We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode Why AI Could Be the Key to Better, Faster, and Cheaper Healthcare with Dean of Stanford Medicine, Dr. Lloyd Minor

Why AI Could Be the Key to Better, Faster, and Cheaper Healthcare with Dean of Stanford Medicine, Dr. Lloyd Minor

2025/2/11
logo of podcast Smart People Podcast

Smart People Podcast

AI Deep Dive AI Chapters Transcript
People
L
Lloyd Minor
Topics
Lloyd Minor: 作为斯坦福医学院的院长,我的工作主要有三个方面。首先,我负责管理医学院的科研和教育工作,确保我们的31个部门、研究所和中心拥有杰出的领导者,以支持他们的研究、教育和患者护理工作。其次,我负责医疗系统的战略规划,包括确定我们的服务范围和卓越领域,以便更好地服务于我们的地区、国家乃至全球。最后,我负责对外事务,代表斯坦福医学和大学,并筹集资金以支持我们的工作。我每天的工作都充满了挑战,但同时也充满了乐趣。

Deep Dive

Shownotes Transcript

Translations:
中文

The Jack Welch Management Institute at Strayer University helps you go from, I know the way, to I've arrived, with our top 10 ranked online MBA. Gain skills you can learn today and apply tomorrow. Get ready to go from make it happen to made it happen. And keep striving. Visit Strayer.edu slash Jack Welch MBA to learn more. Strayer University is certified to operate in Virginia by Chev and its many campuses, including at 2121 15th Street North in Arlington, Virginia.

You know what's smart? Enjoying a fresh gourmet meal at home that you didn't have to cook. Meet Factor, your loophole in the laws of mealtime. Chef-crafted meals delivered with a tap, ready in just two minutes. You know what's even smarter? Treating yourself without cheating your goals. Factor is dietician-approved, chef-prepared, and you-plated. Pretty smart, huh? Refresh your routine and eat smart with Factor. Learn more at factormeals.com.

This is Smart People Podcast, a podcast for smart people, where we talk to smart people, but not necessarily done by smart people. Hello and welcome to Smart People Podcast, conversations that satisfy your curious mind. Chris Stamp here. Thanks for tuning in. Today, we are diving into the future of healthcare with one of the most influential minds in medicine. We are talking with Dr. Lloyd Miner.

He's a scientist, he's a surgeon, and by the way, he is the dean of the Stanford School of Medicine. In this episode, we are particularly exploring the intersection of AI and healthcare, how it's reshaping patient care, streamlining diagnoses, and even revolutionizing drug development. You will hear Dr. Minor break down the biggest challenges in the U.S. healthcare system.

explaining why we have one of the best sick care systems in the world, but not necessarily the best healthcare system. And most importantly, we're going to talk about how AI can potentially change that. We're going to cover the good, the bad, the hype from AI-assisted diagnosis to real-time clinical documentation to privacy concerns and biases.

There's a lot that needs to happen before we can trust this technology. And Dr. Minor, as the Dean of the Stanford School of Medicine, is one of the prominent people leading the charge on the ethical use of AI in our healthcare system. So whether you're a patient, a healthcare professional, or just someone who's fascinated by health, wellness, and where AI might take us next, you are going to want to tune in.

As always, the best thing you can do is share this episode. If you like it, if you learn it, or another episode we have, share it with somebody. Text it to them, a colleague, a coworker. Do us a solid. All right, let's get into it. We are talking with Dr. Lloyd Minor about the intersection of artificial intelligence and healthcare. Enjoy. Enjoy.

