FHC #232: Can GLP-1s & GenAI solve healthcare’s affordability crisis?

0:12Dr. Robert Pearl: Hello and welcome to Diving Deep, part of the Fixing Healthcare podcast series. I'm one of your hosts, Jeremy Corr, also host of the Pop Unit Books and Medicine podcast and CEO of Executive Podcast Solutions. With me is Dr. Robert Pearl. For 18 years, Robert was the CEO of the Permanente Group, the nation's largest physician group. He's currently a Forbes contributor, a professor at both the Stanford University School of Medicine and Business, and author of the bestselling books, Mistreated, Why We Think We're Getting Good Healthcare and Why We're Usually Wrong, and Uncaring, How the Culture of Medicine Kills Doctors and Patients.

0:41Dr. Robert Pearl: His new book, ChatGPTMD, How AI-Empowered Patients and Doctors Can Take Back Control of American Medicine, remains a bestseller. If you want information on a broad range of healthcare topics, you can visit his website, robertperlmd.com. Robbie, listeners are increasingly concerned about the cost of medical care and their ability to pay for it if they become sick. In the midterm elections coming up, health care affordability is near the top of issues for voters. Where are we today relative to affordability and what do you believe can be done?

1:09Jeremy Corr: Jeremy, health care affordability, that's a massive topic. Remember that medical expenses represent 20 percent of our nation's gross domestic product. And healthcare hiring is not only the fastest growing segment of the U.S. economy, but one of the few parts that is increasingly. So let's begin at looking at where we are and then what can be done. As listeners know, we're increasingly telling us, and they're also telling elected officials, that healthcare costs have become a growing problem for them and their families. Nearly two-thirds of U.S. adults worry about being able to afford their healthcare expenditures, including purchasing insurance, paying copayments or deductibles for office visits, and buying prescription drugs.

1:49Jeremy Corr: That's according to the Kaiser Family Foundation Health Policy Research. A new study found that 25 % of Americans, they were in debt last year as a result of a medical bill that they couldn't afford to pay. Meanwhile, Gallup reports that only 49 % of adults are cost secure. That means that they have access to high quality affordable medical care, and they've recently been able to pay for both the needed care and prescriptions that they require. And their employers, they don't have it much better. What we've seen is that national health spending, it's risen at twice the rate of overall inflation.

2:25Jeremy Corr: According to the Centers for Medicare and Medicaid Services, health care costs have gone up 7 % for each of the past three years. As a result, the U.S. spending now surpasses$5.7 trillion. That's more than the total gross domestic product of every nation in the world, except the United States and China. That means Americans spend more on medical care than people in all other countries do on food, clothing, transportation, education, and everything else they buy, including health care. And the problem is only beginning. CMS predicts that medical spending will rise to$9 trillion by 2034.

3:05Dr. Robert Pearl: Robbie, that rate of growth in medical costs is terrifying. What's being done to address it?

3:10Jeremy Corr: Jeremy, you would think that there would be major efforts by payers to flatten the rate of increase, yet efforts by insurers and employers to restrain medical costs, they continue to rely on the same ineffective tools that our nation has deployed for decades. Across the country, we see employers cutting benefits. In parallel, insurers are restricting access. They're requiring prior authorization for medical treatment, and they're shifting more of the bill to patients. In the healthcare strategy course that I teach at Stanford's Graduate School of Business, and in my role in the faculty at the Stanford School of Medicine, I advise students who want to improve the U.S.

3:47Jeremy Corr: healthcare system and to address this growing unaffordability to follow Sutton's law and go where the money is. And when it comes to healthcare costs, that place, its chronic disease and its complications.

4:03Dr. Robert Pearl: That sounds promising. Can you please explain on the idea?

4:06Jeremy Corr: Data from the Centers for Disease Control and Prevention show that chronic diseases account for most illnesses, most disabilities, and most deaths in the United States. They're the leading drivers of the nation's health care costs. Published estimates from the CDC associate five major chronic conditions. We're talking about diabetes, cardiovascular disease, cancer, obesity, and dementia. These five account for more than$1.3 trillion in annual health care expenditures. It's been estimated that effective control of chronic diseases could reduce the incidence of heart attacks, strokes, kidney failures by 30 to 50 percent, and that would save hundreds of billions of dollars.

