This week, Elon Musk and Nvidia’s Jensen Huang discussed AI and the future of technology at the U.S.-Saudi Investment Forum. Here is Jensen Huang discussing radiology:
One thing that I will say, give you some evidence, is that, and I was just telling Elon about this earlier, radiology, for example, has largely been converted to AI-driven radiology. And there’s some really great companies doing that. And the surprising thing is the prediction that all radiologists would be the first jobs to go was exactly the opposite. The trend shows that there are more radiologists being hired now as a result of AI.
This is sheer unadulterated fiction. Leave aside the fuzziness of what “AI-driven radiology” might mean; AI simply doesn’t drive a meaningful part of the radiology workflow. Some AI list triage and a few algorithms to detect intracranial blood or fractures has not changed the game in even the slightest of ways. The only thing that has been in meaningful if still limited use over the past few years that has arguably driven even small efficiency gains is generative AI for drafting impressions based on dictated findings.
There are, of course, more things in rare use and plenty of things announced that might matter, but that is beside the point here: Jensen is wrong.
More radiologists are being hired because there is a shortage of radiologists due to steady, some would say incessant, imaging volume growth. The heavy utilization of CT and MRI in US healthcare has literally—and I do mean literally—nothing to do with AI.
The only thing AI could potentially have to do with hiring more radiologists is faster MRI scanning with some vendors/machines/protocols allowing for more patient throughput, impressive but still in its infancy and with limited market penetration.
And the reason for that, if you take a step back, it’s because the goal of a radiologist is not to study the images. The goal of a radiologist is to diagnose a disease. Now the studying of the images became so productive they could study more images, study more modalities, spend more time with the patients, and as a result, they were actually accepting more patients. We’re doing more radiology all around the world, we’re doing a better job with diagnosing disease.
Sure, the goal of a radiologist is to diagnose disease. But, let’s not pretend that radiology hasn’t, for the past century since its inception, essentially been the art and science of diagnosing disease through studying images and putting imaging findings in context. What Jensen is saying is just nonsense.
Even if we accept that the fraction of radiologists using clot- and fracture-detection tools is doing a better job diagnosing disease (very unclear), we are not, as a field “study[ing] more modalities” (what?!) or “spend[ing] more time with the patients” (less than ever thanks to heavy volumes, long turn-around-times, and the explosion of teleradiology). Current computer vision tools do not make radiologists significantly more efficient unless they inappropriately trust them enough to stop looking at the images.
And so that’s kind of the near term outcome of AI and productivity.
I don’t know if Jensen actually doesn’t know anything about radiology (everyone’s favorite white-collar AI-replacement use-case) or if this is a cynical don’t-fear-the-future puff angle. But either way, he’s wrong across the board.
I am reminded of Michael Crichton’s Gell-Mann Amnesia, where you realize how useless many perspectives and most news are only when confronted with obviously incorrect information in contexts for which you are a subject matter expert. Almost every single news article about AI and radiology is entirely wrong. They’re acting like what the world could look like in the coming years is what has already happened: that we are awash in game-changing, useful AI that has rapidly been deployed across the field and fundamentally altered the practice of radiology. And that’s not true. Ironically, clinicians using LLMs for documentation is probably far more ubiquitous and impactful so far.
Now being wrong about the present doesn’t necessarily mean being wrong about the future. Things are changing fast, and the future is always largely unknowable.
But the reality distortion is just so damn irritating.