As the scientist, the surgeon, the leader in academics and the dean of the Stanford School of Medicine, tell me what a kind of day in the life is for you. Every day is a little bit different, but there are basically three components to my job. One is the traditional life.

components, if you will, related to being a dean of an educational and research enterprise. So we have 31 departments. We have a number of institutes and centers all in, roughly 2,800 faculty. And therefore,

performing research, they're educating amazing students, and they're delivering patient care. So making sure that we have outstanding leaders in the departments, the institutes, and the other entities that are a part of the School of Medicine and Stanford Medicine is

one very important part of my role. A second part is related to the strategy for the healthcare systems that we have, both our adult and children's healthcare systems. So how do we plan the future of those enterprises? Where are we going to deliver our services and where should we build areas of distinctive excellence, for example, to serve not only the people in our region, but more broadly, the

the country and the world. And the third component has to do with external affairs, and that is how I can help to represent what we do in Stanford Medicine and Stanford University outside of the boundaries of our enterprise, and also how I can contribute to raising the funds philanthropically and through foundations and other organizations that enable our people to do their very best work.

On a day-to-day basis, every day is a little bit different, but no day is ever boring. It sounds like it. And one of the things, as you were talking, I realized, you know, I don't fully grasp the difference between, say, what it's like to receive care at a medical institution that's tied to a university as opposed to just a hospital system, per se. What is the differences and the specifics of...

what you all do as compared to like if I were just to go to my local hospital or something like that. - Well, Chris, a lot of our focus is on complex diseases. So severe acute diseases or manifestations of chronic diseases. And one of the things that we do that academic medical centers in general do is to drive the future of healthcare delivery. And so we're constantly looking at how we can bring the very best scientific advances to the benefit of our patients.

We're also looking at how we can learn from every patient encounter, from every aspect of care that we deliver, how we can learn and improve and then disseminate that information more broadly. A role of academic medical centers really is to help define the future of healthcare. A lot of that is focused on severe acute diseases, but increasingly through our initiatives that we call Precision Health,

We want to look at applying technologies, applying genomic medicine, applying data science to improving the preventative care that we deliver and that others deliver as well. So that ideally, just as we have individualized treatments for cancer and other severe diseases,

We'd like to see the day when we can have an individualized approach for each of us to keep ourselves healthy. We already know things we should be doing, exercising, you know, making sure that we have a healthy lifestyle, but how can we make that more specific? And in particular, how can we assess for each of us as an individual, our propensity for developing a disease and then how we can either prevent that disease from developing or diagnose it much earlier.

Those are all things that academic medical centers are engaged in. Certainly on a day-to-day basis, we want to make sure that every aspect of the care we deliver is outstanding and is patient-centered and really recognizes the responsibility we have to those who entrust their care with us.

Now an additional advantage we have at Stanford is that our academic medical center is very much a part of the university. Our two main hospitals are on the campus of the university. Most of our faculty are here on the campus.

I'm sitting in my office today, a five minute walk from where I'm sitting. I can be in the Dean of Engineering's office or I can be in the biology department or the chemistry department. We're all a part of one ecosystem at Stanford. And those collaborations are really helping to drive

science and the technology that's going to define the future of biomedical research and the future of health care. Yeah. I imagine a lot of the innovation comes from places like yours because you have the ability

ability to innovate. You are incentivized to innovate, right? You're funded to innovate as opposed to a lot of the other places which are incentivized simply to treat. So that seems like that's a large part of the difference. Is that fair? That is fair, Chris. I think it's a very astute observation. And innovation runs in the blood at Stanford, of course, across all disciplines.

certainly in biomedicine, but also really across every discipline from the humanities to the social sciences, to the natural sciences, to engineering. People are thinking about, well, how do we

think differently when it makes sense to do so? And how do we look at problems for a new vantage point? And how do we contribute new knowledge through innovation? Let's call it the U.S. medical system often receives a lot of criticism. And I was just talking to somebody recently that said, but also realized that we lead the world in innovation, medical innovation, pharmaceutical innovation.

How do we maintain the level of innovation but start to understand that the costs are becoming untenable? And it doesn't appear that we are receiving the benefits necessarily when we look at kind of health span. It's a great question, and it's exactly the right question to be asking.

We have in the United States a great sick care system for those who have access to that system.