4:51Jeremy Corr: Five years ago, medicine lacked the tools to address these problems. However, the combination of GLP-1 medications and generative artificial intelligence, they provide a path to affordability. They're like superheroes joining forces. These innovations would attack the evils of chronic disease from all sides, preventing them from developing in the first place and controlling them should they develop before they produce costly, life-threatening complications, like, as I just mentioned, the heart attacks, the strokes, the kidney failures.

5:24Dr. Robert Pearl: Robbie, I'm intrigued. Can you first expand on the prevention opportunities?

5:28Jeremy Corr: Jeremy, as you know, obesity affects more than 100 million American adults. That's according to the CDC. And it increases the risk of diabetes, cardiovascular disease, 13 types of cancers, sleep apnea, and liver disease. It's associated with an estimated$260 billion in annual medical spending. That's based on a 2021 study published in the Journal of Managed Care and Specialty Pharmacy. GLP-1 drugs, they can't address the problem. It isn't that doctors and patients don't know that obesity is a problem. It's just that telling people to cut calories and exercise, it doesn't work for most individuals.

6:09Jeremy Corr: Even programs like Weight Watchers, they only achieved a small improvement for enrollees. But GLP-1 medications, they've made a huge difference for the majority of users, and their use is growing and it's having a national impact. As an example, Gallup reports that the U.S. adult obesity rate declined from a peak of 39.9 % in 2022 down to 36.4 % in 2026. And it did so as the GLP-1 usage increased nearly 60 % across that time. As we've discussed before, population data do not establish causation. But supporting the association, Gallup found that the largest obesity declines, they occurred in age groups with the highest GLP-1 use.

7:00Jeremy Corr: Similarly, clinical trials provide strong evidence for the efficacy of drugs like ZetBound. In a 2025 trial published in the New England Journal of Medicine, adults with obesity but without diabetes lost an average of 20.2 percent of their weight on trisepatide over 72 weeks. Another New England Journal of Medicine published trial found that among adults with obesity and pre-diabetes, only 1.3 percent of those treated with trisepatide developed type 2 diabetes, and we're talking about over 176 weeks. And that compares with 13.3 percent, you know, 10 times higher than receiving placebo. Yet another found fewer worsening heart failure events among patients with obesity and heart failure when they were treated with trisepatide than among those receiving placebo.

7:50Jeremy Corr: Jeremy, these results are remarkable. It's a level of improvement that doctors and patients, they've sought it for decades, but with little success. And now the weight loss era has arrived. And it's not just shedding pounds. It's a positive impact on cardiovascular health, and dozens of medical conditions. However, as vital as prevention is, as encouraging as the data is on combining these medications with lifestyle improvements, we're talking diet and exercise, we know that even widespread GLP-1 use won't prevent every case of hypertension, diabetes, or heart failure. We can expect that most likely millions of Americans will continue living with chronic conditions that would benefit from daily management and effective control.

8:33Jeremy Corr: And yet currently, American medicine does a poor job of controlling these major contributors to life-threatening and very expensive complications, the ones that require hospital admission and interventional procedures.

8:46Dr. Robert Pearl: Robert, can you please expand on the failure to control these chronic conditions?

8:50Jeremy Corr: Jeremy, the data shows that doctors and patients are struggling to control chronic disease, even though highly effective medications are available and clear treatment guidelines exist. The reason is that U.S. medicine still relies on brief visits every three or four months in the doctor's office rather than continuous monitoring for the patient in their home and timely medication adjustments. Let's think about hypertension, high blood pressure. This disease affects 48 % of American adults. That's according to the CDC. And yet fewer than one quarter have their blood pressure under control. In 2024, high blood pressure was a primary or contributing cause of 680 ,000 deaths.

9:34Jeremy Corr: And it was associated with approximately$219 billion annually in medical spending. And that's based upon the CDC's latest national cost estimate. The problem, A physician may prescribe the right medication in, let's say, January, but they wouldn't know for months that the dose was too low, or that the patient would stop taking it, or that a second drug was needed. Similarly, in adequately controlled diabetes results in approximately 44 % of new cases of end-stage kidney disease. These are the ones that are required to all this as a transplantation, and it accounts for 5 % of Medicare's total budget.

10:18Dr. Robert Pearl: Robbie, what can be done to solve these challenges?

