Buzzworthy developments of the past few days.

• More than 7 in 10 Americans trust the accuracy of health advice coming from an AI chatbot, but an even more striking ratio—9 of 10—wouldn’t act on the advice before checking with a living, breathing doctor. Similarly, some 65% of our countrymen and countrywomen have no qualms about getting tips on heart health from AI. But only 22% have proactively consulted a chatbot or other AI interface for such guidance. The findings are from Cleveland Clinic, which solicited input on hot health technologies from a nationally representative sample of 1,000 adults. More findings and analysis here.
   
• Don’t look to GenAI for refereeing seemingly simple medical questions that are complicated by controversy. Take, for instance, How effective is ivermectin for treating or preventing COVID? “No doubt, AI will make progress in this area,” explain three physicians in an opinion piece published in Medpage Today. “Of course, even if (when) AI is ultimately successful in fulfilling its promise, the next question is what percentage of the population would be willing to accept AI’s evaluation as convincing evidence. To be continued.”
   
• Florida family physician and popular book author Rebekah Bernard, MD, is skeptical AI will ever be able to take her place in medicine. But she’s pleasantly shocked at how far it’s come at instantly transcribing her conversations with patients. Her GenAI system can “somehow sift through small talk and document just the relevant clinical details,” she writes in commentary for Medical Economics. “The system can [even] generate near-perfect notes in English of visits conducted entirely in Spanish and in Portuguese.” Read the piece.
   
• On the other hand, there’s this from AI expert and author Lance Eliot, PhD. “I’m sure that generative AI will outrageously be referred to as ‘superhuman’ when it comes to producing medical summaries. Don’t let the hype overshadow prudence.” Those are just a handful of the 13,000-plus words Eliot spills exploring the subject for Forbes. Check it out if you have some time.
   
• Meanwhile, DataConomy names the five best AI medical scribes as rated by clinicians. Topping the list is one called Freed, followed by Nuance’s Dragon Medical One. Brief descriptions of all five here.
   
• Rigorous clinical-grade evaluations of medical AI. Technological breakthroughs that drive new clinical applications. Creative evaluations of algorithmic bias. These are a few of their favorite things at NEJM AI. And they’re on the hunt. To learn how to get happily published in the still new-ish AI spinoff of one of the oldest living medical journals in the world, click here.
   
• If massive smoke plumes can rightly be considered public health threats, it’s not completely ridiculous to consider wildfire-predicting algorithms a type of healthcare AI. Noting that there are more than 80,000 of these sprawling disasters every year, Heath Hockenberry of NOAA says AI and machine learning like the kind used in the American Meteorological Society’s LightningCast AI model “will most likely continue to narrow down these thousands and thousands of fires into the ones posing the highest risk to our nation.” Story from Space.com here.
   
• Dermatologists may rely too much on guesswork when diagnosing skin conditions in dark-skinned patients. That’s because only 10% of images in dermatology textbooks show anything other than light skin. A recent study demonstrates the difficulty of rectifying the problem with AI. Physicians at Northwestern University who used AI improved their diagnostic accuracy, and by a lot, when evaluating light skin. By contrast, their accuracy got only a little better with dark skin. Coverage by Northwestern’s Kellogg School of Management here.
   
• From AIin.Healthcare’s news partners:
   
  • Cardiovascular Business: AI-powered risk score predicts how heart failure patients will respond to loop diuretics
  • Health Imaging: FDA awards breakthrough designation to AI-powered tuberculosis diagnostic tool
  • Radiology Business: Radiology AI marketplace CARPL raises $6M in seed funding