Artificial Intelligence: The Vine that Ate Healthcare
Author: Michelle Brooks, MD
Japan invades. Far Eastern vines
Run from the clay banks they are
Supposed to keep from eroding
Up telephone poles
Which rear, half out of leafage
As though they would shriek
Like things smothered by their own
Green, mindless, unkillable ghosts
In Georgia, the legend says
That you must close your windows
At night to keep it out of the house
The glass is tinged with green, even so
As the tendrils crawl over the fields
From Kudzu, by James Dickey, The New Yorker, 1963
I grew up in Louisiana, but our family would drive east on I-10 to Florida to visit my grandparents regularly. Kudzu was easy to spot as it grew everywhere and over everything. My dad always described it as the “vine that took over the south.” The vine engulfed not only the telephone poles described above, but also decrepit houses and abandoned cars. I associated kudzu with Alabama even before I knew how to read a map. The story is that kudzu was brought in as ornamental shade for southern porches. The vine can grow up to a foot a day. In the literal dust of the Depression Era, the Soil Erosion Service and Civilian Conservation Corps encouraged and incentivized southern farmers to cultivate kudzu to combat poor farming practices, to the tune of $8 per acre. Almost 3 million acres of kudzu were planted by 1946 on the government’s dime. Less than a decade later, kudzu was a plague, a fable reminding us on the drive to Florida that good intentions don’t always lead to good outcomes. You can read more about the history and use of kudzu in the south here.
The use of artificial intelligence and machine learning has been in the healthcare system for a while. I finished residency training in Internal Medicine 2010, but we already had computer-generated electrocardiogram readings that were confirmed or over-read by cardiologists. If you go back even further, there was even a very early medical chatbot in the 1960s called ELIZA, designed to mimic a psychotherapist. However, in late 2022, generative language model tools (ChatGPT) became less esoteric and available to the public, exploding the interest in this accessible and sophisticated tool. Many of the articles I’m seeing in the medical literature are optimistic that AI is the answer to the problems faced by physicians today. This review article by Haug and Drazen in NEJM touts faster documentation, quicker and better diagnosis, improved reimbursement, and simplification of research as potential benefits of AI. The authors do mention, “Like any good tool, they can help us do our job better, but if not used properly, they have the potential to do damage.” A few sentences later they concluded that “AI and machine learning can transform medicine.”
Full disclosure: I’m not an expert in AI/machine learning. I’m a geriatric millennial who is open to new technologies. I’ve used ChatGPT to generate emails that I don’t want to write or to make them more empathetic; surprisingly, my written communication is direct and can come across as robotic at times. I’m not alone in this flaw – in several studies, patients like the ChatGPT-generated responses better than physician responses to their questions. However, I’m still on the fence about AI/machine learning being a “good tool” and am interested to see what the proper uses will be. I wonder if planting these AI-based interventions into healthcare without having a clear purpose or goal other than to make work better is as naïve as planting kudzu to prevent erosion. Healthcare is investing billions in AI technology, incentivizing its widespread dissemination, despite AI’s flaws and ethical concerns. I remain hesitant and more pessimistic. From what I understand about the generative language model tools, they’re only as good as the information feeding them, and if they’re fed disinformation, they generate disinformation.
It’s not too far-fetched to say that ChatGPT is changing the landscape of medicine in the same way kudzu changed the south – rapidly and decisively. I agree with the authors above – it will be transformative; only time will tell if the transformation is helpful or harmful. It sounds like future directions for AI’s use in medicine will include closed AI/machine learning systems that use information only from reputable sources. This is a step in the right direction to cultivate AI in a controlled manner. If the growth of AI outpaces our ability to control its uses or improve its inherent flaws in objectivity replicated from human implicit biases, I worry that AI will become a strangulating weed, impossible to extricate. The use of AI in healthcare will become another polarity to manage in medicine, summed up by author Mimi Herman: “For something that can be seen as malevolent or seen as a savior, what I’m always trying to look underneath and understand — what is it really?”
Special acknowledgement: I was inspired to write this blog after reading Alexis Hauk’s article, Of Vines & Villains, in The Bitter Southerner. The poem by James Dickey and quote from Mimi Herman are taken from their article.
About the author: Dr. Brooks is the Deputy Associate Chief of Staff for Education at the South Texas Veterans Health Care System in San Antonio, Texas, and an adjoint Associate Professor at the University of Texas Health Science Center at San Antonio. She is also a Deputy Editor for Digital Media at the Journal of Hospital Medicine. Dr. Brooks is an award-winning clinician-educator with expertise on incorporating handheld technology and social media in medical education for residents and students. She has a background in quality improvement and patient safety and enjoys teaching quality and safety concepts broadly to interprofessional teams.
