Share This Podcast

In a time when AI is creating new realities faster than we can make sense of them, we need to imagine possible future scenarios to prepare. Which is why a new novel by Marschall Runge, Dean of the University of Michigan Medical School, is so prescient. In his book “Coded to Kill,” Runge helps us imagine what could go wrong, even if so much can go right. Abdul reflects on the critical role of imagination in science. Then he sits down with Dean Runge to talk about what his book can teach us about how we prepare for a future of artificial intelligence in medicine.

TRANSCRIPT

[AD BREAK} 

Dr. Abdul El-Sayed, narrating: A new AI based startup says it can detect sexually transmitted infections by looking at dick pics. A CDC study finds that alcohol related deaths are increasing among American seniors. And a new case of bird flu was linked to a dairy cow. This is America Dissected. I’m your host, Doctor Abdul El-Sayed. Happy public health week, everyone. [music break] We’ve covered lots of novel science here on America Dissected, but a novel about science? That’s new territory for us. That said, it’s not every day that a dean of one of America’s most prestigious medical schools writes a novel, which is what forced me to perk up when I heard about it. Before we talk about the book, though, I wanted to explain why we should probably talk more about fiction when we think about science. Science is, of course, about what’s observable, what is known and knowable. So you’d think that fiction, literally content that is not real, would be orthogonal to it, but they’re actually a lot more related than you might think. We’re used to thinking about science in hindsight, using what science has already demonstrated to make decisions about what we ought to do for the future. But that’s what we do with science after it’s already been done. But if you’re doing science, you’re by definition dealing with the unknown. I know it’s important to replicate experiments to make sure we know what we think we know. But scientific breakthroughs happen specifically when we experiment with the unknown. When you’re dealing with the unknown, the most powerful thing in science is how you can use what is known to imagine what is unknown. And when you’re dealing with the unknown, the most powerful tool you have is how you use what you know to imagine what you don’t know. That’s how the best hypotheses are generated. And as any real scientist will tell you, the most powerful thing in science is a great hypothesis. So science really is about imagination, and we need it most when things are moving fast. Right now is one of those moments. 

[clip of unspecified reporter] Last week, the artificial intelligence company OpenAI unveiled ChatGPT4. 

[clip of unspecified reporter 2] The man widely seen as the godfather of artificial intelligence has quit his job at Google, warning of the dangers of AI. 

[clip of unidentified speaker about AI] If this technology goes wrong, it can go quite wrong. Uh. And we want to be vocal about that. 

Dr. Abdul El-Sayed, narrating: AI is fundamentally changing the horizon for science and health care and everything else, too. It’s a lot more than just ChatGPT. When you can train an algorithm on every single chest X-ray that’s been done, you better believe that that algorithm is going to get really good at finding things you couldn’t even imagine finding in a chest x ray. When an algorithm can liberate a physician from the tyranny of their patient portal inbox. You’d better believe that physicians are going to feel some kind of way about it. That’s the upside. We had Doctor Eric Topol, one of the world’s foremost health care futurists, on to talk about AI in health care a few months back. It was a stimulating conversation. Some of you might remember that he was a lot we’ll say more bullish about AI than I was. I’m a child of the internet era. I remember watching one of the first educational live feeds of a trek in the Amazon rainforest back in the sixth grade in 1996. And despite the glitchiness of dial up, I remember marveling at how we could be watching each other in real time. And nearly 30 years later, I’m not so sure that I love the world that unparalleled connection has wrought. So forgive me if I’m a lot more focused on what AI might take away. How it opens up opportunities to be exploited by malign actors, fundamentally changes our relationship with our work, and creates scenarios we can’t even imagine from here. Which is why we need to try so hard to imagine them. And that’s exactly what Dean Marschall Runge is doing in his book, Coded to Kill. In a fast paced whodunit style, Runge explores what happens when an electronic medical record is exploited by the worst people. Speaking of EMRs exploited by the worst people. 

[clip of unspecified reporter] United Health says that it is aiming to bring systems back online by today after a massive health care hack described as the most serious on the U.S. health care system to date. 

Dr. Abdul El-Sayed, narrating: Yeah, that’s why I wanted to have Dean Runge on to talk to us about how leveraging our imaginations can help us see around the corner as health and health care change. About why he wrote this book, and about what it should teach us about the opportunities and threats of AI and other technology in our work. Here’s my conversation with Dean Marschall Runge. 

Dr. Abdul El-Sayed: All right. Can you introduce yourself for the tape? 

Marschall Runge: My name is Marschall Runge. It’s really great to be with you here this evening. 

Dr. Abdul El-Sayed: Thank you so much for taking the time. I um, you know, you and I have known each other for some time, and– 

Marschall Runge: Right. 

