Recently over the transom from an M. D. friend:
I became very interested in [Michael] Polanyi some twenty years ago when I was trying to write a book with an economist on what we previously termed ‘knowledge management’ in medicine. The book never came to fruition (some free advice – never write anything with an economist) but it did prepare me for when I taught healthcare informatics several years later.The challenge with AI in medicine is the temptation to reach the conclusion depicted in Figure 1. to say that generative AI is superior to human medicine.
Figure 1When I was teaching informatics, I used the concept in Figure 2, known as Friedman’s Fundamental Theorem of Biomedical Informatics, which I think applies to Philosophy as well – that human reasoning and AI together are superior to AI alone. Maybe as AI models evolve, it will be a larger amount of AI that accompanies the human thought.Figure 2
“That human reasoning and AI together are superior to AI alone” is certainly a plausible line to take. But the saying and the cute graphics raise some questions. I’ll raise a couple of philosophical questions and then point out a pretty obvious worry for medical practitioners.
1) Is it the brain that thinks when we think? (I mean an embodied living human brain adequately supplied with oxygenated blood and hooked up via the central nervous system with nature’s input and output devices, e.g. eyes, ears, vocal cords.) That our brains think when we think is what the graphic suggests. Might that be a misleading suggestion? It rules out mind-body (mind-brain) substance dualism which has deep roots in religion and philosophy (Plato, Augustine, Descartes, et al.) and is still defended today by a minority of philosophers, Richard Swinburne for example. Or do we think using our brains?
2) But suppose materialism is true and it is the brain that thinks when I think, and there is no immaterial, unextended, res cogitans along Cartesian lines that is the thinker of my thoughts. Then the question arises: Is the human brain an “automatic formal system” that operates in the same way that a computing machine operates? If so, then human psychology and cognitive science will coincide such that natural and artificial intelligence are both equally robust kinds of intelligence . “An automatic formal system is a physical device (such as a machine) which automatically manipulates the tokens of some formal system according to the rules of that system.” (John Haugeland, Semantic Engines, p. 5)
Or is it rather the case that, while materialism is true, and dualism false, the natural human brain and the artificial mechanical ‘brain’ are brains in such different senses of the term that only human brains are genuine semantic engines and truly intelligent. This is essentially John Searle’s view. “Computers are defined in terms of symbol manipulation, and symbol manipulation by itself is neither constitutive of not sufficient for meaning.” (Mind: A Brief Introduction, Oxford UP, 2004, p. 101.)
Whatever the implications of AI for the philosophy of mind, there are other questions ‘closer to the ground.’
3) For example, advanced AI will drastically reduced the need for MDs to do all sorts of things such as interpret the results of CT scans and other diagnostics. You don’t need a face-to-face consultation with a radiation oncologist to learn that a scan has revealed that a neck tumor is lousy with squamous cell carcinoma. And what about the techs who prepare patients for scans? Advanced robotics can replace many if not all of them. And the oncology nurses who administer chemotherapy? Ditto.