In praise of boring AI

Stephen Bradley, a GP, wants his computer to work (doi:10.1136/bmj.q1171).1 He wants artificial intelligence (AI) to help make mundane and sometimes complex tasks easier. When was a drug started, why, and by whom? He wants help checking on outstanding care and reviews, as well as filling in reports and benefit claims. What about identifying clinical problems from a summary care record and making the most of clinical guidelines? None of this is fancy AI, the bleeding edge of innovation’s bleeding edge. But it’s the humdrum functionality that would improve the lives of health professionals and patients.

Technology brings benefits. It makes patient feedback easier (doi:10.1136/bmj.q1243),2 and the potential for a comprehensive electronic patient record is clear (doi:10.1136/bmj.q1385).3 A new online diagnostic tool offers hope of reducing delays in diagnosing ADHD (doi:10.1136/bmj.q1567).4 Updated NICE guidance on management of suspected sepsis aims to standardise care, and the data generated can help improve clinical practice (doi:10.1136/bmj.q1173).5 Each day delivers a new application of technology or AI, and it’s inevitable that Ara Darzi’s “diagnosis” of the ills of the NHS will lead to technology—possibly, technology enhanced polyclinics—being on the “prescription” (doi:10.1136/bmj.q1552 doi:10.1136/bmj.q1501 doi:10.1136/bmj.q1562).678

And herein lies the danger of being seduced by the magnificence of technology, of how it will solve all our problems while cutting costs, time, and carbon emissions. No tech entrepreneur, no early adopter of AI, wants to talk about solving life’s boring problems. The tech dream is transformational and aspirational. Better diagnosis of brain tumours? AI can do that. Safer surgery? We have a robot. Faster peer review? Forget humans, call ChatGPT. Who wants to invest in the boring: making a computer work, seamlessly connecting primary and secondary care, or finding your test results without navigating the Drake Passage? But perhaps we should—and, yes, some tech people do love boring tech.9

Technology is no free good. Digital innovation, to many people’s surprise, won’t necessarily reduce healthcare’s environmental footprint (doi:10.1136/bmj-2023-078303).10 A recent Lancet Commission looked at digital futures and, while embracing technology, identified a lack of engagement with marginalised young people, an imbalance between the profit motive and public values, and an under-recognition of the harms to physical, mental, and planetary health.11 Studies suggest that large language models, the bedrock of AI, would struggle with revalidation (doi:10.1136/bmj.q1507),12 even as deepfakes of doctors offering bogus and potentially harmful medical advice are proliferating on social media.13

To be clear, The BMJ is as enthusiastic as any organisation about digital transformation and working out how AI can improve what we do. But we need a reality check here. Progress won’t happen with our heads full of dreams and our feet stuck in the mud. The first problems that digital technology and AI must solve are the boring everyday ones that destroy efficiency and enthusiasm. Why are we reeling off use case after use case for AI, when doctors’ computers don’t work and patients can’t book an appointment?

True, the intelligent robots of Isaac Asimov’s world began as functionaries and servants who eventually superseded humans. The robots were rooted in mundanity, ethical codes, and order. Boring may not appeal to technology wizards, but it will make healthcare safer and better. And, if the machines are one day to rule over us, as it seems they inevitably must, the boring ones might just be kinder to their creators.