Dr. Abdul El-Sayed: You know, I joked before we started recording that we’re we’re probably recording from, like, not more than a couple miles away from each other, uh and, um and that’s because you’re the, the, the dean of the University of Michigan, medical school, as well as um the vice president for medical affairs at the University of Michigan. And what I’m really excited about is that you’ve written a book, but a lot of deans write books. This one is um is a novel. And I want to ask you right, to step back. Just tell us I want to understand the sort of how you the journey to writing a novel. So tell me a little bit about your day to day as a dean and as a VP of Medical Affairs. Like what is your day to day? What are the things that you work on? Um. What are the things that you got to be thinking about uh when you run a large, health care system like the University of Michigan’s? 

Marschall Runge: Well, I’ll start by saying I just have terrific people to work with, which otherwise it’d be impossible to do a job like this. But, my time is spent part of it’s academics with education, particularly medical student education and graduate medical education, with research and working on building and sustaining our terrific research programs which we have. And uh then, uh part of it is I do, I continue to do a little bit of patient care. I’m a cardiologist, but my main uh efforts in health care are overseeing our health system, which is a large health system. What I like most about it is that I can believe that everything we do is oriented toward trying to improve health. So we’re trying to educate physicians, nurses, allied health professionals to provide health care. We’re doing research that we hope will be impactful for health, and we’re providing health care. And I think that the, most important area and with your background in public health, in addition to your being a physician, I think the future for us is to try to improve health. And if we improve health, we may be able to afford the health care that we provide because, so much can be prevented. And it’s a hard thing, but uh, I think in collaboration with our colleagues here in public health and across the country, we’re very interested in trying to just help improve the, the health and our main impact, of course is in Michigan, and even more so in Southeast Michigan. So all those things tie  together, and on any given day, I may spend um half the day, uh working on certain aspects of our health system, either here in Ann Arbor or across the state. Another day I’ll I’ll spend half a day, in an educational setting. 

Dr. Abdul El-Sayed: You know, the job like you talked about is is really quite multifaceted. You spend a lot of time, I imagine, in meetings. I–

Marschall Runge: Yes. [laugh]

Dr. Abdul El-Sayed: I spend my day leading a much smaller institution than than yours, and I spend almost all my day in back to back to back to back meetings. And one of the things that I find is that you get really good at solving tactical problems, being able to sit in a meeting, try and boil it down to the next thing that needs to get done, and help to guide a team, uh to to doing that, and then following up and checking in and seeing how a process is going. And, you know, if you’re doing it really well, all of those tactical things on a good day add up to a strategy and then, you know, if you’re doing it really, really well, then all those strategies add up to a vision, uh for something that changes on orders of, of of time, of, of years. And the the thing about it, though, is that it really does come down to a lot of like tactical small decisions. Now, one of the things about writing and particularly literary writing, is that you have to architect a vision on a page. And so it really does take a certain, um leap of mind to get yourself out of sort of the tactical meeting movement that you spend a lot of your days doing to be able to look at a an empty page and then fill it out with something. Now one of the the cool things about what you’ve done here is that you’ve written a novel. And I want to ask, hey, how’d you find the time? But then B, why a novel, right? Like I said, it’s uncommon for folks in your area to to write uh fiction. You see a lot more commonly nonfiction books about, you know, an area of of research. But a fiction book is a really interesting piece. So see walk me through the thinking about what led you to this and and why you wrote the book? 

Marschall Runge: Well, it starts with I love reading thrillers. I like medical thrillers. Legal thrillers. And I that’s been sort of my escapism, for a long time. And so at one point, one crazy point, I had a, a person who came to see me as a patient, and he was a, he he could easily have been a down and out lawyer from a John Grisham uh book. And I had never heard of John Grisham. And so he we talked, he said, I’m gonna bring you a book. And I read that book, and I thought, man, this is awesome. I bet I can do this. Well, that was like 20 years ago, but I and–

Dr. Abdul El-Sayed: There you go. It’s a lot of meetings between then and now, though. Just to be fair.

Marschall Runge: [laug] That’s right. So this book, I worked on it off and on for, probably 15 years, and–

Dr. Abdul El-Sayed: Wow. 

