The Occasional Perspective: Much Abuzz About Something
Kevin Fickenscher, MD
Much like climate change, the world of deep learning and artificial intelligence (AI) is on the cusp of tectonic shifts occurring at incredible rates of change!! Whereas in the past, it seemed that temperatures were gradually warming over decades, if not centuries — now weather patterns change seem to change year-over-year. Well, the same can be said for the world of deep learning and artificial intelligence (DeLAI) except that it seems to be changing from year to year if not month to month. Furthermore, the changes could be led by the world of healthcare and sciences according to a new report from Accenture on the rapid evolution of AI with products like ChatGPT on the horizon. The report highlights the fact that 98% of global executives agreed that within this decade organizations will integrate a range of AI strategies as part of their business models. Furthermore, 42% of the surveyed companies were exploring investments in 2023 for the use of ChatGPT. We are on the horizon of a transformation of practices across all sorts of industries — including health care. The implications are clear. We will need to reconsider the traditional approaches toward governance, management, investment strategies, operations, people resources, and care delivery processes as well as approaches to data management more generally.
Let me be clear — the healthcare industry is not immune to these transformational changes. The Accenture report highlighted the impact on health care by noting that the first tectonic shift for the industry occurred in February when researchers from Massachusetts General Hospital (MGH) and AnsibleHealth — a chronic disease management company — reported findings that ChatGPT could pass the United States Medical Licensing Exam (USMLE). The drop in the pass rate is disconcerting enough. The arduous exam is a hurdle that we physicians sweat over for months before finally taking the exam 😊 so that we can “practice” medicine. But, ChatGPT passed the first time around without difficulty. The focus of the MGH-AnsibleHealth work was on the use of artificial intelligence as a “tool” for augmenting clinical care of medically complex chronic respiratory disease patients but the implications are much more far-reaching in my estimation and extend across the entire healthcare spectrum.
While it is clear that tools such as ChatGPT raise many legal, ethical, quality and/or reputation considerations the growth in the use of these tools is like a tsunami sweeping the healthcare landscape. We need to be thinking about these tools as resources for “clinically augmented intelligence”. It’s a term I’ve been using for nearly 15 years as a way of describing the potential for information systems and informatics to collaborate with clinicians on a real-time basis to “augment” care delivery. Now, (SEE hyperlink listed immediately above), I’ve learned that the folks at Harvard have put together a “Group” working on the same topic. Perhaps I was ahead of my time… Regardless, the medical community needs to become engaged on this topic so that we can develop and deploy risk assessment controls in the design of these tools as well as implement appropriate AI management principles across the spectrum of the health care community. It’s absolutely clear to me that we are on the verge of the transformation mentioned above which will require a concerted reinvention of the way we do our work. The future use of AI will no doubt greatly increase the value of the care we deliver and, most likely, provide better results — if we “integrate” the tools with clinical knowledge, understanding, and application.
But, we need to be clear — passing the USMLE is one thing, practicing medicine is yet another!! Researchers conducted further research on the use of ChatGPT which was reported in Cureus. A review of the article reveals that there is still much to do in “teaching” ChatGPT about medicine that goes beyond simply passing a test. In other words, we have our work cut out for us so that simplistic answers and solutions are not deployed inappropriately in the healthcare community. Let me be clear, I am a big believer in “clinically augmented intelligence”. It is the future. But, if we do not engage the future, mold the future, and support the future — it will simply be accomplished over, under, and around the healthcare community. Now is the time to engage.
On a secondary note, I believe we are NOT training our young physicians on how to integrate the tools that will be derivative of artificial intelligence, machine learning, and deep learning. A brief query of the medical school curriculums across the nation reveals that none (i.e. zero) of the medical schools have incorporated training in the use of these tools are part of their core curriculum. My advocacy is that learning about AI/ML/DL is just as important as biochemistry, microbiology, anatomy, physiology, and all the rest of the “basic sciences”. It is preparatory for entry into the clinical realm. Yet, not a single medical school (that I could find) has integrated training in these areas as prep for becoming a medical student in the clinical years or a resident in the specialty years!! This is a massive mistake in my estimation based on the speed and breadth of the change confronting the field of medicine and health care.
Why is my concern so immediate? It emanates from the fact that experts like Google DeepMind CEO Demis Hassabis have indicated that these tools of artificial general intelligence where computers will likely possess human-level cognitive abilities within just a few years. In other words over the next couple of years. If we accept the ongoing trends in the field, specialty knowledge — like medicine and health care — is just around the corner and, I’m predicting will be a part of “practice” by the end of this decade — in a mere seven years.
The final piece that must be considered is that the development of these tools is not confined to the United States alone. We are seeing developments across the globe. As such, there is a need for international cooperation and collaboration in setting standards and regulating these tools — much like we do for other areas of health care. Now is the time.
 The pass rate dropped during the pandemic. For allopathic students the pass rate dropped from 95% (2021) to 91% (2022); for osteopathic students, from 94% (2021) to 89% (2022); and, for international students, from 82% (2021) to 74% (2022)