In my view, there's no better place to receive treatment for severe acute diseases than in the U.S. healthcare delivery system for those who have access to that level of care. But we don't have a very good healthcare system. We don't have a system that's really oriented around keeping people healthy. We're starting to make progress in that regard, and there's a lot of discussion about how we can

can and should be doing more. But by and large, we do a lot of after-the-fact care rather than care focused on, well, how can we prevent a disease from developing in the first place? Or if we can't prevent it, diagnose it much earlier and therefore treat it more effectively. And also, how can we more effectively engage each of us in

taking ownership of our health and maintaining a healthy lifestyle, doing the things that we know we have good evidence will help us to have a better, higher quality lifestyle and life for a much longer period of time. There are lots of pieces to that, but one of the things I think that's going to make a difference is how can we apply the same innovation

that has led to advances in the care of the most complex cancers, that has led to incredible advances in the treatment of heart disease at all different stages. How can we take that approach to innovation and apply it at the front end

And that really, Chris, maybe is a transition for us at this point or whenever in our conversation to talk about, well, what can AI do to address some of the challenges you just mentioned about the fact that we spend a lot on healthcare in the United States, far more than other OECD countries. And although there's been some curtailment

or there has been in the past some curtailment in the rate of rise of those costs, we still see a lot of expense. We also

drive innovation, as you mentioned, in the development of new therapeutics. And new therapeutics have been so incredibly impactful in diseases such as, well, let's take one example that's front of mind, I think, for many people today. Former President Jimmy Carter just recently passed away. It's important to remember that former President Carter was

as has been reported in the press, was diagnosed with widely metastatic melanoma, metastases to the brain, in I believe 2015. And thanks to a newly developed immunomodulatory drug, newly developed at that time, he was able to have roughly an additional decade, much of that high quality of life without

you know, without the burden of widely metastatic melanoma. That was unimaginable prior to this class of drugs that we call checkpoint inhibitors. So we have seen advances, but they've come at large expense and we haven't necessarily been as focused in the past on keeping people healthy as we have on treating severe acute diseases.

Ryan Reynolds here from Mint Mobile with a message for everyone paying big wireless way too much. Please, for the love of everything good in this world, stop. With Mint, you can get premium wireless for just $15 a month. Of course, if you enjoy overpaying, no judgments, but that's weird. Okay, one judgment. Anyway, give it a try at mintmobile.com slash switch. Upfront payment of $45 for three-month plan, equivalent to $15 per month required. Intro rate first three months only, then full price plan options available. Taxes and fees extra. See full terms at mintmobile.com.

This episode is brought to you by Indeed. When your computer breaks, you don't wait for it to magically start working again. You fix the problem. So why wait to hire the people your company desperately needs? Use Indeed's sponsored jobs to hire top talent fast. And even better, you only pay for results. There's no need to wait. Speed up your hiring with a $75 sponsored job credit at indeed.com slash podcast. Terms and conditions apply.

You just gave me a vantage point that I don't think it's talked about enough, which is, you know, there's oftentimes this criticism, exactly what you said about healthcare versus sick care. And why don't we focus more on healthcare? And what it made me realize was,

Let's take somebody like Stanford. They're doing it right. You all are doing it, but you also have to care for the sick and there's only so much that can go around. Right? So we're saying, look, we are trying to innovate on all fronts, but the person who

like Jimmy Carter has this terminal diagnosis. They now are past the healthcare component. What, what are we innovating for the sick care? So when we hear this stuff about, you know, institutions or companies publicly traded, always focused on sick care because it's maybe it's profitable or whatever. There's also this element of it's keeping up with the modern challenges of disease and life.

It's just never as simple as it seems, I guess, is what it's making me realize. You got to do both. And there's only so many of the yous.