Marschall Runge: It was 90% done when I moved to Michigan. But trying to finish that last 10% took a lot of time. So what I found about it is I love writing fiction. It’s just fun and it’s interesting. You can write whatever you want, and there’s no known risk of, excuse me expanding on the truth. Because it’s not truth it’s fiction. And so that’s how I got going on it. And I would write in the morning and, and sometimes on weekends. I’m an early morning riser anyway. But then at some point, I here when I came here, Mark Schlissel, who was the president of the university at the time, said we need every executive to go out and do something to help um promote the university. And some people go to Washington and some people get involved in politics. Some people get involved in delivery of public health, all kinds of things. But I decided I would try to write op eds. That was one of the choices, and I enjoyed that. But I realized it just wasn’t having much impact and I thought huh? Op eds; novel. So I thought, well, maybe I could talk about what I found to be really important topics in a way that people who read fiction might read it, and it’ll trigger them to think about these areas. That was part of it. Another part of it is this this novel is about things that could go wrong with electronic medical records. And it’s very analogous to. What we are looking at today in terms of this transformation with artificial intelligence. This was an earlier version of connecting medical records all over the place, but I was involved in having to deal with, not here previously with faculty members who had inappropriately accessed medical records of a high profile person. And I I realized how easy that was to do. And, you know, now things are much more secure now. But I thought, well, if all this promise of having all this genetic data and incorporating it into health care, the benefit if you have a huge electronic medical record is you you have all that data, you can try to assimilate it. And I think AI will push it way faster. But the other concern is what if somebody gets into that database? What if they decide to do bad things with that database? Then as I learned that all the devices in a hospital are connected. So I didn’t realize this. But if if someone writes, if a doctor writes in order for an IV solution, it doesn’t go to a pharmacist anymore. It goes to a robot. And those robots are connected to the letter medical record. And so if somebody was in a medical record, they could presumably disrupt what was going on. So that I tried to build on that premise, but making it a fun, exciting kind of beach novel, but one that had a message, which is, let’s be careful as we think about gigantic databases and protecting, protected health information, which I think is for doctors is just sacred. 

Dr. Abdul El-Sayed: I want to I want to get back to the question of of doctor as um uh, as novelist later on. But I do want to turn to the content of, of the book. Right. Because you wrote a very particular book sitting at this intersection between complex bureaucracy, the Internet of things, and AI and connectivity. As you think about where we’re headed what is your general approach? What is it that you want the public to understand about the intersection of these things and the potential risk that you lay out quite uh neatly and I would say in a very thrilling way, in the book. 

Marschall Runge: Well, thank you for the for the kind words about the book. So here’s what I’m worried about. Where we’re currently in in what has been called a Promethean moment. And I had to look I had to look up well who is Prometheus. So and Thomas Friedman, who’s a well-known writer for The New York Times, I like to use his description, which I’m not going to quote exactly, but I want to attribute it to him. He said, we’re at a we’re at a point in time when there’s a technology that’s coming in that’s going to that is such a shift, it’s going to change everything we do and the way we do it. And others have computer science experts here at the university have said this will be well this will be more impactful then the formation of the internet. So, you know, you think of wow, they can compare 1990 to now and what we can do on the internet. But this is going to be yet another step forward. And the power is unbelievable. Anybody who’s played around with ChatGPT or any of those, it’s just mind boggling. But what it can do in terms of, generative AI. But the real power in medicine is going to be analyzing vast amounts of data. And so that takes vast numbers of people and all kinds of data. But I’m I’m convinced that generative AI will be able to analyze and put together patterns that we couldn’t possibly do before. So there’s huge promise there. Overall, what worries me is two two factors. One is uh the old adage that about computers. Garbage in, garbage out. These are entirely dependent on the quality of the data that goes in. Make sure that’s high fidelity data that it doesn’t carry biases that might have been part of creation of that data. Um. So the fidelity of the product that comes out is really important. Secondly, the by its very nature, generative AI can put together patterns that we could never put together, even with the most powerful computers in the world. And I had heard this five or ten years ago, and I believe it is absolutely the case now that even before AI, a computer scientist gone gone rogue at a big data aggregator like Google or Amazon. They know who we are. They could know who I am. I think that gets multiplied by a thousand when AI starts to be able to analyze these data. And then the last part and you know, this is a position I know this is a position is that people are different, everybody’s different and patients are different. And so there was such a there has been such an emphasis on care protocols. So a 60 year old has these risk factors for cancer or for heart disease or whatever. This is what you do. You do this test and that test and the next test. And and then what AI might be able to do is say, well, here’s what how you might treat that person, but we all know that maybe that person doesn’t want that treatment, or maybe that person has a real fear of certain kinds of medications or certain kinds of procedures or taking vaccines. And as we. I don’t see how AI can bring that all together without that really sort of personal interaction between physicians and their patients, which I think stands the potential of being lost. 

Dr. Abdul El-Sayed: Yeah, I really appreciate a couple of the threads that that that really link a lot of those worries. And I agree very much with them. I think the biggest conceit of AI is that somehow, because of its profound pattern recognition capacity, that it’s going to buy us out of the mistakes we make. But I actually think that the problem here is it’s just going to make those mistakes faster, and it’s going to do so in a way that are less recognizable. And I think the the challenge with it is that we look at our systems now and say, oh man, that system is broken because so and so human biases and because it’s happening at a slower pace, we can actually see the way that our biases shape those systems and then try and fix them. The problem is, is that when all of that’s happening inside of a black box at near warp speed in ways that are incoherent to us, or at least not identifiable to us, that we are going to assume that somehow this is all on a higher order, rather than, as you said, junk in, junk out. Right. Or a function of the problematic incentives that shape the system as it started and then leveraged in a certain way and I, you know, had a good conversation with Doctor Eric Topol, whom I know, you know, and is one of the foremost futurists in medicine and one of the, the, the, the challenges that we sort of came to and he agreed with. He’s a much more of a techno optimist than I think you or I are, is that there’s a real risk that the way that AI is moving right now is that’s being captured by a small number of very, very, very large, very powerful organizations. And that’s actually going to be even harder given how much computing power you need to be able to train AI models. And so you’re going to have like the top 0.000001% of corporations who are then leveraging this incredible power to their own incentive set. And I worry that that’s just going to make it a lot worse. And one of the things that I loved about the book is that, you know, because you’re able to create characters who sit on different sides of a system, you you’re able to almost slow down and demonstrate the frustrations of that from the perspective of, of a different, um a number of different perspectives of these protagonists in the book. I want to, you know, as you think about this, you illustrate some of the big dangers. And obviously it’s a thriller that’s intended to, entertain. But but the kernel of truth upon which it’s based is, is really quite real. So I wanted to ask you, you know, what did you find, um you were able to say in a book as, as your imagination sort of took hold and you wrote this thing out that you wish more people were saying in real life?

