tested ChatGPT’s ability to diagnose examples of patients’ symptoms, the AI was highly accurate, similar to physicians’ level of accuracy.
The researchers cautioned that AI could show bias and perpetuate harmful inequities present in the medical field. Despite the potential drawbacks, “given the interactive nature of these chatbots, we can see a future where people frequently turn to these types of tools for advice,” the researchers wrote.
“For physicians, AI tools could become a standard part of clinical care to reduce misdiagnosis, which unfortunately remains much too common in health care,” they added.
Beam also co-authored a preprint study, published February 1 on medRxiv, about the accuracy of the Generative Pre-trained Transformer 3 (GPT-3) artificial intelligence model in medical diagnosis and triage.
Read the op-ed in STAT: ChatGPT-assisted diagnosis: Is the future suddenly here?