Fineberg asked how Apple reconciles the trade-offs between customer privacy and scientific discovery. She said the company is exploring technical architectures like “private cloud compute” that allow de-identified, encrypted analysis at scale but prevent access to individual-level data.
Health care access and technology
As a moderator for one of the panel discussions, Ron Claiborne, an Emmy-winning journalist and former correspondent for ABC News, noted a recent survey in which about 20% of Americans reported taking medical advice from AI that later proved incorrect. He turned to the panelists for their reactions.
Fatima Rodriguez, MD, associate professor of cardiovascular medicine and associate director of the Stanford Center for Digital Health, said her research group has found that ChatGPT provides more accurate health information than a Google search. Still, she added, “I would say, ‘Use with caution.’”
Sinead Langan, PhD, co-director of Health Data Research U.K., said that in this “era of misinformation,” patients might trust ChatGPT more than a physician. Providing information to the public that counters distrust in health care institutions and explains the limitations of large language models could be helpful, she said.
“Because I think right now, everybody’s excitedly getting on with it and asking their questions and potentially then not following through with a medical appointment because they think they’ve already got the answer,” she added.
Langan also warned that digital health care systems, such as online scheduling of medical appointments, risks silently excluding older adults and people lacking an internet connection. There’s widespread anecdotal evidence, she said, that digital health tools present obstacles for such patients.
Jack Resneck, MD, chair of the World Health Organization and chair of dermatology at the University of California, San Francisco, said physicians think of AI in terms of augmented, rather than artificial, intelligence. “It’s not because we’re threatened by AI taking our jobs,” he said. “It’s actually because we do have a vision where patients, doctors and AI tools are kind of collaborating together to advance health.”
Ron Claiborne, Fatima Rodriguez, Sinead Langan and Jack Resneck. (Photo by SF Photo Agency)
Think distribution, not just innovation
A panel of investment experts described some of the barriers innovators encounter when they face the business side of health technology. Speakers noted that great ideas and innovations aren’t usually enough to launch or sustain a business; strong business plans are also necessary, especially ones that consider how a product will be distributed and fit into existing clinical workflows.
“If I find a founder that just says, ‘You know, this is going to be transformative because it’s amazing for the end consumer,’ it doesn’t show the intellectual interest, humility and wherewithal to actually figure out how to navigate the health care delivery system,” said Emily Melton, founder and managing partner of Threshold Ventures.
Jay Rughani, an investment partner at Andreessen Horowitz, agreed. “There’s been no shortage of technology innovation in health care,” he said. “But the primary bottleneck, at least in the digital health sector, has been a distribution.”
He added: “Someone who’s really excited about innovating on that distribution mechanism is what we’re looking for.”
Food fights
Christopher Gardner, MD, the Rehnborg Farquhar Professor and director of the Nutrition Studies Research Group at Stanford Medicine, discussed his experience working on the Emmy-winning Netflix documentary You Are What You Eat: A Twin Experiment. The film reached 50 million people in one month. It triggered national debate, backlash, copycats, political responses and deepfakes. In his talk, Gardner argued that when it comes to modern-day health science, evidence is necessary but insufficient; research impact depends on whether you’re able to translate that evidence into a compelling, mass-scale communication effort capable of surviving a rough-and-tumble information ecosystem.
Dariush Mozaffarian, MD, DrPH, director of the Tufts Food is Medicine Institute, noted that poor nutrition is the single largest, preventable driver of disease and health care spending nationwide. He argued that the health care system must treat food as a core therapeutic intervention — not a side issue — by integrating the idea that food is medicine into health policy, reimbursement, clinical care and digital tools.
Health care innovation
Speakers from medicine, robotics and engineering had additional insights into ways to improve health care with technology:
- Jeff Tangney, CEO and co-founder of Doximity, an online networking service for medical professionals, said health care innovation succeeds when “docs and dorks” — that is, clinicians and technologists — work together. He also advocated for putting physicians in leadership roles so that digital tools serve clinical reality rather than reimbursement logic.
- Michael Yip, PhD, director of the Advanced Robotics and Controls Laboratory at UC San Diego, described how humanoid robots are emerging as viable, general-purpose clinical tools that can help solve critical workforce shortages. These robots can be operated remotely and perform autonomous care-delivery tasks, he said.
- Tom Soh, PhD, professor of radiology, of electrical engineering and of bioengineering, spoke about how real-time, continuous molecular monitoring inside the body — analogous to how jet engines stream diagnostics — is becoming technologically feasible. Such advances could enable early disease detection and precision dosing of therapeutics.
- Mykel Kochenderfer, PhD, associate professor of aeronautics and astronautics, discussed how methods of sequential decision-making that developed for safety-critical fields, such as aerospace, can be applied to medicine to guide clinical decisions under conditions of uncertainty, as opposed to relying on intuition or one-step reasoning.
- Mohsen Bayati, PhD, a professor in the Stanford Graduate School of Business, said that AI hallucinations can lead to lethal harm in health care. Scaling AI models or feeding them more data won’t fix such problems, he said. Rather, AI in medicine becomes safer only when clinical structure, constraints and safety checks are imposed on models’ outputs.
In closing remarks, Linos reflected on a theme that stood out during the symposium — namely, the need to balance technological innovation with clinical purpose.
“Several of our speakers noted that progress in digital health cannot be defined only by sophistication of the algorithm or device,” she said. “What matters is purpose, empathy, creativity, trust and the relationship between patients and clinicians.”
link