Marschall Runge: That’s a that’s a profound question. And thank you. And nobody’s asked me that. Um. What I like about fiction so much is they take something that is true, and then they just make it big. And so, for example, what I learned and and I, I owe any any technical knowledge I have in these areas, I know to I owe to other people, but I had lots of great conversations. So I, I for example, I said, would it really be possible for somebody to, hijack a pharmacy robot or would it really be possible for somebody to erase a person’s allergies, from their electronic medical record without a trace and even put in a history that makes it seem like they don’t have any, really severe, potentially lethal allergies. And they walked me through. And, yes, it is possible. I hope and I believe that we have the systems to prevent that and that there’s nobody so nefarious that would try to do it. But I want people to understand that what we do is not risk free, and we do depend on systems that work. I mean, they work every single day, but they could be caused to malfunction. I don’t know. I have yet to read any reason that the AT&T network went down in a big way last week. Something happened there. But I am very familiar with instances where, electronic medical records have been hijacked by bad actors who, either insert ransomware so you can’t use your medical record, or they even insert disinformation, or they grab the information from your medical record and they put it on the dark web. And, I think if you think about any of those. What and what I end up believing is, man, we have to redouble our efforts to protect all of this information. And so this novel is sort of I don’t know that it’s the worst case scenario, but it’s a bad case scenario of that happening. And, you know, I think, for example, University of Michigan, Michigan Medicine, we get over 500,000 attempted hacks into our medical record every day. 

Dr. Abdul El-Sayed: Wow. 

Marschall Runge: And those are all screened out. And so far, you know, knock on wood, thank goodness. Whatever. We have not had a hijacking of our electronic medical records. But if you think of it, even even just from a purely day to day logistics, when you and I were training. Well, maybe you’re not as old as me but when I was training, everything was on paper. So you wrote medical orders on paper. Medical records were on paper. Now there’s we don’t even have the paper to put I mean, we do have a sort of disaster plan, but uh if all those things grind to a halt, you can’t write medical orders, you can’t get laboratory results. It’s really a bad situation for people that are in the hospital. There was a case of a hospital that was hijacked for, ransomware for about a week. And, you know, basically all they could do was try to transfer the patients out of the hospital because they couldn’t really care for them. [music break]

[AD BREAK] 

Dr. Abdul El-Sayed: You know, it’s it’s interesting because this example exactly illustrates one of the real risks I worry about, and one that you actually somewhat highlight with one of your uh characters, uh who is a medical resident and a daughter of the hospital president, and she’s one of the protagonists in the book. And we’re it’s almost that we’re at a tipping point. So I graduated medical school in 2014. Uh uh. Marshall what [?]. 

Marschall Runge: Wow, you’re a lot younger than me. [laughing]. 

Dr. Abdul El-Sayed: When did you graduate? Would have been?

Marschall Runge: ’83. ’83, ’84.

Dr. Abdul El-Sayed: ’83. Okay. So so in your time, I would imagine. And you correct me if I’m wrong, but but but but but computers themselves were just starting to make their way into clinical practice, I would imagine. Right. And for very, very few purposes, I would imagine. 

Marschall Runge: Barely I mean, for, for interfacing with uh pathology machines like to generate the [?] or whatever we called it back in the day. 

Dr. Abdul El-Sayed: Yeah. And then in my case, we had fully moved to to EMRS, but the EMRS weren’t as smart. They didn’t auto populate. 

Marschall Runge: Right. 

Dr. Abdul El-Sayed: And then my little brother is a first year resident at the University of Michigan. 

Marschall Runge: Yeah. 

Dr. Abdul El-Sayed: He’s a he’s an intern. 

Marschall Runge: Go blue. 