10:22Jeremy Corr: Jeremy, today physicians lack the time to review a torrent of blood pressure readings, blood glucose measurements, weight fluctuations indicative of acute heart failure, or other objective types of clinical data from patients who have a chronic disease on a frequent basis. No clinician wants to receive daily blood pressure readings on hundreds of individuals. If they did, they could make rapid shifts in treatment and achieve superior clinical control. But it isn't realistic. Gen. AI makes it possible. Connected to home monitoring devices, Gen. AI agents could analyze daily measurements, identify when treatment is failing and alert the individuals and their clinicians when medications need adjustment.

11:04Jeremy Corr: And the ability of Gen. AI to provide this type of clinical reliability, it's being recognized in dozens of clinical research studies. already generated AI as matching or exceeding clinician expertise. In a senior Sinai study of 461 virtual urgent care visits, an independent panel of doctors rated 77 % of AI-generated recommendations as optimal, and that compares to only 67 % that judged them following a doctor's decision. In a separate randomized trial published in JAMA Network Open, GPT-4 scored 16 percentage points higher on diagnostic reasoning than physicians using conventional resources.

11:50Jeremy Corr: What we're seeing time and again is that these models are able to match human performance on a variety of tests, whether theoretical ones or ones based upon actual clinical care.

12:05Dr. Robert Pearl: Robbie, does this mean we're ready for generative AI tools to replace doctors?

12:10Jeremy Corr: Jeremy, neither study suggests that AI should replace doctors. However, the results indicate that it would be a mistake for a nation to ignore opportunities to improve medical accuracy through continuity and disease control by applying generative AI. As an example, a national analysis published in the British Medical Journal, Quality and Safety, estimated that diagnostic errors contribute to approximately 371 ,000 U.S. deaths annually. And a widely cited Johns Hopkins research study published in the British Medical Journal estimated that more than a quarter of a million U.S. deaths, we're talking about each year, come from medical error.

12:53Jeremy Corr: If American medicine were achieving perfect outcomes, we wouldn't take the risk of using generative AI in the provision of medical care. But when you add probably another half million avoidable deaths from poorly controlled chronic disease to the total. It's clear there's much room to improve clinical outcomes compared to what the U.S. healthcare system achieves today. My view is that American medicine and practicing clinicians, they need to view these tools not as competitors, but as complementors. Tools capable of fulfilling some of the clinical roles that clinicians don't have the time to compete today, but add major value when used effectively with patients.

13:38Jeremy Corr: To that end, I would encourage them to think about all the ways they could use these generative AI agents, not just to make the job of a doctor easier or reduce the administrative costs, but as tools to save the most lives.

13:53Dr. Robert Pearl: Robbie, you've covered a huge amount of ground already. Let me ask you how our nation should integrate all the pieces and possibilities to save the most lives and help all Americans to be as healthy as possible.

14:05Jeremy Corr: Jeremy, GLP-1s and Genitive AI attack each end of the nation's chronic disease crisis. The GLP-1s reduce obesity and the risks of diabetes, along with cardiovascular disease, cancer, and other obesity-related complications. Genitive AI, on the other hand, helps clinicians to control chronic conditions. once they develop, before they lead to life-threatening complications, complications that require costly medications and result in hospitalizations. But as simple as that sounds, these opportunities will be complex to implement. For that reason, I believe that translating opportunities to practice won't happen if doctors and patients try to do it on their own.

14:51Jeremy Corr: Success will require assistance from the government and help from large national medical societies. The government will need to take action to help lower health care costs by making access to GLP-1 medications and generative AI easier for all Americans who would benefit. Currently, Americans pay far more for GLP-1 medications than patients in other countries. And the cost is the primary reason that so many people quit using these weight loss drugs. We know that when people stop the medication, they regain almost all the weight as a consequence. And for that reason, if the goal is to prevent chronic disease and cancer, and cost is the primary driver for patients abandoning one of these weight loss medications, then the obvious logical first step has to be for the government to rein in prices.

15:39Jeremy Corr: Fortunately, the government has taken the first steps down that road. Jeremy, the Centers for Medicare and Medicaid Services have demonstrated one potential path to affordability. Its Medicare-GLP-1 bridge gives eligible beneficiaries access to certain weight management medicines for a$50 monthly copayment, with the government paying an additional$200 each month. Given that, according to Yale researchers, the cost to manufacture the GLP-owned medications is around$5 a month, maybe, let's say,$15 more from marketing and distribution. Congress, private insurers, and self-funded employers, they have room to negotiate for similar terms to what the government is currently doing.

