U.S. Air Force//Benjamin W. Stratton
Andrew Ng, the artificial intelligence researcher, tweeted a letter overnight from a worried MD now undergoing specialist radiology training — to become one of the folks who analyze X-rays, MRIs and CT scans: "I'm 3 years into specialist radiology training. Should I quit and do something else?" Ng did not reveal the doctor's name, but did query the AI community: "What would you say to him?"
What worries the doctor is the encroachment of AI into medical imaging: Radiologists are said to be among the most threatened workers on the planet, along with truck drivers and retail workers. IBM's experimental Avicenna software reads X-rays and CT scans, and China's Enlitic is already using its imaging products in health centers there.
Among hundreds of replies, the upshot was that AI, after replacing technicians and outsource radiology labs, will threaten senior radiologists, too, in a decade, said Jonas Lamis, a tech VC now with Nok Nok Labs.
But ... the bottom line: Luke Oakden-Rayner, an Australian radiologist and Ph.D student who studies machine learning, said humans will still be needed to read images because "many pathologies are too rare to develop large data sets in near future." But if you are looking for a guarantee, he suggested that Ng tell the doctor, go into interventional radiology, which uses imaging to direct medical procedures. It's a "sub-speciality as safe as houses!" Oakden-Rayner said.