Dr. Abdul El-Sayed: Go blue. And he he just graduated, and EMRS today have AI built in so they can do a lot of complex calculation, which we somewhat take for granted. Now, the interesting thing is, if all three of us were in a situation where, like you talked about a ransomware attack happened, you would be able to operate in a scenario where we were still using pen and paper. I think my brother and I would be like, what is going on right now? Right? We’d have to find some other IT enabled capacity, simply because that’s a reality that that none of us have ever experienced. And I think the same is true for a lot of the calculation that happens inside a medical encounter. And I worry that we’re hitting a tipping point on which we’re losing the scaffolding on which we’ve built all of our knowledge. And so you end up in a scenario where when computers aren’t just enabling you to take a note, right, and store certain information and make sure it gets coordinated and goes to a certain place. But the the computer itself is actually doing a lot of the most complicated thinking for you in real time, that you lose the ability to understand what’s happening inside that calculation. And when that calculation goes away, you then realize that you were almost a glorified human I hate to say uniform for the AI, right? You kind of made the AI look human versus actually being the calculator inside, which is how we think about physicians. And, you know, and you talked about a lot of your role. And as a dean, your big role is to train the future of physicians. And I know a lot of that is about being able to bring the humanism back into medicine. I think that’s one of the best things that AI will enable, and I fully concede that point. It lets you be a human again rather than have to just–

Marschall Runge: Right. 

Dr. Abdul El-Sayed: –think all about all of this. But I do worry that when you get enough crutches, you forget how to do the basics. And, you know, I talked to to physicians, um my my father in law is a retired nephrologist. And he said, you know, physical exams are dead. I don’t see any medical students graduating who know how to do a good physical exam, because the reality of it is, you do the physical exam to generate a differential, and at the end of the day, the cost of doing the test is either lucrative enough or lucrative enough for the health care system or not expensive enough that you’re just going to do the test anyway. So a lot of these arts are sort of dead, and I worry that like that may be true for like a physical exam. Now, what happens when you’re talking about even generating a differential. Right. You can dictate a note and you’re going to get a differential, which I would argue is like the whole goal of graduating medical school is knowing how to generate a differential and knock down some of the most important ones, right. So what happens when that goes away? And like when we get this point where we lose our ability to do the things that physicians of yore used to be able to do? Do we have an ability to adjust, and do we even have an ability then to counteract the the AI inside? I’d love to get your sense on how is there a way for us to train physicians to be future proof, or should we just embrace the future where, you know, doctors aren’t really gonna have to think that much because the AI is going to do it for them?

Marschall Runge: To my mind that you’ve hit exactly the crux of the issue. So I’ll give you one example. Um. It was a great thing when the earlier versions of Epic and other electronic medical records were able to look at drug interactions, drug drug interactions. So potentially look at a list of medications and say, oh, maybe you shouldn’t prescribe this medication because it could interact with another medication or say, caution, you shouldn’t give that medication. You said your father in law is a nephrologist. You shouldn’t give it to a person, this person, because his renal function is abnormal. Um. But the first oh, I would guess for me, the first five or 10 years I was, I was at the University of North Carolina. And like Michigan, we had our own homegrown electronic medical record. Then we transitioned to Epic, but over about a ten year period, five years on our own. Five years on Epic there it did that, that ability for the computer to help us matured. But at first it it wasn’t. And I worry about that transition time with AI. So I think it’s going to be transformational. But it may not be perfect. It may not be perfect for a while. But I think what you’re hitting is an even more important point. And that is um as physicians, we have to understand the disease process. People sometimes ask me, why do we still have to study some of these more basic concept in medicine, concepts in medicine, and where I think that will move to a higher level. So it won’t be biochemistry so much, but uh it’s all about trying to understand the disease process. So if you get a crazy result, I hope that we will all think, wait a minute, that doesn’t quite fit with what I’m looking at and talking to. I don’t know how that’s going to how AI will fix that. And I, I really I think probably it can and I, I think but I so I think we need to be training our students and our house staff and everybody else to embrace it and use it. But to learn everything they can about medicine and embrace that as well. Well, one example that, I think a lot of us are concerned about, um has to do with an interface that’s really important. With the advent of electronic medical records and use of the patient portal. For many care providers, physicians and others they will they they get so many messages in that portal to respond to that it has really changed how much time they can spend with patients, because they have to go spend another two hours that day answering portal messages. So we’ve done a pilot in Michigan and these are being done all over to see could AI help us? And, it’s like any other use of AI. You can cut and paste so the patients, person sends a query that is two paragraphs. You can cut it in, cut it and paste it into AI and say, please respond. And it will respond. And well, what we have found is that those responses are excellent but they’re not perfect. And so but I worry that a person will get in a hurry. And, so at least I don’t know what percent, 5% or something that they’re giving the wrong advice. Maybe not that high, maybe 1%. So we have to be very diligent to look at that and it’s the same with analyzing images, radiologic or pathology images by AI. I do I think it’s uncontestable that that’s very much speeds up what a radiologist or pathologist can do because it, it can really screen out most things and say focus on this area. Um. But still, despite what the hyperbole in the, medical media as well as, lay media, you can’t just substitute a machine I don’t want I don’t want a machine making the final diagnosis on whether that thing I’m worried about in my chest is normal or not. I’d like to be sure a radiologist looks at that. 

