"AI Technology has been in its current form since the late-60s, early-70s."
That's clearly not right and the general public really started taking notice when IBM's Watson computer won Jeopardy. The next big shock will be in 2 years when they start operating self-driving cars in limited locations as a taxi service. The vast majority of what retail pharmacists do is something that the newer AI techniques excel at, which is pattern matching. For DUR and counseling they could use off-site pharmacists.
The Turk - Wikipedia
Define "newer". Do you know how much background preparation that Watson had to have to go on Jeopardy? Yeah, if you use the Turk programmers and maintainers, anyone in computer science can get that to work as a one-time matter, but not consistently without supervision. And no, Watson still is basically the same underlying technology (cluster/classification/neural networks) that past systems used and they do date from the 1960s-1970s (but has computer advantages in pressing the buzzer correctly to timing if you knew how Jeopardy works). Yes, computing power has increased, but not really our understanding of how to get those techniques to work. And yes, the self-driving cars are like plane autopilot. It works, but it has an error rate that has to be sustained, and none of them work well enough in inclement weather (and I don't mean the usual rainstorm or snow, I mean situations where the actual answer is not to drive at all) yet that it's releasable. I think this is a solvable problem to an extent, but like all AI, there's only so much you can do to train the algorithm? Watson was great PR, but if you notice the news from that time, Watson (and AI in general) never really has done very well in situations where you cannot define the problem very well, which hilariously enough, is what medicine is like.
That's kind of why Google can't get rid of their PageRank/TrustRank/HITs curators. Yeah, the underlying algorithm is quite explainable, but it requires lots and lots of manual curation, else you get the anti-Semitic AI bot:
Microsoft’s Chatbot ‘Tay’ Just Went on a Racist, Misogynistic, Anti-Semitic Tirade – Adweek
AI has been hyped every decade as solving many problems. Most forget that there was an AI Winter:
AI winter - Wikipedia
It's not that AI is not done in medicine, what is clinical decision support really if not AI rebranded? But, it's what the applicability of that AI is to your actual problem. What we train humans at present is to recognize and sort the problem into the right clinical decision support tool, and that turns out not to be easy.
I can tell you from very direct experience (to the point that I was present for the Watson tests) is that Watson has no credibility outside of the scientists and programmers that run the system:
IBM Watson Jumps Onboard Biden's Cancer Moonshot Initiative
VA signs $6 million contract for IBM Watson to advise PTSD treatment
Both turned out to be negative in terms of actual use. Mainly for the reasons just discussed, except the additional one is that cancer informatics could really, really use some work in terms of secondary data construction:
As I lay dying
and that prevents any meaningful work as there is not a good enough input situation to Moneyball cancer work.
The main challenge in using AI has always been to define a solvable problem and even harder, to define when a situation is out of context and you're relying on instinct. If the curriculum really were that deterministic, you all would never have done your practical 1600 hour internship. Unless you really had other experiences, that's really when you find out whether or not a candidate can hack it as a practitioner. Most of that is training that we cannot seem to pin on the curriculum, those situations where you should know that you are out of your depth, and those situations where ranking what's wrong takes a real backseat to the immediate situation. If you do have the resources to keep up a bunch of programmers to be the Turk, then it'll also work, but is it sustainable for the problem at hand?
Don't worry, your job won't be going away. It'll just get worse, such that only the most efficient pharmacists can keep up, and not for career long periods of time. AI will put a floor down on terms of how unproductive a pharmacist can be before termination, but won't replace a pharmacist willing to work to the limit. John Henry is a decent tall tale to describe what the future will be. Humans still exist, but will always be on the marginal brink.