Every study that I'm aware of indicates that as our population ages, and it is in the United States and worldwide, the incidence and the prevalence of cancer across almost all types of cancer is going to grow. And the need to provide care to patients with cancer, the need to diagnose it earlier, to treat it more effectively,

to take care of patients for a long period of time who are either their cancer is, in quotation marks, cured, or it's in remission, or it's controlled, that need is going to grow. It's not going to diminish. And so we do have to keep up with those, if you will, sick care needs, in addition to applying the same sort of scientific approaches that has enabled advances like the

the checkpoint inhibitor that former President Carter was treated with, or the many other new cancer therapeutics that we have that are showing real promise, not only promise, but results. We have to keep up with that and other diseases, neurodegenerative diseases, for example, that will rise in incidence and prevalence as well as the population ages.

We can't turn our backs on those needs, but we have to in parallel look at lifestyle, look at disease prevention, early detection, which in the past has not received as much attention as the severe acute disease area has.

What is your stance on this common refrain nowadays that the research community or the scientific community, healthcare community, whatever you want to call it, tends to focus too narrowly instead of the body as a whole? There's this big push. I mean, I'm sure you're well aware and I'm a fan of it, of like holistic focus.

care. And then they say, well, that's because mainstream or whatever only focuses on one part, not the body as a whole. Do you believe that to be true? What is the trajectory on that? How is it viewed in a place such as Stanford? I think it's viewed with a lot of credibility, openness, and a desire to learn more and have greater impact. That is,

How do we really think about, as you described it, holistic care? How do we think about nutrition? How do we think about the role of exercise? How do we think about the role of our mental, psychological health in terms of our overall well-being and in terms of the prevention of disease? We know that each of those components that I just mentioned has an important role in

in preventing disease or in improving the outcomes once a person does develop a disease. Embracing that as

as a part of the care that we deliver in our healthcare system, that is including an emphasis on psychological wellbeing, including an emphasis on nutrition, embracing that is extraordinarily important as well as studying it. So how can we be more effective at, for example, we can talk a lot about nutrition, its importance, we understand it, we teach it, but then how do we really change

or provide the opportunities for people to change their nutritional profile so that they are

promoting their health. That's been a real challenge in our country. We've known for ages that obesity is linked to carbohydrate ingestion. And we've seen a shift in the U.S. diet over the past 20 plus years towards a higher proportion of carbohydrates, less protein. So how do we

study, you know, how do we study the interventions that might lead to a change in our diets or how do we make healthy diets more available or educate people about how to achieve a healthier diet. That becomes then also part and parcel of our mission as an academic medical center, as well as driving the latest developments in cancer therapeutics and other therapeutics for severe acute diseases. Okay.

How would you recommend the average person, right? Who's trying to decipher, like, how do they know what to do, what to trust, where to go with more information? It's actually gotten harder. It's actually, even though we should have better information due to these advances, right?

We also have more. How do we know what is best given the information we have at this time? It's a great point. First is a realization that we're constantly learning.

And I think for those of us putting out information to the public, whether that's an academic institution like Stanford or it's a federal agency, having a large component of humility that what we're saying today should represent our best knowledge as of today, but also a realization that that knowledge and that understanding today

may and very likely will evolve over time and how we communicate the level of "certainty" about a fact or about a particular approach and how for some things this represents our best understanding at this time

but it's very it very well may change in the future i think communicating that uncertainty and a desire to continue to advance knowledge and advance the dissemination of knowledge is a very important responsibility of those of us who are involved in both generating the research generating the knowledge and then communicating it more broadly now in terms of specific sites and sources

you know there are a number of national organizations the american cancer society the american heart association federal agencies you know each of those have websites federal agencies the fda the nih have websites i found them in general to be carefully curated and to have information that at the time it's it's put out represents best knowledge at the time but again may change as we learn more and and

And that engagement, that sort of communication of this is the best we know today, but it may change in the future. I think those of us involved in generating that knowledge and communicating it need to do a much better job of indicating to the public

that this is the best we know right now, but we may have a different understanding in the future. One thing this podcast has taught me is the people who are really doing the work, like dedicating their life to it, they always couch every answer in, this is what we know. There are no certainties when we're talking about the complexities of, say, the human body or food and nutrition or the newest innovations. So

That is also some discernment on the part of, you know, the consumer or the average person, if you will. I also think there's

some distrust in the larger institutions these days. I don't know how much it's warranted, but again, it is clickbaity. It is loud. It is out there. And so I think it's people just like to jump on that, but it does get confusing. It does. And it's really important for those of us at institutions and those of us

who are dedicating our lives to making sure that institutions are serving the public in a responsible way, it's important that we're clear in our communications that we're transparent at how we generate the information that we're putting out.