Dr. Abdul El-Sayed: This is what makes it so tough is in medicine. You know, we, one of the old adages, uh is when you hear hoofbeats think horses. But sometimes they’re zebras. 

Marschall Runge: Right.

Dr. Abdul El-Sayed: And one of the things or the promises of AI or if if I could sort of sum up the promises of AI with respect to horses and zebras, is AI should help you clear the horses faster and identify the zebras more accurately. And if it were to do those two things, it would solve a lot of the problems in medicine. The problem, though, is that horses and zebras sound the same. And I worry that if so much of of medical practice is about identifying the zebras from the horses, and that’s what you know, that’s what you really look for in the most top notch doctors. Most everybody can, can, can address the horses. It’s the one who you want to find that one minute detail to hold on to and say, no, no, no, this one does not sound like a horse. I think this is a zebra. I worry, though, that, you know, as physicians spend less time having to suss that out themselves because the AI is doing it, we’re actually going to lose the ability to second guess the AI. Because here’s the thing. There are going to be times that, you know, you talked about that 5% or 1% of the times it’s wrong. That’s going to look like it’s saying it’s a zebra and us saying it’s a horse or vice versa. 

Marschall Runge: Right. 

Dr. Abdul El-Sayed: And I worry that over time, as the AI becomes more and more common, our ability both uh intellectually but also bureaucratically to remand the AI goes away. Right? And intellectually, it’s just that we just haven’t had to do the hard thinking because we just take the AI answer as rote. And bureaucratically you know, we talk about you always want a doctor on the loop now, it’s kind of like a self-driving car. We say we we believe that humans have to be on the loop because humans are better drivers than self-driving cars, which is actually not true. Like if you had a whole road of self-driving cars, yes, you would have accidents, but you’d probably have far fewer than then a bunch of humans driving who, you know, check their texts and [?] the radio and don’t pay attention and have fights in the car with their their significant other, whatever else it is that people do in cars, right? The automated driving is actually better. We just don’t trust it because it’s not normal. But there’s going to come a tipping point where enough of these automated cars reduce the cost enough that actually you could imagine a world where you get in trouble because you put your hand on the wheel. 

Marschall Runge: Yeah. 

Dr. Abdul El-Sayed: It was the car the automated car is better than you, and you made a bad decision. So you can imagine the same thing with doctors. And this is the moment that I actually think like it breaks all of medicine, which is to say, you’re a doctor who’s like, I’m sure this is a horse, not a zebra, or I’m sure this is a zebra and not a horse. But if I second guess the AI that I’m going to get in trouble, so I might as well just let the AI do what the AI is going to do. And that’s the thing I’m worried about. So I don’t I don’t know what happens to doctors as doctors moving forward because of this, this reason the the litigation, the liability involved, that second guessing a machine that on average is just more accurate than you are. 

Marschall Runge: Yeah, a big, big issue. Abdul, let me talk about a different aspect of AI because I think it’s very informative. So if you look up what what how how is AI going to help us, help us and help? One of the big applications is in the discovery of new medications. So we all know that there are certain kinds of whether their genetic or otherwise, medical problems that are said to be undruggable. Nobody can find, Pharma has been looking at it forever. They can’t find a drug that will inhibit that particular pathway and prevent that particular disease. So now many of the biotechs and big pharma have decided this has started several years, at least two or three years ago decided, well, we could use AI and we can just take our chemists and have them develop, the perfect drugs. And what they found was the coolest thing was that, AI would generate um chemical structures that nobody has ever seen, and they could be made. And so I’m familiar with one setting in which, in the period of a month or so, AI generated Iike 30 or 40 drugs that should address a specific disease. Uh. And that would have taken years with medicinal chemists. Now, not all of them work. They had to be tested. Uh. But I believe that the conclusion that the uh people who are experts in drug development, uh and in uh AI and these are, these are massive uses of computer power, by the way. Um. The conclusion they’ve come to is you can’t find the drug by AI alone. You have to combine AI with the biology of the disease. And so it’s foolish to think go down the AI path and then try to put some of those drugs in humans. If you don’t really understand the biology of the disease and put those together, and that the reason I give you that example is it is one that’s out there. But I think it I think if we could figure out and I don’t I can’t figure it out right now. How do you do the same with AI medicine. So you can’t just have the AI. The AI can terrifically, uh augment what a a doctor or nurse or another health care provider can do, but you have to put them together. Uh. And I guess, I’m, I’m really enthusiastic about the future of generative AI, but I think we have we’re going to have a period of time, I don’t know how long is it going to be? Five years or ten years while we’re still trying to understand how the two augment each other. I think I think you’re absolutely right. But whatever process might cause a zebra in me might only be a horse in you. And uh, we just have a lot to learn. 

