Man vs Machine: A Doctor's Defence of Medical AI

And the results are in… Med-Gemini, Google’s medical AI model, has shown an astonishing ability in a variety of medical tasks. It’s recently published open paper shows the AI can achieve 91% accuracy in the MedQA (USMLE) dataset.

And perhaps more impressive is Med-Gemini’s utilisation of medical knowledge. Google’s medical AI has pushed the boundaries in correctly identifying chest x-rays normal x-rays and 65% in abnormal cases in certain datasets.

Given time, medical AI has the potential to become more accurate and better at diagnosis than the most accomplished medical specialist.

What does this mean for us mere mortals? The human doctors on the shop floor trying to bully our brains into absorbing the ever-changing guidelines and technologies? Are we now inferior to a machine?

Possibly.

Is this a good thing?

Possibly.

Here goes: I’m pinning my colours to the mast and voicing my support for medical AI. Here are my reasons why I’m optimistic about the rise of the medical supercomputer:

Not all doctors are great

One of the most quoted arguments against medical AI, is that it can’t have the same personal, empathetic touch as a human doctor. But having been part of countless medical teams for the last 20 years, I have seen my fair share of fallible clinicians. Clear-thinking and deportment suffer under stress and the extreme working conditions of today’s healthcare systems, can lead to erosion of good medical care.

If a medical AI is going to make the safest decisions, I think I would take that over a well-meaning but inaccurate human doctor, let alone over an obviously incompetent one.

AI as an adjunct

I see AI as an evolution in the tools doctors already have. I don’t pretend to know the whole of Kumar and Clark’s ‘Clinical Medicine.’ I am, after all, not a robot. But I do rely on reference texts and guidelines when I need them.

In the same way, I look up the dosing of medications in the BNF, I feel the use of AI could be a cross-check adding an extra safety layer in my practice. Using Med-Gemini’s example, if I think a chest x-ray looks normal but the AI says differently, I will probably take a closer look at the image, especially if the AI’s accuracy is as good as a Radiologist’s more often than not.

We haven’t cursed the release of many transformative medical technologies, even when they usurped previous gold standards of treatment or diagnosis. I am incredibly glad an MRI can better determine the nature of a patient’s lump, even though that makes a machine a better diagnostician than me. So too, I believe, AI will become another tool in my medical arsenal, something to be embraced rather than resented.

Redirect attention

It will take some time for AI to become the true deus ex machina, which replaces the human in the medical interaction. However, the hope is that by AI automating the administrative burden, a doctor will have more time to concentrate on the actual patient encounter. With no more endless clicking to order a test, hours of doctor’s time can be spent doing more important things and regaining a sense of balance and health in the working day.

‘Worst’ Case Scenario

Say the feared thing happens and doctors become obsolete, what then? Whilst I don’t think that is likely to happen, these highly talented minds can be directed to other things - maybe focusing on preventative healthcare or maintaining mental wellbeing and flourishing, rather than treating disease. Maybe we’ll all become bona fide creatives, after all, medicine is both an art and a science. I have no doubt us doctors will find something good to do with our time, if we’re no longer needed.