Intended for healthcare professionals

Feature Healthcare Innovation

How does medicine assess AI?

BMJ 2023; 383 doi: https://doi.org/10.1136/bmj.p2362 (Published 09 November 2023) Cite this as: BMJ 2023;383:p2362
  1. Chris Stokel-Walker, freelance journalist
  1. Newcastle upon Tyne, UK
  1. stokel{at}gmail.com

Artificial intelligence is touted as saving time and improving accuracy of diagnosis, among other benefits in medicine. How are such claims being evaluated? Chris Stokel-Walker reports

Mindaugas Galvosas, a doctor in Switzerland, is excited. “Personally, I am extremely excited for what’s ahead for the field of acoustic epidemiology, and I have no doubt that cough counting will be as mainstream as step counting is—or even as thermometry is.”

He’s talking about a cough monitoring app, Hyfe, that is powered by artificial intelligence and that its manufacturer says will be able to use a patient’s pattern of coughing to diagnose disease. “Machine learning allowed for automating cough monitoring, with the new technologies being able to run in the background of any device with a mic and timestamp coughs as they happen to a millisecond,” says Galvosas, medical officer for Hyfe.

Hyfe is one of many AI powered tools claiming to supercharge medicine, saving time and labour for overstretched medical staff. Mention of generative AI (box 1) has become inescapable in the year since the research firm OpenAI released ChatGPT to the world in November 2022. Every week it seems like a new tool promising to aid doctors is released by big and small companies alike. In July 2023, for instance, Google claimed that one of its AI systems was more effective than doctors at analysing medical scans, while also reducing doctors’ workload by 66%.1

Box 1

What is generative AI?

Generative AI is a collective term for chatbot-style tools that can create text, including prose, poetry, or computer code, on demand, often through use of a simple “chat box” window. Their ease of use by experts and non-experts alike, and the seemingly fluent output—much of which can fool most readers into thinking the author was human—has led to widespread adoption by companies and the public.

Such …

RETURN TO TEXT
View Full Text

Log in

Log in through your institution

Subscribe

* For online subscription