Artificial intelligence: augmenting or replacing doctors?

“ARTIFICIAL intelligence [AI] could put doctors and lawyers OUT of a job in FIVE YEARS’ time,” trumpeted a headline in Britain’s Express earlier this year.

Could it be true?

There’s no doubt technology is radically changing the nature of work across most areas of human endeavour.

One of the many challenges confronting us as a society is the question of how to distribute meaningful employment, and the wealth associated with it, in a world where we just won’t need as much human labour as we used to.

We often assume that the jobs that will be lost will be mostly unskilled, but perhaps we’re kidding ourselves.

As machines become more “intelligent”, able not just to follow the rules we’ve given them but to continually improve their performance by learning from mistakes, they will inevitably expand their reach into new areas.

Intelligent systems are being developed that may match or even outperform radiologists in examining brain scans for stroke risk, or pathologists in assessing biopsies.

While that might be worrying news for the next generation of diagnostic specialists, it does offer hope for greater equality in access to health services (to diagnosis at least, if not to resulting treatment).

Intelligent machines will be able to diagnose far more quickly than any human, potentially processing thousands of images a day and making early diagnosis available to anybody with a smart phone.

A group of Stanford computer scientists and medical experts published an article in Nature this year that showed that their AI system matched dermatologists in its ability to distinguish both melanomas from benign naevi and keratinocyte carcinomas from benign keratoses.

Another study found that a machine-learning algorithm was able to distinguish subtypes of non-small cell lung carcinoma with similar accuracy to that of expert pulmonary pathologists. Combining the predictions of the algorithm with those of the humans led to an 85% reduction in error in detecting metastatic breast cancer in lymph nodes.

Perhaps that’s where AI has the greatest contribution to make, not in replacing humans but in augmenting them.

That may be how previous technological advances have operated, though some believe the disruptions likely to be caused by AI are in a league of their own.

Physicist Stephen Hawking, for example, famously warned that the development of thinking machines could spell the end of the human race.

Even if you don’t buy such doomsday scenarios, it’s worth asking what we may lose in harnessing the undoubted power of the machine for medical diagnosis.

With every new technology come fears of our dependency and resulting loss of our cognitive skills. Calculators took away our ability to do mental arithmetic. Google has destroyed our memories. Spell check made us forget the letters in algorithm.

Interestingly, one of the computer scientists in the Stanford dermatology study, Professor Sebastian Thrun, has also been one of the brains behind Google’s effort to develop driverless cars — an initiative that could, despite its obvious benefits, raise similar concerns.

When Google filmed “drivers” using its test cars, it found that they overtrusted the technology. Despite moving at high speed on the freeway in what they had been told was only a prototype, drivers were observed playing with their electronic devices without even a glance at what was happening through the windscreen.

Could the harnessing of AI in medical diagnosis lead to similar complacency?

Oncologist and author Dr Siddhartha Mukherjee has examined the likely benefits and risks of such developments in a thought-provoking essay in the New Yorker.

The algorithms may be better than humans at identifying pathology, but they would not have our capacity for enquiry, he writes.

“… in my own field, oncology, I couldn’t help noticing how often advances were made by skilled practitioners who were also curious and penetrating researchers.

“The chain of discovery can begin in the clinic. If more and more clinical practice were relegated to increasingly opaque learning machines, if the daily, spontaneous intimacy between implicit and explicit forms of knowledge … began to fade, is it possible that we’d get better at doing what we do but less able to reconceive what we ought to be doing, to think outside the algorithmic black box?”

It’s a good question. Whatever the future holds, it’s certainly going to be interesting.

Jane McCredie is a Sydney-based science and health writer.

 

To find a doctor, or a job, to use GP Desktop and Doctors Health, book and track your CPD, and buy textbooks and guidelines, visit doctorportal.