Dr. Abdul El-Sayed: Yeah. No, I, I really appreciate that perspective, and I, I think there’s always going to be a role for humanity to do some things. The, the real question is how and then what are the dangers we don’t yet see. And I, I guess it might be just a function of my, my cohort. Right. I’m in that last group of people who remember a time before the internet, and I remember sitting in my sixth grade classroom and my sixth grade teacher telling me that the internet was going to be this amazing thing that that changed everything in my life. And he was right. It changed everything in my life. I mean, I get to host this podcast because of the internet. 

Marschall Runge: Right.

Dr. Abdul El-Sayed: And at the same time, I look at the world that the internet has wrought, and I don’t know that I like this world better, and that cognitive dissonance of the impact of this thing for me personally, versus the emergent impact across the scale of all of us. That’s the thing that breaks my mind a little bit. And I I think AI is just going to do that even more. And the analogy of breaking my mind that like that may be may just be it. Right? It’s just like [laughter] well, listen, like this thing that you relied on, that was your own personal supercomputer that all of us have between our ears, is just not as super as you thought it was. And um, and I think that should give us pause. It should also give us awe. And um, one of the things I appreciate about your book is that you force us to both to appreciate both of those at the same time. I want to just, you know, finish [?] to note, about about doctor as novelist. You know, one of the most powerful lessons that I learned in medical school, and I don’t practice medicine every day was the power of a story. Both the story that a patient tells a physician, the ability of a physician to understand that story and sit inside of that story with that patient, and then the story that the physician tells the patient when they give them a diagnosis and walk them through what, what the potential, outcomes could be next. Um. I want to ask, you know, as someone who’s now written a novel where you allow your mind to imagine a whole set of scenarios and you sort of pick the one you want to tell, how has that changed your medical practice? 

Marschall Runge: Um. Well, that’s another good question. So I think you’re right, though. Doctors aren’t inherently involved in storytelling or story understanding. And I’ll give a kind of terrible example about myself. Um uh I think what it has really encouraged me to do is to spend that extra time listening to people and what they have to say. When I’m seeing them and not rushing through it and thinking, okay, well, let’s do this, this and this. So the terrible story about myself is oh, 15, 20, 20, 25 years ago. Sometimes I’d be covering a massive cardiology service over the weekend, and it was all I could do to just see all the patients, much less that they needed something done, that I would do a procedure or something. You know, it was it was tough. And so I got in this mode where I would just try to go in and get just the [?] I needed and then scoot out of there. Um. And uh you know, I was dealing with with that shortage of time. And I think what it tells me is a few things. First, I think we’re the the the physicians of your generation are right. We shouldn’t be telling people to see an unreasonable number of patients. It’s it’s not good medical care. Furthermore, if you just listen for a little while, uh you’ll you’ll they’ll tell you things people will tell you things that they might not tell you otherwise if you bond with them. And and we know that about any kind of relationship. Um. And it reminds me of, you know, I am another generation or so back uh there were people who had during their careers, some of my professors had been in the era where you did a lot of house calls. And I don’t know if you ever heard these discussions, but they used to talk about this to us as students and say, you learn a lot about a person when you go to their house and you see where they live and you see what the environment is like, and we don’t do that anymore. Or rarely, uh but I think that kind of getting to know a person, really um it’s it’s good for the relationship. I think the doctor patient relationship is important to people and it’s important to health. Uh. But it also you learn things that you wouldn’t learn if you were if you were the old me just [?] saying, okay, any chest pain? No. Any problem with your groin? No. Where they had the catheterization? Great. Be back to see you later. Um. I mean, it wasn’t quite that abrupt, but it was pretty abrupt. 

Dr. Abdul El-Sayed: And, you know, the irony of all this is that if there’s one thing that I think everybody agrees that AI will mock is it may enable us to be human again. 

Marschall Runge: Yes. 

Dr. Abdul El-Sayed: And I think sometimes that’s that’s what was lost in having to see all these patients in so little time. Make decisions and then process them through the system when there is, you know, we call it health care. And there is something about that care part that um that is bigger than did you do the right procedure, get the right diagnosis and move that patient forward? It was this act of caring. There’s no other way to me to explain the placebo effect, except for that I think this thing will help me. And so much of that is in another human um taking the time to listen and care. And so I really appreciate you taking the time uh to share with us and tell us a little bit more about, your book, why you wrote it, the juxtaposition of intersecting trends at the heart of it, um and and where you feel like we go from here. Our guest today uh was doctor, Marschall Runge. He is uh the dean of the University of Michigan Medical School, as well as the vice president for medical affairs at Michigan Medicine. We really appreciate you taking the time. Thank you and once again, the book is Coded to Kill. I hope folks will check it out. 

