AI has already exceeded doctors in terms of diagnostic accuracy and throughput with older models. With newer models, better computational power, more widespread use and ability to analyze infinitely more data sets than humans, the future of basic and even advanced medical diagnoses is going to heavily use AI, if not completely depend on it (it has shown promise in diagnosing unknown diseases far better than humans). Is your contention that AI wont be used to a great extent in these areas?
AI these days can perform language modeling to perform diagnostics. There's no reason to believe that AI wont get better and more specific to the medical industry. ChatGPT a generic AI can already do some diagnosis, imagine what an AI trained specifically for medical purposes can do in the future?
And yes, that paper is a little outdated and a meta-analysis of older AI models. I only posted it as an example of AI already out-performing human diagnosticians in some areas. The blurb you quoted simply says that AI needs to be trained by doctors and can be used to make their jobs much easier.
For example, a recent 2023 study shows that AI has been shown to effectively diagnose heart, liver, lung and skin diseases, as well as breast cancers:
A broad range of medical diagnoses is based on analyzing disease images obtained through high-tech digital devices. The application of artificial intelligence (AI) in the assessment of medical images has led to accurate evaluations being performed automatically, ...
www.ncbi.nlm.nih.gov