The Rise of the Self-Driving Surgeon
Autonomous surgery, chatbots, and image recognition — oh my!
Autonomous Surgery is Now
In May of last year, a team of surgical researchers from Children’s National Medical Center in Washington, DC described something never before seen in the operating room. In a paper in Science Translational Medicine, they described a “Smart Tissue Autonomous Robot” (STAR) system capable of performing truly autonomous soft tissue surgery—and performing it better than human surgeons:
“We demonstrate that the outcome of supervised autonomous procedures is superior to surgery performed by expert surgeons and RAS [robot-assessed surgery] techniques in ex vivo porcine tissues and in living pigs.”
It’s still early days: in 40% of experiments with STAR, the human surgeons had to jump in to correct a problem. But STAR is the kind of development that makes even surgeons realize, as the STAR team admitted, that fully-autonomous surgical robots could, in the future, replace human ones.
In a conference call with reporters, research team member and surgeon Peter Kim made the analogy with self-driving cars:
Now driverless cars are coming into our lives. It started with self-parking, then a technology that tells you not to go into the wrong lane. Soon you have a car that can drive by itself.
AI is Coming Faster Than You Think
In general, I’ve found that most people in healthcare and related fields have little or no idea that AI-driven technologies are being developed that can do things that used to require a doctor or other provider.
For some reason, though, those same people aren’t too surprised to learn that there are now algorithms that can read an x-ray better than radiologist, or that there are chatbots that can provide mental health counseling. Perhaps these activities are so close to what we’ve seen in our personal lives (e.g. automatic face recognition on the iPhone or in Google Photos, or the now-ubiquitous WhatsApp conversations) that it doesn’t seem that remarkable.
But the existence of autonomous soft tissue surgery — an incredibly complex, multifaceted task combining image recognition, robotics, pressure sensors, feedback loops, and more, as demonstrated by the STAR team — sets audiences back on their heels (none more so than doctors) and has led many to question the future of the medical profession itself.
Are they right to worry? Could it be that these AI-driven tools could eventually become so easy to use, so fool-proof, that no doctor (or nurse, nurse practitioner, physician assistant, etc) will be needed at all?
Let’s Agree to Disagree
Intelligent and reasonable people disagree strongly on this question. Writer and surgeon Atul Gawande, who knows a lot about healthcare and medicine, and presumably less about AI, recently dismissed it as “dumb” on Twitter:
(though to be fair he was writing specifically about the Human Diagnosis Project, not AI in general) but the exceptionally successful (and decidedly non-dumb) technologist and venture capitalist Vinod Khosla, who knows a lot about technology (including AI) and business, claimed in a 2012 article in Forbes and Fortune magazine that
“technology will replace 80% of what doctors do.”
And that was before the recent explosion in machine learning successes.
The Next General Purpose Technology Revolution
Part of the difficulty in deciding how AI might affect healthcare, medicine, and our health is that it shows early signs of being what economists refer to as a “general purpose technology” (GPT), with wide-ranging effects, as discussed here by Tobias Huber (“Is Artificial Intelligence Really the Next Technological Revolution?”)
In economics, a GPT is defined as a generic technology, which (1) can be improved, (2) can be widely used and applied, and (3) expands the space of possible innovations and investments. Similar to historical GPTs, such as the steam engine, electricity, or microelectronics, these new interconnected and interdependent AI-based technology systems and markets have not only the potential to enable innovations in products, processes, and organizational structures — as previous GPTs did — but also to radically transform our economic, social, political structures.
With regard to healthcare, the ability to 3d-print a customized cast for a patient with a fractured arm is useful, but no one expects it will change the nature of what hospitals do — or threaten their business model and existence. With AI, as with the steam engine or electricity, there is the potential for dramatic transformation of the entire healthcare system model (indeed, of society).
All Roads Lead to Disruption
In the end, it may be that Khosla’s vision and Gawande’s aren’t that far apart: both may be of a future in which we need far fewer doctors per population than we have now.
If Khosla is right, and machines do 80% of what doctors now do, does that mean that we’ll only need 20% of the doctors we have now (perhaps adjusted for population growth)? Or that doctors will find other things to do?
If Gawande is right, and AI will just be a useful tool that makes doctors more efficient, won’t that also mean that we need fewer of those more-efficient doctors to accomplish the same thing?
And if the number of doctors is reduced, and communications and augmented reality and networking continue to improve, will it make sense for doctors to be in a central location (e.g. the hospital) and have patients come to them? Or will all or most visits to those remaining docs be virtual?