First things first
Every few weeks, it seems, there's another story of disappointment regarding IBM Watson Health. By most accounts, it was over-marketed, over-hyped, over-priced and, worse of all, under-performing. The latest, from a few fays ago, comes courtesy of the Wall Street Journal.
But that doesn't mean we should write off the role of artificial intelligence in health. Not by a long shot, in fact. Rather, as with many AI discussions, we probably just need to realign our expectations of what's possible and what can actually be helpful today.
We don't need to get rid of doctors, or to offload treatment plans to machines. We do need to help doctors spot conditions earlier or identify possible disease markers that they can investigate deeper. We don't need voice recognition or, it turns out, systems that can read and understand thousands of journal papers. We do need systems that can diagnose quickly and accurately, in order to make health care more efficient and bring it to presently underserved places.
Finding like these make me pretty confident we'll get there:
- Artificial intelligence platform screens for acute neurological illnesses at Mount Sinai (Mount Sinai):
- Artificial intelligence model “learns” from patient data to make cancer treatment less toxic (MIT):
- A major milestone for the treatment of eye disease (DeepMind):
- Artificial intelligence 'did not miss a single urgent case' (BBC News):
Yes, there are going to be some imperfections as research turns into products, but we all know perfection is far from the case today. If hospitals and drug companies take a measured approach to deploying AI, hopefully very few (or none) of these bumps in the road will be life-threatening or otherwise catastrophic.