By ROBBIE PEARL
Shortly after Apple released the original iPhone, my father, an unlikely early adopter, bought one. His plan? “I’ll keep it in the trunk for emergencies,” he told me. He couldn’t have predicted that this device would eventually replace maps, radar detectors, AM radio traffic reports, CD players, and even coin-operated parking meters—not to mention the entire taxi industry.
This was a typical reaction to the revolutionary technology. We look at innovations through the lens of what already exists and insert the new into the familiar context of the old.
Generative AI is on a similar path.
As I planned the early April release of my new book, ChatGPT, MD: How AI-Powered Patients and Doctors Can Take Back Control of American Medicine, I was thinking deeply about the promises and dangers of generative AI in medicine. Initially, I worried that my optimism about AI’s potential might be too ambitious. I envisioned tools like ChatGPT transforming into centers of medical expertise within five years. But when the book hit the shelves, it was clear that these changes were happening even faster than I expected.
Three weeks before ChatGPT, MD debuted at number one on Amazon’s Best New Books list, Nvidia surprised the tech and healthcare industries with a flurry of headline-grabbing announcements at its 2024 GTC AI conference. Most notably, Nvidia announced a collaboration with Hippocratic AI to develop generative AI “agents” that are said to outperform human nurses at various tasks at a significantly lower cost.
According to data released by the company, the AI bots are 16% better than nurses at detecting a drug’s impact on lab values; 24% more accurate detection of toxic dosages of OTC medications and 43% better detection of disease-specific negative interactions of OTC medications. All at $9 an hour compared to the average hourly wage of $39.05 for U.S. nurses.
While I do not believe this technology will replace dedicated, competent, and compassionate caregivers, it will aid and support their work by detecting when unexpected problems arise. And for patients at home who can’t get information, expertise and support for medical concerns today, these AI nurse bots will help. Although they are not yet available, they are intended to be used to make new diagnoses, treat chronic diseases and give patients a detailed but clear explanation of the doctor’s advice.
These rapid developments suggest that we are on the cusp of a technology revolution that could reach global ubiquity far faster than the iPhone. Here are three key implications for patients and physicians:
1. GenAI in healthcare is coming faster than you can imagine
The human brain can easily predict the rate of arithmetic growth (where numbers increase at a constant rate: 1, 2, 3, 4). And geometric growth (a pattern that increases in a constant ratio: 1, 3, 9, 27) can also be understood reasonably well.
But even the brightest minds have difficulty understanding the implications of continuous, exponential growth. And that’s exactly what we’re experiencing with generative AI.
For example, imagine a pond with just one lily pad. Assuming the number of lilies doubles every night, the entire pond will be covered in just 50 days. But on day 43, you would hardly notice the green plants as only 1% of the pond surface is covered. It seems hard to imagine that just seven days later the lily pads would completely cover the water.
Experts predict that AI computing progress will roughly double every year, if not faster. But even with conservative projections, ChatGPT and similar AI tools will be 32 times more powerful in five years and over 1,000 times more powerful in a decade. It’s like riding a bike as fast as a car and then riding a rocket.
This pace of progress is difficult for both healthcare providers and patients to understand, but it means now is the time to prepare for what is to come.
2. GenAI will be different than previous AI models
When evaluating the transformative potential of generative AI in healthcare, it is important not to allow past failures like IBM’s Watson to cloud our expectations. IBM has set ambitious goals for Watson, hoping the company would revolutionize healthcare by helping diagnose, plan treatment and interpret complex medical data for cancer patients.
I was very skeptical at the time, not because of the technology itself, but because Watson relied on data from electronic medical records, which lacks the precision required for reliable “narrow AI” diagnoses and recommendations.
In contrast, generative AI uses a broader and more useful range of information sources. Not only does it draw on published peer-reviewed medical journals and textbooks, but it will also be able to integrate real-time information from global health databases, ongoing clinical trials and medical conferences. It will soon incorporate continuous feedback loops from actual patient results and physician input. This comprehensive data integration will enable generative AI to continually remain at the forefront of medical knowledge and be fundamentally different from its predecessors.
However, it will take several generations before generative AI can be widely used without direct clinical supervision. But Nvidia’s bold entry into healthcare signals the long-overdue willingness of technology companies to overcome the legal and regulatory hurdles in healthcare. Once an AI clinic chatbot is available, several other companies will quickly follow.
3. GenAI in healthcare will be ubiquitous (hospital, office and home)
Just as my father never imagined that his iPhone (stowed in his trunk) would become an essential tool for navigating life, many Americans have a hard time imagining the transformative impact generative AI will have on healthcare.
The concept of affordable, reliable and convenient access to medical advice and expertise 24/7 represents such a departure from current healthcare models that we can easily dismiss it as far-fetched. Yet it is becoming increasingly clear that these capabilities are not only possible, but probable.
Every day I receive feedback from both clinicians and patients who have interacted with current generative AI tools. Almost all report that responses, especially when effectively prompted, closely match the doctor’s recommendations. This is a testament to the increasing accuracy and reliability of generative AI in healthcare and promises to revolutionize medical care in the near future.
A decade from now, we’ll look back on today’s skepticism in much the same way I think about my father’s initial underestimation of his iPhone. We are on the cusp of a major shift in which generative AI will become as integral a part of healthcare as smartphones have become a part of everyday life. The question is whether clinicians will lead the way or leave this opportunity to others.
Robert Pearl MD is the former CEO of Permanente Medical Group, writes the Monthly Musings Newsletter and hosts two podcasts, Fixing Healthcare and Medicine The Truth. His latest book is ChatGPT, MD: How AI-Powered Patients and Doctors Can Take Back Control of American Medicine