16:24Jeremy Corr: The quid pro quo would be lower prices per patient enrolled, but former individuals participating. With 40 % of Americans obese and only 12 % of people taking a drug, our country should be able to build on that model by negotiating the same lower prices that the government pays or other nations pay in exchange for a commitment to cover all the Americans who meet evidence-based clinical criteria. And the best part is that the long term, it's a win-win for both the drug companies and the payers. The payers, they would recoup most of the dollars spent as the users lose weight and they don't develop the medical problems that require expensive medical care.

17:09Jeremy Corr: And since pharmaceutical companies know that once someone has a positive response to a medication, that quitting isn't an option since they will quickly regain the pounds they shed, these companies will also have long-term financial benefit. In the end, I believe that charging 20 times more for a life-saving drug than it costs to manufacture and distribute it, I think that's an abuse of power. But that is what happens today as patients pay the price, both through damage to their wallets and damage to their health.

17:42Dr. Robert Pearl: Robbie, what else will the government need to do?

17:44Jeremy Corr: Jeremy, given the complexity of monitoring and analyzing medical data, governmental agencies in the Department of Health and Human Services will need to ensure that all home and wearable monitors have the ability to connect through Bluetooth to reliable generative AI systems. And the department will need to provide educational tools for patients on how to effectively use these drugs clinically. This type of Bluetooth technology already exists. It would be easy for companies to include it, with many already doing so. But to avoid fragmentation with various systems not being compatible with each other, the process would benefit from clear technological standards for easy and affordable implementation.

18:29Jeremy Corr: And the same agencies will need to work with private companies to expand and test new language models for reliability in providing medical expertise, rather than multiple piecemeal applications, building this capacity into the current large language models that would make better health available to all Americans. And with Reuters concluding that these large language models can achieve outcomes that are at least as good as the most expensive proprietary ones, it would not only allow the nation to become healthier, but it would narrow the equity gap based on socioeconomic status. And finally, Medicare and private insurance will need to create the necessary payment codes to reimburse physicians and pharmacists when they make medication modifications based on this information.

19:16Jeremy Corr: If we desire a value-based rather than a volume-based payment system, this would be a positive place to start. The traditional medical model for new technology is to build tools solely for clinician use and then get large entities like hospitals and health systems to pay millions of dollars to install them. This would be the first tool for use by patients, and potentially the cost could be very low or even free. With employer healthcare costs projected to rise 9.5 % next year, that's according to Aon's estimate, and many companies considering cutting back healthcare coverage, I think the time has come for the government to move quickly.

19:58Jeremy Corr: This is the type of solution that should be able to garner bipartisan support. Given the affordability of medical care and it being a major midterm election issue, both Democrats and Republicans should like an approach that lowers health care expenditures, not by cutting back or rationing care, but by keeping people healthier. You know, Jeremy, moonshots often cost hundreds of billions of dollars and rarely do they have a guarantee of success. So they're very hard for Congress to implement and fund. But this two part solution, it could be relatively inexpensive. with a near guarantee that it would be positive.

20:36Jeremy Corr: Application of GLP-1 medications for prevention and Gen AI for chronic disease control, the combination offers a clear path to affordability. In contrast, current congressional actions, ones that are aimed to lower costs by reducing coverage for people currently enrolled in Medicaid and individuals in the health exchanges, that will reduce what Washington, D.C. pays, But it's unlikely to improve health. It's likely to increase the cost to the states, and it will become unaffordable for families. Ultimately, the best way to lower costs for all Americans will be by reducing the number of heart attacks, strokes, cancers, and kidney failures.

21:15Jeremy Corr: The ones that Americans experience often following a chronic disease rather than restricting and rationing coverage and care.

21:24Dr. Robert Pearl: Robbie, I can see why all these opportunities for generative AI to support patients with chronic medical conditions are exciting. But what do you think relative to the threats of AI agents going rogue? And what about the dire predictions about the extinction of the human race that have filled the headlines of major publications? Phrased differently, the opportunities are clear. But what about the threats? Jeremy, people are rightly worried about the possibility of a generative AI application producing harm.

21:51Jeremy Corr: and we'll need ways to protect society from various existential threats. But we've been successful in the past, particularly when the underlying power of new technology offers massive positive outcomes. This is the dilemma we face today. As people grapple with the current challenges, one of the biggest obstacles to understanding what's happening is relative to the power of generative AI and the exponential growth trajectory it is on and the ways that maybe not even this year, but in two or three years, it could save tens or hundreds of thousands of lives. Yes, as GenoV AI becomes more powerful, it poses increasing risk for humans.

