)
)
Why are humans still in the loop with advancing AI capabilities?
Large language model generative artificial intelligence (AI) systems have opened Pandora’s box, beating human benchmarks across a range of tasks. AI systems alone can outperform humans who have access to those same AI systems on complex medical reasoning tasks. Such studies, of course, do not reflect the breadth nor depth of our clinical practices. Similar to physician licensing and board examinations, however, they isolate complex reasoning processes previously thought to be in the exclusive domain of human experts. These results fly in the face of the fundamental theorem of informatics: human+computer, when combined, will outperform either alone.
(Re)training, redirection or replacement are essential areas for future studies of human+AI collaboration. Many participants in the above studies had never used advanced generative AI systems. Advances in technologies that may now surpass humans in isolated capabilities warrant patient-centred studies in live clinical settings and revisiting what qualities human professionals can uniquely provide.6 Why do we still keep humans-in-the-loop? What makes a doctor a doctor? See table 1.