And then when we discover we've made a mistake or we discover something that we thought before turns out not to be the case, then we correct it as quickly as we can and as transparently as we can. Sure. Yeah. Well, Lloyd, as we mentioned at the beginning, want to get into AI and its impact on healthcare and what's coming. And I got to tell you, one thing I have said for maybe the last year since AI has really, really come on the scene is I can't say that

technology or technological innovation has on the whole benefited humanity outside of health and quality of perhaps that, that life and health span. You know, when you think about food availability, right? Less people being food insecure or, um, not necessarily lifespan, but perhaps the ability to, to deal with some really difficult illnesses. Right.

So I will say that if any area I'm excited about what AI can do, it is healthcare. What are you most excited about the potential for this technology and how it might actually help humans live? I think there are three areas that, three broad areas that I'm very excited about in terms of the application, the deployment of AI in biomedical research, all

all the way up through the delivery of healthcare. First is improving the efficiency and the effectiveness of the care we deliver. A lot, you alluded to it before, Chris, in one of your questions, a lot of the care we deliver, healthcare we deliver in the United States isn't necessarily objectively effective if

if we measure effectiveness as we should be based upon quality of life and length of time where a person is maintaining quality of life. So how can we have more effective care and how can we have more efficient care? Again, if we're not delivering the right care to the right person at the right time, it's not going to be effective and it's not an efficient use of healthcare resources. I think AI

offers the prospect of improving both the efficiency and the effectiveness of healthcare. Second is access. Let me give you an example of how I think AI can and will improve access to high quality healthcare. If a person in a rural area has a rare tumor, it may not be all that rare, it may just be rare in that area of the country where they're being diagnosed.

The doctors in that area are not necessarily going to have today access to, well, what is the current knowledge regarding the treatment of that tumor?

AI can assist that both in making the diagnosis through a review of the slides of the tumor or a review of the radiology scans or a tumor board that brings together experts enabled by AI to decide what

the evidence suggests is the very best treatment for that patient at that moment in time. So access to high quality care also through the deployment of telehealth and other related technologies, I think is another area where AI can have profound impact. And then the third also related to access is health equity and how AI can break down some of the traditional barriers to access and make sure that everyone is receiving the

the type of care that they should be receiving and that that care is delivered in an equitable way so those are broad categories related mainly to the delivery of health care there also is a whole field a whole approach to how ai is going to transform biomedical discovery science and the

and getting from discoveries to new therapeutics, new diagnostics or improvement in the healthcare we deliver. And in that area, we've seen it, for example, with predicting the structures of proteins through alpha fold, for example. Now those approaches are being extended to nucleic acids. And right around the corner, I predict, I don't know whether right around the corner is three years, five years, or perhaps longer, but I suspect it's

more in the three to five year time frame, right around the corner, I believe we're going to see a much more efficient and effective

series of processes for the discovery of new drugs, new therapeutics, and the ability to design therapeutics specifically for an individual patient and their disease at that moment in time. That's all within our grasp. As with a lot of these advances, Chris, you know, in life, we tend to overestimate what we can accomplish in a short period of time and underestimate what we can accomplish in a long period of time.

So has AI been overhyped in areas related to healthcare, biomedical discovery? In some instances, perhaps. But I think over the intermediate to long term, those implications and those applications are going to be profound. This episode is brought to you by Shopify. Shopify.

Forget the frustration of picking commerce platforms when you switch your business to Shopify, the global commerce platform that supercharges your selling wherever you sell. With Shopify, you'll harness the same intuitive features, trusted apps, and powerful analytics used by the world's leading brands. Sign up today for your $1 per month trial period at shopify.com slash tech, all lowercase. That's shopify.com slash tech.