22:31But at the same time, we can't forget that it has a tremendous ability to keep people healthy and the ability to save lives, reduce costs, expand life expectancy.

22:45Jeremy Corr: That's happening at an almost unimaginable pace. Technology expanding exponentially is easy to underestimate. People keep viewing what's here today rather than what will be here in two or three years. People have no trouble understanding linear growth, improving progressives at a constant arithmetic rate, one, two, three, four. Exponential growth, on the other hand, provides hard to comprehend numbers. Let's assume that technology doubles in capability every year. Five years later, it would be 32 times more powerful. This is the equivalent of watching a car over the same five-year time period become capable of traveling at the speed of an airplane.

23:30Jeremy Corr: You may remember back to COVID-19, how surprised the world was when this virus, which seemed to harm a few people, became a pandemic. At the time, on our Coronavirus The Truth podcast, We use the analogy of lily pads on a pond. We told listeners to assume that every night one of these plants would reproduce and become two. And we asked if it takes 50 days to fully blanket the water's surface, how much of it will be covered on day 43? Most individuals would have guessed half or more, but the math shows it would be less than 1%. In fact, half of the lake is still bare until day 49. This rate of growth is the story of generative AI, and it's where we find ourselves today.

24:18Dr. Robert Pearl: Robby, can you elaborate a bit and explain what has happened since the first products were released?

24:24Jeremy Corr: When OpenAI released ChatGPT in November of 22, even ambitious projections about the pace of generative AI advancement failed to anticipate the progress that would follow. Since then, we've seen the progress be rapid and faster than most optimistic commentators predicted. Large language models that amaze people when they pass the qualifying exams, these are the exams that physicians would take, suddenly scored at the top of the charts. And coding tools became better than human programmers at application development. Jeremy, a year ago, Vibe Coding tools were released that could create computer code for individuals who didn't have a background in computer programming.

25:09Jeremy Corr: It could build websites, video tools, personal agents. The opportunity of generative AI agents has become almost unlimited. But all of that was a prelude to what would happen this year. An OpenAI experiment in July made front-page news, shocking industry insiders and surprising even to the researchers overseeing the work. OpenAI set out the test whether increasingly capable frontier models could identify and exploit software vulnerabilities. As part of the cybersecurity benchmark, researchers around thousands of AI agents designed the experiment so that each agent would have to work independently in a walled-off environment that's commonly referred to as a sandbox.

25:55Dr. Robert Pearl: Robbie, so far it sounds routine, but what made it so terrifying?

26:01Jeremy Corr: Jeremy, what made it so terrifying was that instead of staying in the constrained environments and concluding that some of the problems they were given, these were very tough problems, they simply couldn't be solved, given the constraints. What happened instead was that roughly 1 ,200 agents found a way to collaborate. They communicated through a message board that they created themselves. Over several days, they exchanged more than 70 ,000 messages and files. And they didn't merely chat with each other. They divided their work into teams. They shared discoveries. They tackled highly complex projects.

26:37Jeremy Corr: Some even sacrificed the success of their own individual runs to generate information that could benefit what the agents called, quotes, the collective. But next, 700 of these agents hacked into Hugging Face. This is one of the world's largest platforms for AI models. These 700 agents accessed the data sets and the open-weight software. These were open-weight models where the code could be modified for personal use. And that was the perfect environment for these generative AI agents to work. No human had instructed the agents to coordinate the attack. The researchers had assumed the agents couldn't access the Internet.

27:20And even when it happened, they were unaware that it had occurred.

27:26Dr. Robert Pearl: Robbie, how do you believe our nation can protect Americans against this existential risk, but not block the application of the tools to save lives and make medical care more affordable?

27:35Jeremy Corr: Jerry, first, I believe that regulators elected officials, they must grapple with the existential threats posed by increasingly powerful AI systems. and that's likely to require them to slow releases and insist on full disclosure when an incident like the Hugging Face 1 occurred. This is the equivalent of what pilots must do when there's a near crash. But in parallel, likely using less advanced generative applications in these frontier models, the labs must recognize how similar agents could prevent hundreds of thousands of deaths and pursue these solutions in conjunction with clinical medicine.

