Last month, TechCrunch published an op-ed by technologist Kai-fu Lee arguing AI “stands to outclass doctors” and will eventually become superior to physicians in diagnostic capability. There is no denying AI’s value and promise for advancing healthcare. However, while AI will certainly transform the practice of medicine, it will not — and cannot — do so in the form factor it’s being offered today.
We cannot substitute the human brain with an AI point solution – especially in healthcare, an inherently human-centered practice that requires the experience, expertise, and empathy of human professionals.
Recent patient surveys reflect this. By and large, patients prefer providers over algorithms and are especially reluctant to embrace AI without human oversight. So while Lee centers his argument around the idea that AI will become superior to humans and eventually replace the need for them entirely, we say that can never be true for healthcare.
Rather, the true power of AI in medicine lies in its potential to augment the knowledge and expertise of human healthcare professionals. AI should never replace healthcare professionals, but it can and should be used as a tool that allows them to focus their time on higher-value interventions.
Not too long ago, the application of AI in EKG screening technology stoked fear among many in the cardiology community. Some saw the development of AI-interpretations of EKGs as a potential threat — one that could eliminate the profession of cardiology. Rather, AI-assisted EKGs have enhanced the field of cardiology, simplifying diagnostic tasks and enabling cardiologists to shift their focus to higher-value work at a higher volume. In addition, AI has allowed cardiologists to spend less time sifting through miles of EKG tape and more time thinking about and performing strategic interventions.
We’ve seen similar fears of technology replacing humans stoked in radiology. For example, after MRI/CT was introduced, the images were “so clear” that many argued that radiologists wouldn’t be needed to interpret them. Rather, image resolution improvements and data proliferation led to a drive toward sub-specialization and an expanded use of imaging as standard of care, and the need for radiologists increased drastically.
Likewise, AI will bring about the second era of subspecialization in radiology. There will likely be hundreds, if not thousands, of AI models for healthcare in the future. Their application, and the insights they produce, will require the steady hands of healthcare professionals.
The future of AI is one in which AI is used to amplify physicians’ intelligence, value, and clinical output — not replace them.
Why AI will amplify, not replace
Most AI models are probabilistic and work best when trained on a wealth of historical data. Unfortunately, this makes AI less valuable in instances where new, infrequent, and outlier situations occur— and dangerous in the cases where sparse training data could introduce bias into algorithms.
It does, however, make AI invaluable for automating tedious administrative work. With a bevy of training data, computers can be programmed to perform time-consuming, mundane tasks, ultimately freeing up professionals to focus on producing high-value work.
In personalized medicine, every patient — and every case — is unique. Interpreting and applying the aggregate of big data and individualized data to personalize treatment requires critical thinking and creative reasoning of the human mind. These skills are at the heart of medicine. AI cannot and should not replace them.
Reducing physicians to a role “resembling compassionate caregivers” — as Lee says AI will do — devalues their intelligence and misinterprets why they became physicians: to practice medicine. Lee’s argument completely misses or misunderstands what it means to be a physician.
The idea that AI can completely automate away the practice of medicine, where human lives hang in the balance, is not feasible — not least because of the moral, ethical, and legal complications Lee raises without adequately addressing.
AI will be most valuable to medicine when it is embedded in the workflow: automating tasks that make sense to automate and otherwise used to assist physicians in providing high-quality care to their patients. In this way, AI can be seen as the engine, workflow as the chassis, and physicians as the driver.
As AI is used to streamline administrative healthcare work and enhance the healthcare experience, human judgment and creativity will become even more critical. But, at the end of the day, physicians must have their hands on the wheel.
While AI will play an enormous role in the future of healthcare, it will never be able to do what humans do: practice medicine.