This transcript has been edited for clarity.
Hi. I’m Art Caplan. I’m at the Division of Medical Ethics at NYU’s Grossman School of Medicine, located in New York City.
A story broke not too long ago that absolutely took my breath away. I was so astounded that this would even be proposed, but it was Elon Musk doing the proposing, and if you follow his activities, I suppose nothing should be surprising. He has all kinds of interesting ideas and pursuits, ranging from trying to go to Mars, to deciding how to cut the federal government’s budget, to projects like Neuralink, where he has been trying to use AI and computer technology to wire people’s brains to computers.
In this instance, he called for people who are on X (the Twitter site that he bought) to upload medical test results such as CTs, bone scans, and neural imaging to a new AI chatbot he’s trying to develop, called Grok, so that it can learn better how to read these test results — in other words, to replace doctors who are doing that work in trying to diagnose based on different types of scan technology with an artificial intelligence. He basically put out a call saying, try submitting an x-ray, a PET scan, an MRI, or other medical images to Grok through X, and we’ll try to make Grok a better, if you will, doctor.
There’s been some experience already with AI, and I’m just going to say that it’s been a long way from accurate. Many of you will have heard the term “AI hallucinations.” If you go out and take random information submitted by a subpopulation of people, not representative of everybody, you’re going to get many false findings.
Although we’ve seen some success using AI and analyzing blood samples, we’ve also seen AI misdiagnose breast scans, tuberculosis tests, and Pap smears. It is not anywhere close to ready for prime time. Elon Musk is going to agree and say, that’s exactly why I want you to send me this information, because we’ve got to improve; we’ve got to make sure that we do better.
Here come the ethics problems. To begin with, if you just go to X and say “Send me images,” you don’t know who’s sending images. You don’t know if they’re accurate. You don’t know if the people who use that are typical of the population. It’s not a reliable data source to teach about medical diagnostics.
You need to have an adequate sample. You’ve got to make sure that you’re not getting a bunch of samples from healthy White men or young people, who are on X more than others, if that’s true. Asking people to submit their own test results — who knows if they’re accurate or they’re telling the truth? Feeding junk to Grok is not going to produce a very intelligent alternative to a radiologist or some other form of physician diagnostician.
Second, when sharing this information, it’s easy to deidentify it. They’re going to know who you are. Musk is linked up to healthcare. I would say do not share anything without promises of absolute privacy. I don’t see them associated with this call for information.
Just like 23andMe, the genetic company that asked people to send their DNA to them for ethnic determination, ancestry determination turned out to have no ability to protect its privacy when it went bankrupt. There’s no guarantee that sending your images and scans as a patient is going to give you any form of protection down the road, should third parties want to take a look at what Grok has accumulated.
The last big issue is, why should we be doing this for free? Elon Musk is a gazillionaire. If he wants information, why doesn’t he go out and pay a representative sample of people to undergo tests, establish what a normal baseline looks like, and then try to explore what disease baselines look like?
That’s what we need to have good automated technology to help diagnosis — and note that I said help it, not replace it. If there’s no baseline and people are just randomly firing in medical tests, you’re not going to have an accurate AI diagnostician; you’re going to have a mess.
I think there are many reasons right now not to honor Musk’s request. As a wealthy and intelligent guy, he can do much better than saying, “Send me random tests and I’ll teach my chatbot to be a much better pseudodoctor.” I doubt that’s going to happen.
I’m Art Caplan, at the Division of Medical Ethics at the NYU Grossman School of Medicine. Thanks for watching.