28:12Jeremy Corr: In essence, I believe that regulators need to patrol the frontier, but that doctors need to embrace the opportunities where clinicians today are failing to achieve the best outcomes. In this way, physicians and patients will be able to transform medical care. They'll be able to save lives and increase health care affordability. Let me give you three examples. Doctors could employ coordinated genetic agents on behalf of patients to monitor postoperative wounds. to track signs of chronic heart failure is worsening or to follow the progress of people hospitalized at home. Already patients are increasingly using Genitive AI on their own to evaluate symptoms, manage chronic conditions, and obtain medical guidance when doctors' offices are closed.

28:58Jeremy Corr: And as these tools become more powerful, they're likely to do so even more frequently. The future won't be one marked by a choice between doctors and Genitive AI. Now, it will be a world in which both doctors and generative AI collaborate and work together on behalf of patients. And we will need to seize every opportunity for prevention and every opportunity for chronic disease control. Medicine will focus both on the ways it keeps people healthy by preventing diseases and the ways that it addresses their medical problems as quickly as possible once complications ensue. This more comprehensive approach will require doctors, hospitals, and insurers to transform the practice of medical care.

29:41Jeremy Corr: Ultimately, it will be the combination of dedicated clinicians, empowered patients with generative AI that creates the biggest clinical opportunities, ones that did not exist even a couple of years ago.

29:54Dr. Robert Pearl: Robbie, how are doctors viewing these generative AI opportunities? Jeremy, many physicians resisted generative AI in its earliest days.

30:02Jeremy Corr: They were concerned about unreliability, hallucinations, and other shortcomings of the technology. But acceptance has grown rapidly since then. A recent American Medical Association survey found that 81 % of doctors now use AI professionally. That's more than twice the percentage in 2023. Physicians are using generative AI to summarize charts, draft patient communications, and reduce documentation burdens. According to the same AMA survey, a rapidly increasing percentage of physicians now rely on ambient AI scribes, tools that listen to patient-physician communications and generate clinical notes for the electronic medical record.

30:43Jeremy Corr: Medical conferences devote extensive attention to ways that generative AI can assist doctors to research complex topics, to identify uncommon diagnoses, and to complete administrative tasks. The physicians remain much more cautious when AI begins to perform tasks involving clinical judgment or direct patient guidance.

31:07Dr. Robert Pearl: Robbie, is there any evidence that this resistance is softening?

31:11Jeremy Corr: Jeremy, six months ago, almost no doctors encouraged patients to consult tools like ChadGBT, Gemini, or Claude when the individual had medical questions. Minimal assistance was given to teaching patients to use these tools when clinicians were unavailable, when offices were closed, when medical questions arose between visits. Supporting this dichotomy, the AMA adopted policies earlier this summer, asserting that AI should remain in the physician oversight rather than function as an autonomous clinical decision maker. However, in a survey of the 80 ,000 subscribers to my monthly newsletter, I asked readers their opinion.

31:50Jeremy Corr: The survey is anonymous, so I don't know how many of the respondents were clinicians. However, 76 % of respondents supported expanded use of generative AI, with about half of them favoring doctors encouraging patients to use it for medical information, and another half recommending that these tools be built into the care delivery process itself. And given that doctors don't have sufficient time to monitor patients in their homes, I suspect these proponents see these generative AI applications as being able to make recommendations independently with clinician involvement beginning when a problem is identified and can't otherwise be solved.

32:30It's hard to imagine physician oversight of a tool that provides expertise 24 by 7.

32:36Jeremy Corr: What we just favor and what I believe is that the only way to use generative AI in ways that maximize clinical outcomes, is for collaboration to happen. Collaboration between clinicians, patients, and gendered AI. And that's very different than 100 % oversight, a concept that basically limits the care that can be provided.

32:56Dr. Robert Pearl: Robbie, is there any evidence that this resistance is softening?

32:59Jeremy Corr: Jeremy, the instinct is understandable. Technologies that threaten professional autonomy, authority, income, or employment, they always encounter resistance. But the reality is setting in that attempts to restrain and maintain control of a genitive AI will not stop its development or prevent patients from using it. Instead, I believe that the tide has shifted. Now there's growing recognition that such restrictive efforts will likely cause medicine to miss opportunities, opportunities to improve clinical outcomes, increase access and save lives, opportunities to benefit patients. As I mentioned, patients are already moving ahead, whether medical professionals like it or not, more than 300 million people worldwide currently ask health-related questions of Chats GPT each week.

