AI in medicine – balancing prudence with possibility

The existential threat of artificial intelligence is overstated at present

The advent of so-called artificial intelligence (AI) has prompted an extraordinary period of introspection, speculation, anxiety, and – at times – panic. ChatGPT, an advanced language processing AI developed by OpenAI, has generated particular interest. This application is designed to produce human-like text based on the input it receives. It was trained on a large volume of text from multiple sources. Certain versions of ChatGPT are free to access and are well worth a look.

Perhaps the most interesting feature of the current response to AI is how predictable the response is. Every new technology prompts a fresh bout of existential hand-wringing, as each generation believes that its technology presents vastly greater threats than any technology of the past. This was true for steam power in the 19th Century and, later, radio, television, video, computers, and the Internet.

The cycle is very predictable: Anxiety, panic, gradual settling down, and incremental adaptation of new technology. Today, we fear AI, and the cycle is starting again. Soon, more rational appraisals will emerge, followed by a degree of boredom, culminating in a certain amount of change, likely in ways we cannot yet predict.

One of the many ironies is that while AI is certainly artificial, it is far from clear if it is intelligent. AI is certainly, however, a triumph of marketing. Many technology companies appear to be generating a sense of panic based on an odd mix of modest technological progress, waves of apocalyptic hysteria, and canny existential button-pushing. AI tech shares have rocketed in recent times. Nobody ever got rich by assessing technology rationally or underestimating the human propensity to panic.

The truth is that each generation honestly believes that the technological changes it witnesses are not only incrementally bigger than those of the past, but vastly greater to the point of presenting an existential threat to the human race. We have believed this about everything from steam engines to nuclear power, from the printing press to the Internet. We inevitably think that this time is different, more, and worse.

Despite this Sisyphean cycle, our brains remain resolutely ‘unrotted’ by the telephone, radio, television, video games, the Internet, and AI. Change occurs, but rarely when or how we predict. Perhaps that’s what change really means: Unpredictable jolts forward from time to time. The truth about technology is usually quite mundane and not tremendously exciting. Change is incremental and we cope with it. The same will be true this time round.

Like all new technologies, AI presents real opportunities in many areas of human activity, including the practice of medicine. The possibilities range from medical education and research to providing care and helping patients to understand diagnosis, treatment, and prognosis. Used wisely, AI can help with many of these tasks.

Medicine can be an information-heavy endeavour, especially during undergraduate and postgraduate training. If models of AI become more reliable over time, it is likely they will prove to be useful tools for finding and summarising information, although they will still be based on information generated by humans, rather than AI. In addition, AI will not dispense with the need for careful, critical thinking by healthcare professionals. The waves of information provided by the Internet increased the need for critical thought, rather than diminishing it. AI will do the same.

For patients and families, the benefits are likely to be significant. Certain models of AI can already provide healthcare information in clear, written formats or spoken aloud (for people with poor hearing). They can summarise information in different ways, translate it, re-state points for greater clarity, make output larger or smaller (for people with poor vision), and add as much or as little detail as a person might request. The possibilities are vast. Used mindfully, AI will hopefully help to save time for healthcare professionals to devote to different tasks that necessarily require direct human input.

Careful, critical thought is also vital for patients and families as they explore AI. But this caveat is not a disadvantage of AI, but a note about how to use it. With this in mind, it would be wise to emphasise thinking skills in primary and secondary schools, and third level, in the future, to help ensure that AI and other technologies are used appropriately and well.

Inevitably, the new technology will bring certain changes to society, both positive and negative. Ultimately, however, the potential of AI more than justifies its examination, adaptation, and careful review, especially in medicine.

The existential threat of AI is overstated at present. As matters stand, AI does not show any sign of disproving Bernard Levin’s comment in The Times in 1978 that “the silicon chip will transform everything, except everything that matters, and the rest will still be up to us”. It was ever thus.

Prof Brendan Kelly is Professor of Psychiatry at Trinity College Dublin and author of Resilience: Lessons from Sir William Wilde on Life After Covid (Wordwell, 2023).