While ChatGPT has entered American culture and lexicon as quickly as the latest TikTok trend or Taylor Swift hit, how it will work long-term in everyday life is still a mystery.
Some believe it will revolutionize the way we live, learn, and work. But many, including the Biden Administration, believe that before we see what it’s capable of, we need to fully understand what it is not.
For example, one lawyer recently found out that legal research is perhaps not ChatGPT’s strong suit – when the cases cited in his AI-crafted brief turned out to be entirely made up, which he had to explain to a furious judge. And while this is merely detrimental to his client’s case, and certainly this lawyer’s reputation, this kind of false or mistaken information could be deadly in healthcare. So as AI starts to move from its traditional use in nonclinical, administrative settings to more clinical settings, let’s look at what might be in store.
While older versions of AI are already in use for things such as reading certain scans and voice-to-text records, a recent study tested the empathetic answering capability of ChatGPT with a list of 195 common patient questions posed to both the AI chatbot and a physician. A panel of physicians was then asked to blindly rate which answer to each question was superior. The judges preferred ChatGPT’s answers in 79 percent of instances; not only were they considered high-quality and empathetic, but ChatGPT’s answers were about four times longer than the physicians’ answers.
But how practical is this in a clinical setting really? It’s not a slam dunk – the same issues that plagued our unfortunate lawyer exist in this setting. ChatGPT continues to make basic mistakes in math and often includes incorrect or falsified facts in its answers.
But what it might mean for providers is assistance, particularly in the face of provider burnout and increased patient messaging stemming from the increase in telemedicine usage in recent years. A provider might really benefit from running a question from a patient through ChatGPT and then simply editing as necessary. This could free up more of the provider’s time for in-person patient care. Other suggested uses have included summarizing patients’ medical history and writing clinical care notes and discharge instructions. Many question, however, how ChatGPT interacts with issues such as patient confidentiality, bias, accessibility, and protected health information.
And what about the legislative and statutory landscape surrounding this emerging technology? The Biden Administration has certainly not steered clear of the issue, releasing an AI Bill of Rights, and incorporating acknowledgment of the risks of this technology in many of its executive actions. Recently, the Administration also asked for public input on mitigating risks of AI, and hosted a listening session with workers who use AI in their positions, including those in healthcare.
Senate Majority Leader Schumer has met with a bipartisan group of senators to begin crafting comprehensive legislation to regulate AI, coming on the heels of several congressional panel hearings – including one that saw testimony from the CEO of the company behind ChatGPT. It’s clear that a legislative response is being mounted, but specific details are currently scarce on what guardrails, if any, will be put up on the technology.
When asked about its role in the healthcare industry specifically, ChatGPT stated that it was a, quote, “groundbreaking development poised to reshape the medical landscape.” While humility might not be another of ChatGPT’s attributes, you’re unlikely to find anyone who thinks that the technology isn’t here to stay, in at least some capacity.
So, for now, the healthcare industry is teasing out the opportunities and waiting for definitive answers on the how, when, and where.
