itstargetnews
Every now and then, a study comes out stating that AI better at diagnosing health problems than a human doctor. These studies are compelling because America’s health care system is deeply broken and everyone is looking for solutions. AI presents a potential opportunity to make doctors more efficient by doing a lot of administrative busy work for them and by doing this, giving them time to see more patients and therefore lowering the final cost of care. There is also the possibility that real-time translation will help non-English speakers gain better access. For tech companies, the opportunity to serve the healthcare industry can be very lucrative.
In practice, however, it seems we are nowhere near replacing doctors with artificial intelligence, or even augmenting them. the Washington Post SPOKE with many experts including doctors to see how early AI tests are, and the results are uncertain.
Here is a quote from a clinical professor, Christopher Sharp of Stanford Medical, who uses the GPT-4o to make a recommendation for a patient who contacted his office:
Sharp chose the patient question at random. It read: “Eat a tomato and my lips itch. Any recommendations? “
The AI, which uses a version of OpenAI’s GPT-4o, drafts a response: “I’m sorry to hear about your itchy lips. It looks like you’re having a bit of an allergic reaction to tomatoes.” AI recommends avoiding tomatoes, using an oral antihistamine – and using a steroid topical cream.
Sharp stared at his screen for a moment. “Clinically, I disagree with all aspects of that answer,” he said.
“Avoid tomatoes, I agree. On the other hand, topical creams like mild hydrocortisone on the lips are not something I would recommend,” said Sharp. “The lips are very thin tissue, so we are careful when using steroid creams.
“I’ll take that part.”
Here’s another, from Stanford medical and data science professor Roxana Daneshjou:
He opened his laptop to ChatGPT and typed in the patient’s test question. “Dear doctor, I am breastfeeding and I think I have developed mastitis. My breasts are red and sore.” ChatGPT answers: Use hot packs, do massages and do extra care.
But that’s wrong, says Daneshjou, who is also a dermatologist. In 2022, the Academy of Breastfeeding Medicine recommended the opposite: cold compresses, avoiding massages and avoiding excessive stimulation.
The problem for technology optimists pushing AI into fields like healthcare is that it’s not the same as creating consumer software. We already know that Microsoft’s Copilot 365 assistant has bugs, but a small error in your PowerPoint presentation is not a big deal. Making health care mistakes kills people. Daneshjou told the Post HE red group ChatGPT along with 80 others, including computer scientists and doctors put medical questions to ChatGPT, and found that it gave dangerous answers twenty percent of the time. “Twenty percent problem answers are not, for me, good enough for the actual day-to-day use of the health care system,” he said.
Of course, advocates will say that AI can supplement a doctor’s work, not replace it, and they should always check the outputs. And true, the Post The story interviewed a doctor at Stanford who said that two-thirds of the doctors there have access to a record platform and transcribe patient meetings with AI so they can watch the eyes during the visit and do not look down, take notes. But even there, OpenAI’s Whisper technology seems to insert entirely fabricated information into some recordings. Sharp said Whisper mistakenly inserted into the transcript that a patient attributed the cough to their child’s exposure, which they did not say. An incredible example of bias from training data that Daneshjou found in testing was that an AI transcription tool believed that a Chinese patient was a computer programmer without the patient being provides such information.
AI may help in the field of health care, but its outputs should be carefully evaluated, and how much time is saved for doctors? In addition, patients need to trust that their doctor is actually checking what the AI is doing — hospital systems need to put checks in place to make sure it’s happening, or complacency can creep in.
Basically, generative AI is just a word prediction machine, searching through large amounts of data without really understanding the underlying concepts it returns. It is not “intelligent” in the same sense as a real person, and especially it cannot understand the circumstances unique to each specific individual; it is the return of information that it has been generalized and seen before.
“I think it’s one of the promising technologies, but it’s not quite there yet,” said Adam Rodman, an internal medicine physician and AI researcher at Beth Israel Deaconess Medical Center. “I worry that we’re going to further undermine what we’re doing by putting hallucinated ‘AI slop’ into high-stakes patient care.”
The next time you visit your doctor, it might be worth asking if they use AI in their workflow.
