I doubt if there’s any of us, that have stood across the counter from a patient, who hasn’t been told what ChatGPT told them. Recently I had a patient who uploaded all their MRI scans and got a concerningly different result from the professional report. Given that their child was in considerable pain, for a protracted period, it was a natural and understandable approach to take. ChatGPT have put in minor barriers to health diagnosis. Yet, this person who is by no means technically literate, was able to circumvent the limited guardrails and produce what seemed to be an interesting report. Now, I am no radiologist but it was clear that the AI had picked up on an area of spinal stenosis that AI suggested was missed or glossed over. Whether this is relevant, I have no idea, but it certainly puts the patient on a collision course with the medical team.

A recent article in this magazine by Alan Reilly, about the European AI Act, was an eye opener, in that it forced me to think about where we are at (IPU Review, September 2024). Last year the EU grasped the thorny nettle of AI regulation. They decided to put in a few, on the face of it, very sensible safeguards. The flip side is that our AI development competitors, which currently would be the USA and China, have seemed to have put fewer, or no, safeguards in place. Whether we look at this on either a societal level, or in a very specific healthcare setting, it is a major concern.

Like it or hate it, AI is here to stay. All the evidence in front of us shows that it is almost like the Wild West. There are enormous amounts of capital pouring in. It has almost solely accounted for the gains on the US stock exchange for the last year. The rush being an expectation that all this will result in some sort of AGI (artificial general intelligence) at some stage in the very near future. This is when everybody, rightly, gets nervous. We all remember HAL in 2001: A Space Odyssey. If we have something that is actually more intelligent than us, because it has access to more information, it is quite possible that it may be capable of doing as much harm as good. While these concerns have been staples of the science fiction doom mongers, they are also the genuine concerns of some of the so-called parents of AI. Yet, when we look at the data, the overwhelming majority of AI is being used for personal queries. Very few businesses have made the step to getting useful returns from AI systems. This, the investors are betting, will change. However, life must go on. We must look at our own little niche, in the community pharmacy sector, and see what AI is bringing to us on a practical level.

If we aren’t, we should be expecting that many of our patients will be using AI to investigate, second guess, or help treat any condition that they, or their family members may have. We have always had informed patients. We are now at a whole new level. Anybody can open their phone, select from a growing number of AI specific browsers, and produce questions to which they feel that we need to respond. The only way that this can be satisfactorily resolved, from a professional point of view, is if we have access to more reliable resources than the public.  A type of ‘my AI is better than your AI’ situation.

Unfortunately, I haven’t seen anything that is making my life much easier in the dispensary, other than quick queries to help counsel patients. There are now several companies marketing AI to us, particularly on how they can improve efficiencies and improve sales. Yet, it is hard to justify an investment in time and energy for something that may offer little advantage over a spreadsheet and a few dedicated hours of work. There are days when I feel like I am working in a system that is almost pre-industrial revolution. Prescriptions need to be manually entered on the PMR. Checking prescriptions is a manual process. Most hospital prescriptions are handwritten and must be interpreted prior to data entry. Furthermore, many hospital prescriptions are riddled with inconsistencies and questions that place barriers to safe, timely treatment.

Currently, the number one item I dream of is something that helps me navigate drug shortages. Yet, the monkey on my back reminds me that AI uses data to train itself. Until solid safeguards are in place, the last thing any of us want to see is Mrs Murphy’s prescription records being regurgitated for little Johnny’s homework. Currently, we must not allow AI systems to interrogate our dispensary data. It is akin to publishing confidential medical information on the web. Until this major barrier is surmounted, safely, we can but wish.