Marschall Runge: Thanks very much. Really a pleasure and privilege to speak with you tonight. [music break]

Dr. Abdul El-Sayed: As usual, here’s what I’m watching right now. As we discussed today, AI is going to revolutionize the way we do health care. But I don’t think anyone thought it was going to go down like this. HeHealth launched back in 2022 with the we’ll call it audacious goal of using AI to detect the presence of STIs from dick pics. It was marketed to people with penises, hence the name. But they’ve just launched a new platform called Calmara, marketed to Gen Z women who can get this use the app to take a picture of their partner to check that it’s, quote, “clean.” To address the obviously creepy scenario, the app claims the pics are deleted immediately. Except that really doesn’t help. And it’s not just that, most STIs are asymptomatic. You can’t identify them on the shaft of a penis, so it’s bound to create a lot of false negatives. So what’s the upshot of all this? People sending a bunch of dick pics to an AI they have no control over and know nothing about. What could go wrong? The story highlights just how absurd the wild, wild west of health based AI can be, and the inherent risks of trusting a product simply because it’s got those two magical letters in it. Apps like this won’t keep you safe, but they’ll have you sending sensitive information to a cloud based supercomputer in the name of it. Moving on, a troubling new study in the Morbidity and Mortality Weekly Report, the CDC scientific journal, found that rates of alcohol related deaths are increasing in a group of folks we really usually don’t associate with aggressive alcohol use, seniors over the age of 65. The study found that rates of alcohol related deaths more than doubled over the past two decades. There are a number of potential explanations here. First, remember who this generation is. If you’re 70 this year, that means you were born in 1954, you were 20 in 1974. These are classically the boomers. And while boomer may take on a particular connotation these days, the boomers made quite the impression in their younger years. They were the hippie generation after all. Unlike their straight laced parents who bore them, they are a generation known for their substance use. And so perhaps it shouldn’t come as a shock that they use more substances as seniors too. But that’s not the only explanation. We’ve talked a lot about how the social milieu, driven by algorithmically amplified social media, has left young people more alone, anxious and depressed. But seniors use social media, too. And if you don’t believe me, just go on Facebook. And while they may have more wealth than life experience to cope, it’s likely that it’s having an impact on them too. We also just lived through a pandemic, a moment when seniors were among our most vulnerable people and therefore among the most likely to be socially isolated from their families. How to cope? Unfortunately, based on these numbers, it’s likely that a lot of folks found comfort at the end of a bottle. And when we talk about mortality, remember seniors are just more sensitive. Their bodies don’t rebound the same way, and chronic use over a long time horizon can have a cumulative effect. Finally this happened last week. 

[clip of unidentified news reporter] A person in Texas has been diagnosed with bird flu, which has prompted some concern. 

Dr. Abdul El-Sayed: We’ve been tracking this strain of H5N1 avian flu, the same one that led to a historic cull of egg laying hens and mink last year. Now it’s made its way into dairy cows and from there, into a human. Before you freak out, there are two really important things to remember. First, there’s little risk that those of us who don’t spend large amounts of time with dairy cows might get this via milk, for example. Though virus has been identified in the milk of infected cows. Unless you’re drinking raw milk our milk supply is pasteurized specifically to kill any potential pathogens it could carry. That said, it’s not impossible. What about meat, though? Interestingly, the virus has only been detected in dairy cows and not beef cattle. Not yet at least. Second, there is no evidence of human to human transmission, which is the most worrying potential outcome as efficient transmission between humans could kick off a serious outbreak or worse. But this is worrying, cows weren’t previously thought to be at high risk, though scientists speculate that the virus may have made an interspecies transfer through accidental ingestion of the feces of an infected bird by a cow. We know it’s affected at least 13 herds across six states, and though it’s possible that cows are independently being infected by birds, it’s also possible that it’s spreading between them. A more concerning possibility, and probably more likely. That’s also concerning because the virus would have more opportunity to evolve in its dairy cow hosts. Thankfully, analysis of the strain from the infected person doesn’t show the kinds of mutations that would make it concerning for efficient person to person spread. We’ll keep you posted. Finally, before we go, happy Public Health Week to everyone and Eid Mubarak to all of you celebrating. That’s it for today. On your way out, don’t forget to rate and review. It really does go a long way. Please take the time. And if you love the show and want to rep us, do drop by the Crooked store for some America Dissected merch. And don’t forget to follow us at @CrookedMedia and me at @AbdulElSayed no dash on Instagram, TikTok, and Twitter. [music break] America Dissected is a product of Crooked Media. Our producer is Austin Fisher. Our associate producers are Tara Terpstra and Emma Illick-Frank. Charlotte Landes mixes and masters the show. Production support from Ari Schwartz. Our theme song is by Taka Yasuzawa and Alex Sugiura. Our executive producers are Leo Duran, Sarah Geismer, and me. Doctor Abdul El-Sayed, your host. Thanks for listening. [music break] This show is for general information and entertainment purposes only. It’s not intended to provide specific health care or medical advice, and should not be construed as providing health care or medical advice. Please consult your physician with any questions related to your own health. The views expressed in this podcast reflect those of the host and his guests, and do not necessarily represent the views and opinions of Wayne County, Michigan or it’s Department of Health, Human and Veteran Services.