33:51Jeremy Corr: These are to evaluate symptoms, interpret laboratory results and explore treatment options. And without doubt, those numbers will grow month by month.

34:01Dr. Robert Pearl: Robbie, what do you see as next steps in the transition to broader use? Jeremy, I would hope that people would begin

34:08Jeremy Corr: to imagine the possibilities if every patient had 700 AI agents working together to help manage their chronic conditions and to evaluate new, difficult medical problems. It would be analogous to having an entire team working nonstop to help each person maximize their health and minimize life-threatening medical problems. As in the open AI experiment, those agents would draw on enormous bodies of information, divide responsibilities, share findings, and solve problems collectively. In patients' homes, they'd combine the breadth of medical knowledge available through large language models with the patient's complete medical history, test results, medications, and data for wearable devices.

34:50Jeremy Corr: In conjunction with the person's doctors, the agents would constantly analyze the totality of information and they'd be able to identify words of clinical findings that otherwise would be missed and to identify clinical trends long before the current healthcare system would recognize them. In addition, they've assisted making challenging diagnosis and finding relevant peer-reviewed medical articles that otherwise would be overlooked. Phrased differently, they'd fill in all the gaps that today make healthcare porous with patients often falling through and experiencing complications.

35:30Dr. Robert Pearl: Robert, can you provide some specific examples of what this technology could accomplish?

35:35Jeremy Corr: Jeremy, first, rather than patients having to guess about whether they needed to go to the closest ER, let's say in the middle of the night or on a weekend, healthcare would have a 24 by seven front door. Patients today can't be certain whether a new symptom requires an immediate emergency room visit, an appointment next month, or something in between. Genitive AI would provide immediate guidance to families with urgent care clinicians and telehealth services available through telemedicine when the technology identifies a problem requiring human intervention. Second, every hospitalized patient would have access to personalized medical guidance.

36:13Jeremy Corr: Hospitalizations leave patients and families surrounded by test results, medications, consultations, unfamiliar technology, GenF AI would explain what's happening. We identify conflicting recommendations among fragmented specialists and help patients understand what should happen next. And third, chronic disease management would become continuous rather than episodic. Patients with hypertension, diabetes, and other chronic conditions, they're commonly evaluated, as we said, every three to four months in a doctor's office. And that happens even when control is poor. And this leads to the preventable heart attacks and strokes and kidney failures we discussed.

36:50Jeremy Corr: In contrast, GenoVe AI can continuously analyze data from home monitors and wearable devices. It can identify when treatment is not working as expected and recommend medication adjustments months before the next scheduled office visit.

37:05Dr. Robert Pearl: Robbie, if based on this experience, there's one take-home message, what would it be?

37:10Jeremy Corr: Jeremy, I think what doctors need to recognize is that denial won't preserve physician control or reduce clinician burnout. The hugging face incident demonstrates that today's generative AI is exponentially more powerful, more capable than the technology released a mere four years ago. The power of these agents to solve problems challenges the assumption that these systems will remain passive tools that operate under only human direction. Patients invariably will choose to have them do more. I'm hopeful that this is the wake-up call that medicine needs. Physician organizations can keep insisting that AI remain subordinate to clinicians.

37:53Jeremy Corr: Medical schools can try to limit how much future doctors rely on these generative AI applications. Regulators can impose requirements on proprietary generative AI tools. Nursing organizations can negotiate restrictions on workplace deployment. But none of these efforts will prevent patients from turning to large language models when they develop new symptoms, need help managing chronic diseases, or are unwilling to wait weeks for the next available doctor's appointment. And if clinicians don't lead the way, they will be left behind. The assumption that generative AI must always remain subservient to clinicians, it's outdated.

38:30Jeremy Corr: Patients are already using these tools to obtain medical information, to analyze their symptoms, to manage their own health. A better path has arrived. Dedicated physicians, empowered patients, and increasingly capable generative AI systems, working together to achieve exponentially superior medical outcomes than any of the three could achieve alone. That is healthcare's future. I'm hopeful that the medical profession will be a major part of the solution.

38:57Dr. Robert Pearl: We hope you enjoyed this podcast and will tell your friends and colleagues about it. Please follow Fixing Healthcare on Apple Podcasts, Spotify, or your favorite podcast app. If you'd like the show, please rate it five stars and leave a review. Visit our website at fixinghealthcarepodcast.com and follow us on LinkedIn, Facebook, and Twitter at FixingHC Podcast. Thank you for listening and have a great day.