Page 10 - CUAJ Dec. 2022
P. 10
Leveridge
Consider whether the feel of the input need be congruent with the output. As it
stands, we might think, “I want the July CT report,” or “I want to move this grasper to
hoist up the prostate.” Currently, we search for the “imaging” tab or menu item and
click to get a step closer or make arm movements that glide in an arc that the robotic
arm mimics. In each case, we perform a process towards an outcome, but there is no
reason the process has to be so literal. The computer was always capable of these out-
puts, it was just programmed in such a way that the user is granted a sense of agency
by the input process. Programmed another way, one might speak into the air, “I’d like
the July CT report,” and the same output comes up as before (yes, I’m describing Siri).
Perhaps a “surgeon” at a screen uses the telestrator function (or any other interface
— a sensor-studded glove, a grid of IR beams, arms waving with a VR headset, etc.)
to simply point where they’d like the robot to go and which move to apply on arrival.
Taken further, if the computer can coordinate the moves we conceive of, perhaps
goal states (ligated DVC; stone fragments ≤2 mm; adenoma enucleated) instead of
processes will come to be standard. A mentor of mine envisioned a type of CNC
(computer numerical control) machine for partial nephrectomy, using high-resolution
imaging input to bloodlessly remove a renal mass like an engine block might be hewn
from a cube of aluminum. We see artificial intelligence programs like DALL-E and
GPT-3 already rendering cogent art and prose from basic text instructions. Of course,
surgery and medicine are complex and not just complicated; the substrate changes
during the act of care and biological variability is staggering, so it’s hard to compute
(pun intended) a machine having the adaptiveness to modify a surgical plan on the
fly and think clinically, as well as procedurally. Understand though that surgeons may
hone their skills and intuition through perhaps 1000 of a given case over a career; a
computer may have access to recordings of any robotic prostatectomy ever recorded
in building its skillset.
Well, we’ve managed to go from tapping A & B buttons on ‘80s consoles to a bleak
future of autonomous robo-surgeons in 1200 words, but our AI replacement is not
imminent. There are still only 30-odd robots in Canada, and the state-of-the-art prostate
imaging takes a $6 million machine the size of a car 45 minutes in a 2000 square
foot electromagnetic cage to render an image that looks like a jacked-up cookie your
toddler made. It’s just an angle on innovation that reminds us that our ability to take
care of patients is tariffed by interface decisions outside our ken and control. We reap
both the efficiencies and the fragility.
Correspondence: Dr. Michael Leveridge, Department of Urology, Queen’s University, Kingston, ON, Canada; Michael.Leveridge@kingstonhsc.ca
384 CUAJ • December 2022 • Volume 16, Issue 12