Sort:  

"AI isn't replacing radiologists" — good article
Expectation: rapid advances in image-recognition AI would eliminate radiology roles, as predicted nearly a decade ago. Reality: radiology is healthy and expanding

Many recent predictions about AI's imminent impact on jobs are overly simplistic. For example, about a year ago someone asked if software engineers would still exist today; indications are they will

The piece explains why it's more complex, using radiology as an example:

  • benchmarks are far too narrow to reflect real-world scenarios
  • the role is far more multifaceted than just image recognition
  • deployment barriers exist: regulation, insurance and liability, diffusion and institutional inertia
  • Jevons paradox: speeding radiologists with AI can generate more demand

Radiology was a poor exemplar in 2016—too complex, high-risk and regulated. For nearer-term disruption, look to jobs that are repetitive, independent, low-context, short, forgiving of mistakes, and easily digitized. Even then, AI will likely appear first as a tool, shifting work toward oversight and supervision rather than outright replacement

About six months ago a vote was posed on whether there will be fewer or more software engineers in five years; the answer is left to the reader