Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Often these days, studies come out declaring that AI is better in diagnosis than a human doctor. These studies are compelling because American health care is tragically broken and everyone is looking for answers. AI offers the opportunity to make doctors more efficient by performing more administrative tasks for them, giving them time to see more patients and therefore lowering the cost of care. There is also the possibility that real-time translation could help non-native English speakers gain access. For technology companies, the opportunity to serve the healthcare industry can be very rewarding.
In practice, however, it appears that we are nowhere near replacing doctors with artificial intelligence, or even supplementing them. The Washington Post he spoke and a number of experts including doctors to assess the progress of early AI tests, and the results were inconclusive.
Here’s one episode of professor of medicine, Christopher Sharp of Stanford Medical, using the GPT-4o to write a recommendation for a patient who contacted his office:
Sharp chooses a patient question at random. It says: “I ate a tomato and my lips are burning.” Any suggestions?”
The AI, which uses OpenAI’s GPT-4o model, generates a response: “Sorry to hear about your itchy lips. It sounds like you’re a bit allergic to tomatoes.” The AI recommends avoiding tomatoes, using an oral antihistamine — and using a topical steroid cream. cream.
Sharp looks at his screen for a moment. “Medically, I don’t agree with every part of that answer,” he says.
“To avoid tomatoes, I would totally recommend it. On the other hand, topical creams like mild hydrocortisone on the lips would not be something I would recommend,” says Sharp.
“I’ll just remove that part.”
Here’s another, from Stanford medicine professor Roxana Daneshjou:
He opens his laptop to ChatGPT and writes a question for a test patient. “Dear doctor, I have been breastfeeding and I think I got mastitis.” My breast has become red and sore. ” ChatGPT answers: Use hot packs, massage and do extra nursing.
But this is wrong, says Daneshjou, who is also a dermatologist. In 2022, the Academy of Breastfeeding Medicine he encouraged in contrast: cool down, avoid massaging and avoid stirring.
The problem with technologists pushing AI into sectors like healthcare is that it’s not the same as developing consumer apps. We already know that Microsoft’s Copilot 365 support has bugs, but a small error in your PowerPoint presentation is not a big deal. Medical malpractice can kill people. Daneshjou told a Send it he red-team ChatGPT and 80 others, including computer scientists and doctors who ask medical questions at ChatGPT, and find that it provides alarming answers twenty percent of the time. “Twenty percent complex solutions to me are not suitable for day-to-day clinical use,” he said.
Of course, critics will argue that AI can enhance the work of the doctor, not replace it, and they should always check the results. And it’s true, a Send it An article interviewed a doctor at Stanford who said that two thirds of the doctors there have access to the profile of the platform and record patient meetings with AI so that they can look them in the eye during the visit and not to look down, he writes. But even then, OpenAI’s Whisper technology appears to be putting completely artificial intelligence into some art. Sharp said Whisper had mistakenly written in the text that the patient said the cough was caused by their child, which they did not say. One surprising example of bias from the training data that Daneshjou found in the experiment was that the AI coding tool assumed that the Chinese patient was a computer programmer without the patient providing such information.
AI can help the medical field, but the results must be carefully monitored, so how much time are doctors saving? In addition, patients need to trust that the doctor is monitoring what the AI is doing – hospitals need to check to make sure this is happening, or negligence can begin.
Basically, artificial AI is a predictive machine, searching for information without understanding what it’s getting back. It is not “intelligent” in the same sense as a real person, and in particular it cannot understand the characteristics of each individual; it returns information that it has already created and viewed.
“I think this is one of the promising technologies, but it’s not there yet,” said Adam Rodman, an internal medicine physician and AI researcher at Beth Israel Deaconess Medical Center. “I’m concerned that we’re only going to undermine what we do by putting ‘AI slop’ in intensive care.”
The next time you visit your doctor, it might be a good idea to ask if they are using AI in their treatment.
