‘Good afternoon, the AI will see you now.” This could be the future of family practice medicine as machines become better equipped to do many of the things that traditionally made up the GP’s job description.

Machines have always threatened human jobs, but up to the 21st century they were only a serious threat to repetitive, low-skilled jobs. Times have changed and the higher skilled jobs today are threatened by the advance of artificial intelligence (AI). In the past few years, intelligent machines have begun to perform complex tasks far better and quicker than intelligent, skilled humans.

The ‘good GP’ we know will always know his patients well; their medical history, their personality, details about their family and their foibles, and will retain this background information and call on it when assessing patients. The good GP would know, for example, that Mr Murphy is probably upset, rather than clinically depressed, because his father died last week and he has no depression history.

However, unlike the good GP, who may be overworked, and stressed by the demands of a busy clinic, computers with built-in AI have an almost unlimited capacity to store information in medical records, and to recognise patterns that may have been missed by the GP. They are also excellent at measuring things, such as blood pressure, and analysing results, such as a routine blood test.

In recent years, scientists working in AI made a breakthrough when they developed software based on the working of the human brain, rather than, as was the case before, sticking to a rigid rule book. This software could learn from the environment, making mistakes, and correcting them, like a child learns. The world saw its first artificial ‘self-learning systems’.

There were implications from this for the GP, and other professionals, as suddenly it wasn’t just the old story of machines replacing low-skilled workers; their jobs were on the line too. GPs faced the prospect of being confined to a decreasing number of areas where they are still superior to artificial rivals, or worse still, to become redundant; a charming medical relic of a bygone age.

The role of specialist medical consultants – not just GPs – is also coming under threat too from sophisticated ‘deep learning systems’ which are already outperforming doctors in specialised areas of medicine, says Professor Barry O’Sullivan, director of the Insight Centre for Data Analytics at the Department of Computer Science, in University College Cork, and a leading AI researcher.

“Dermatology is one such field,” he says. “A very recent paper that appeared in (research journal) Nature showed that a deep learning system could outperform human dermatologists at identifying skin cancer by quite a margin.”

Another area where AI is already better than doctors is radiology, adds O’Sullivan. A radiologist can recognise patterns in scans, based on years of experience. However, no matter how experienced the radiologist, he can’t store the amount of scans that AI can, while comparing them to a particular scan in order to assess, for example, the likelihood of a stroke or epilepsy. Meanwhile, there are further reasons for GPs not to be cheerful about the future with plans in the UK, supported by the NHS, to use smartphones to bypass the GP clinic altogether. A pilot scheme, run by private firm Babylon Health is encouraging people to consult a ‘chatbot’, equipped with AI, rather than a human being when they contact the 111 non-emergency line.

Patients key in symptoms, and AI determines how urgent each case is, and whether the user should be told to go straight to A&E, the chemist or simply go home to bed. The AI system makes decisions based on 300 million pieces of information, and the company says trials have shown the system is faster and more accurate than doctors. Certainly, most experts agree the future of medicine is ‘personalised medicine’, with disease treatments being tailored to suit personal needs. This is an area that AI excels, as it is very good at looking at medical data and finding specific treatments for patients which can, in some cases, even save their lives.

“For example, organisations such as Cancer Commons have assisted cancer patients survive conditions that were regarded as fatal in their situations by combining AI with extremely rich and individual-level data,” says Prof O’Sullivan. “The founder of Cancer Commons, Marty Tenenbaum, is one such survivor.”

Brendan Kelly, professor of psychiatry at Trinity College Dublin, says: “Computers and electronic diagnostic aids are delivering more and more information every year in medicine. And it is likely that computers and other devices, with their essentially infinite data storage and pattern recognition abilities, will continue to add increasing value to medical care.

He adds: “AI is especially useful in diagnosis: recognising skin lesions, monitoring measurable indices of body function, and identifying known patterns of symptoms and signs.

“It is less useful for other dimensions of medical care: contextualising findings in the context of the patient’s life, moving from recognising a pattern to agreeing a narrative between doctor and patient, and providing reassurance.

“AI is also less useful for picking up on the unexpected elements in a patient’s presentation: noting, for example, that a patient presenting with a lesion on his ear also has swollen ankles. Or noticing that someone with chest pain smells strongly of alcohol at 11 in the morning and might have an alcohol problem (which might or might not be related to the chest pain). Or noticing that a woman who comes to have a prescription renewed brings along her child, whom the nurse remembers has missed a vaccination, and can receive it today.”

Kelly isn’t worried about his job just yet though.

“Medicine has always welcomed new technologies and benefited hugely from them: stethoscopes, x-ray machines, MRI. But these tools amplify the effectiveness of healthcare professionals, rather than replacing them.”

The doctor will still see you now

Dr Mark Murphy, GP, and chair of Communications with the Irish College of General Practitioners, responds:

“It is unlikely that AI will make a meaningful difference in the GP setting, given the significant human factors at play in a consultation.

“There is a requirement for the GP to conduct face-to-face consultation to develop a rapport, to physically examine a patient and read nonverbal cues.”

“AI systems cannot get rid of the complex uncertainty and greyness at the heart of the undifferentiated presentations to General Practice.

“However, AI and machine learning will help with the handling of information in the healthcare system, and may be able to assist with complicated ‘black and white’ diagnostic process, such as imaging or to make sense of the complex genomic factors with rare-disease treatment.”