“Algorithms,” he noted, can be problematic because they “evolve after they are put into practice. It is unlikely that an accurate algorithm will stay accurate when deployed in real life over time. It needs adjustment and measurement of its operating characteristics continuously throughout the life cycle. How do we get that done?” That adjustment and measurement are not getting done now, he added.
Dr. Califf cited large language models as an example. “I see the regulation of large language models as critical to our future,” he said, referencing ChatGPT. The artificial intelligence writing tool has emerged this year as a popular tool among consumers and technophiles alike, with some in healthcare espousing its transformative potential.
Despite his concern, Dr. Califf is bullish on the potential of this technology—if it is handled appropriately: “Large language models are that next step that appears to be ushering in the revolution many of us are hoping for,” he said. “Imagine a world in which your questions were answered immediately in a language appropriate for your literacy and numeracy; also your clinician can actually talk with you rather than spending all their time cutting, pasting and writing clinic notes; I could go on and on.”
Misinformation
Regarding the ongoing threat of health misinformation spread, which has been an acute problem, especially since the pandemic started, Dr. Califf cited his visit to South Carolina the previous weekend, when he attended a funeral with relatives. The time he spent with them in his home state reminded him of the discrepancy of information they are usually exposed to compared with what he typically consumes in Washington, D.C., where he works and lives.
Many emerging technologies threaten to expand problems with misinformation in the U.S., he said. For example: “Technologies like large language models give almost everyone the potential to produce false narratives or even so-called deep fakes—fabricated images and voices.”
Support Structures
Overall, “a key part of a successful transition to digital health is an effective regulatory scheme to guide digital technologies,” he said. “Across industry, digitization and insertion of machine learning and other types of mathematical algorithms into everyday life is making a profound difference, but government agencies lag behind private industry. Quite simply, we need to assemble the resources to put in place these policies and tools and adaptively align our digital health efforts to support public health and regulatory innovation.”
