I was watching a Hasan Minhaj video recently in which he interviews journalist Joanna Stern about how most people still don't feel AI is doing anything great for their lives. Two points she makes in the video gave me pause. Pause enough that I felt compelled to note them down.
They are discussing how AI is being used in radiology to catch cancer. In a recent mammogram, AI flagged 3 spots that Joanna's radiologist, Dr. Laurie Margolies, had not seen.
First point: When Joanna asked Dr. Laurie if AI had caught things that she missed and saved a life, Dr. Laurie admitted "absolutely!". But she also mentioned the other side: She has caught things the AI missed and that saved lives too. In my, admittedly, cursory reading on the "AI will make radiologists obsolete" narrative, I've never heard the flip side.
Second point, the one that worries me: Hasan asks her if she foresees physician brains atrophying because of reliance on AI... and she says
"I have seen it happen in my own work, and so I have to assume it will happen in all work".
What an obvious but insightful observation.
These days I see job postings looking for the 10x or a 100x engineer. Imagine a job post for the 100x radiologist. An engineer relying on AI leading to a bloated codebase or a PM shipping a 10-page PRD instead of 3 pages is one thing, but a physician misdiagnosing or missing out on something because they over-relied on AI would be bonkers!
Here's the Black Mirror version: the "we won't need as many radiologists" narrative turns off people from pursuing a career in this field, thinning out the pipeline. So fewer people entering a field that depends on a steady supply of folks to train, sharpen, and eventually catch what the machine can't. The few who remain are overbooked or extremely expensive. So your insurance plan only covers "AI diagnosis" and "getting a physician diagnosis" is out of network.
In case you want to watch the section yourself: