Oncology services across the UK and Europe are under significant pressure. Cancer diagnoses are projected to increase by 21% by 2040. With a system already at breaking point, the need for tools to support safe, evidence-based decision making has never been greater.
Large language models (LLMs), artificial intelligence (AI) systems trained on vast quantities of text and encodable with clinical knowledge, may help address this by automating time-consuming tasks such as clinical information retrieval, supporting clinical decision making, and generating clinical summaries.
LLMs are becoming popular among patients and clinicians very quickly. Where our patients would have once said “I turned to Dr Google”, they now say “I asked ChatGPT.” The interest and excitement in this technology have been palpable, reflected in both the commercial funding LLMs are attracting and the volume of academic publications, with a recent systematic review identifying 4609 peer-reviewed studies on LLMs in clinical medicine published in under 4 years.
LLMs undoubtedly hold promise for health care, but we are yet to see the impact on productivity that technology enthusiasts predicted. The challenge is not one of technological performance alone, but whether health care systems already under strain can assess, validate, implement, and sustain this technology. There is a pressing need to narrow the gap between technological promise and real-world adoption that has compromised past digital health investments.
