Artificial intelligence tools such as ChatGPT are being touted as proxy experts on many subjects, including medicine and disease, for their ability to pull from large data sets in a short amount of time.
While the promise AI holds within healthcare and other fields is self-evident, serious questions remain about both accuracy and transparency, studies have shown.
A comparison of Google’s and ChatGPT’s answers to queries about Alzheimer’s exposed a variety of knowledge gaps and inconsistencies, researchers said.
The findings have somewhat disturbing implications for caregivers, the study authors noted, because they are often the ones seeking information on how to make important healthcare decisions, and also have the responsibility of care.
While Google offered more current information, the results were skewed toward product providers seeking customers. ChatGPT had more “objective” information but, despite showing no sources, appeared to be pulling from outdated material, the study authors concluded.
In addition, both query results showed poor readability, and they could be difficult to interpret even if the information itself was accurate.
Perhaps because of the latter, other AI tools are now being developed to simplify medical jargon in reports.
However, ongoing issues with AI tools around potential biases, and concerns about privacy, need to be addressed.
The Alzheimer’s study was conducted by a team at the University of California, Riverside.
At least one expert in medical IT, Mayo Clinic President John Halamka, MD, has expressed skepticism of AI’s current capabilities in healthcare.
Halamka noted in a recent interview that the varied responses AI can give, even to the same query, made it difficult to assess quality and accuracy, particularly at the moment of use. One solution, Halamka suggested, was for healthcare companies to build their own AI models and tightly control the data-sets the algorithms were pulling from.
For better or worse, while many healthcare organizations are expressing optimism about AI, the hype lags behind concrete, actionable plans to utilize it, as Fierce Healthcare recently reported.
