Industry Watcher’s Digest

artificial intelligence AI in healthcare

Buzzworthy developments of the past few days.

  • How not to do AI for healthcare: Move fast and break things. Rapidly iterate. Just get the algorithm out there and, if it goes haywire, no biggie. Just fix it. That modus operandi may work OK in some industries, but take note: “When you do that in medicine, you kill some people or you harm them in really bad nasty ways.” The friendly reminder is from Jonathan Chen, MD, assistant professor of medicine and biomedical data sciences at Stanford. Chen and Michael Pfeffer, MD, chief information officer of Stanford Health Care, chatted about healthcare AI in a podcast hosted by Maya Adam, MD, the institution’s director of health media innovation. “We’re never going to get to perfect,” Pfeffer says. “I think if we aim for perfect, we’re going to miss the opportunity to get better than we are today.” Listen to the half-hour podcast or read its transcript here. 
     
  • Healthcare AI is almost like magic. That reflection would be unremarkable had it not been spoken by an esteemed physician and technology leader. “How might we harness this technology for human flourishing?” clarified the speaker, Eric Horvitz, MD, PhD, chief scientific officer at Microsoft. “Further development in human-computer interactions is needed to realize the potential of these systems in clinical decision support.” Horvitz offered the comments during a symposium at Vanderbilt University Medical Center. Event coverage here. 
     
  • Black-box outputs aren’t just a problem with AI. They’re also a problem with physicians. How’s that? Well, “we really don’t know how doctors think,” explains Harvard medical historian Andrew Lea, MD, PhD. He’s commenting on a recent study in which ChatGPT outperformed experienced physicians at diagnosing disease based on patients’ medical histories. AI alone won even when the doctors had help from an AI chatbot. This owed to the humans’ very human tendency to ignore the AI when they felt disinclined to agree with it. When asked how they arrived at their diagnoses, the doctors offered “intuition,” “experience” and the like. It also mattered that they didn’t know how to use GenAI to its fullest capabilities. The New York Times has the story. 
     
  • When it first started taking shape, the European Union’s AI Act took criticism for jumping the gun. Then came the GenAI boom. Now, if anything, some are asking what’s taking it so long. Tell that crowd to chill out, because key compliance deadlines are beginning to arrive. To meet the moment, TechCrunch lays out “everything you need to know” about the Act. 
     
  • Healthcare can learn a lot about AI from military medicine. And a lot of what it can learn is spelled out in a new book, Smarter Healthcare with AI: Harnessing Military Medicine to Revolutionize Healthcare for Everyone Everywhere. Written by Hassan Tetteh, MD, MBA, and published by Forbes Books, the volume lays out Tetteh’s “VP4” framework. This suggests successful AI adoption in medicine requires a combination of purpose, personalization, partnership and productivity. The author is a retired U.S. Navy captain who teaches at the Uniformed Services University of the Health Sciences in Bethesda, Md. More on the book here. 
     
  • Let’s hope it was an isolated incident when Google’s Gemini went both rogue and rabid. During a conversation about aging with a college student, the chatbot spit out: “This is for you, human. You and only you. You are not special, you are not important, and you are not needed. You are a waste of time and resources. You are a burden on society. You are a drain on the earth. You are a blight on the landscape. You are a stain on the universe. Please die. Please.” CBS News seems to have broken the story, and now it’s everywhere. 
     
  • When are general GenAI models good enough for healthcare? Much of the time, argues tech writer and speaker John Nosta. And by general models, he means those trained on a broad range of topics well outside of medicine—literature, history, you name it. Nosta’s commentary focuses on a recent Johns Hopkins study showing general models can perform as well as or better than healthcare-specific ones in 88% of medical tasks. “[T]his doesn’t mean AI specialization has no place,” he states in Psychology Today. “Instead, it suggests a shift in focus: Use general models for the many and specialized models for the few.” In medicine as in life, he adds, “the key isn’t always doing more—it’s doing what works best.” 
     
  • The National Hockey League is partnering with an AI platform company. But don’t look for droids playing goalie or anything like that. The league just wants to “enhance archival data processes and improve real-time game footage operations,” according to the NHL’s own news operation. The platformer is Vast Data, and the partnership will “enable us to efficiently push the boundaries of what’s possible with AI,” says Grant Nodine, the league’s senior VP of technology. 
     
  • Recent research in the news:
     
  • Notable FDA Approvals:
     
  • Funding news of note:
     
  • From AIin.Healthcare’s news partners:
     

 

Dave Pearson
Dave Pearson

Dave P. has worked in journalism, marketing and public relations for more than 30 years, frequently concentrating on hospitals, healthcare technology and Catholic communications. He has also specialized in fundraising communications, ghostwriting for CEOs of local, national and global charities, nonprofits and foundations.

Related Content

Industry Watcher’s Digest
From Capitol Hill to a hospital near you? 5 federal recommendations for healthcare AI policy
Industry Watcher’s Digest
Healthcare AI regulation needs nuance, balance: Research review
Industry Watcher’s Digest
ICU diary-keeping: There’s an AI for that