What do you see is currently benefiting people that is new AI technology? Like what's an example of something we can point to and say, see, this helped, this worked, this is new, this is cool. That's actually being implemented today. There's several, but they're all, each of them is in an early stage. And I'm going to mention the examples. And then we can also talk about

some of the problems with each of these examples, because there's nothing that I'm aware of today related to AI and its deployment in healthcare delivery or biomedical discovery that doesn't come with some associated challenges today. But let me talk about a few.

One is the use of ambient AI, so voice recognition transcription AI, in preparing a clinic note from an encounter that a patient is having with a healthcare provider.

It's the bane of every physician, every healthcare provider's existence, and certainly not a good thing for patients that when today all too commonly when you walk in the room to see a physician, the first thing they start to do is to type into a computer terminal the documented note describing the encounter.

They're not looking at you in your, looking you in the eye when they're having a conversation with you. You can even question how much they're focused on what you're saying, on what you're relaying, and how much they're focused on making sure that they get it documented because documentation is important and because they're going to have 10, 15, 20 other patients they have to see during that day and they want to make sure they don't miss something. Now,

There are technologies in existence today. We're deploying some of them today in our delivery system where

a conversation between the patient is transcribed and then organized through ambient AI. And at the end of the encounter, the physician and the patient get to review the note, correct it. And oftentimes it does need to be corrected because this technology is not perfect by any means. But during the encounter,

The physician is focused on the patient. The physician is making eye contact. There is that connection that is so incredibly important in the delivery of healthcare that unfortunately and inadvertently technology separated that connection when we moved to entirely electronic technology.

health records and when the need to have documentation grew. So that's one example being deployed today. Other examples using AI to help in the interpretation of radiology images that's being used today and increasingly also being used in the interpretation of pathology slides when

when there is a slides made of a tumor for example and knowing is this cancer or it isn't or

has the surgical resection actually gotten the complete margin? Those can be tricky determinations even by skilled pathologists. And that's an area that's ripe for transformation with AI because the AI has the benefit of learning from far more pathology slides than any one human being is going to see during their professional lifetime. So that's another example. A third example would be

a scenario that happens every day um three o'clock in the morning a patient comes into the emergency department um and they have a number of chronic medical conditions and they're coming in with a complaint either they have a fever or they have chest pain and the emergency

The emergency department physician has probably never seen that patient before. It's three o'clock in the morning. They could try to reach the primary care doctor, but they have to make some decisions pretty quickly. And there's this huge electronic medical record that either they're going to have to spend a half hour, 45 minutes longer reviewing that medical record to try to distill from it

the elements that are most related to what brought the patient to the emergency department. Now with AI assistance, we're able to distill from that electronic record the key points related to the complaint that led to the patient coming to the emergency department. That is being used today and is a tremendous aid, again, not substituting for

actually reviewing the medical record, but bringing things to the fore that the physician and other healthcare providers need to be aware of immediately as they decide how best to treat the patient. So those are some examples, again, early stage, but how I think AI today is having impact. We know that these technologies will improve. The ambient AI for transcribing clinic notes learns from every

from every correction that's made from something it hasn't done well. And that knowledge then improves the performance of the algorithm. These are early stages, but I think already very encouraging. Yeah. You know what I'm noticing is a lot of these things, especially at this point, are

unknown to the average person, but probably very well received within the medical community. When I think about the note taking component and I look, most people being treated by health care professionals, their biggest complaint is they didn't see me as a person. They didn't try to understand the depths of what's going on.

And it's not that the patient doesn't understand the challenge. It's just, they're in a vulnerable state. And so now with the assistance of AI, so number one, I don't have to capture notes. Number two, I can more readily understand your medical history and then focus on today to make the best determinant. That's a really, you know, I think that's a positive outlook on, um,

man and machine coming together, if you will. Yes, I do too. And again, we have to be very much aware of the challenges. We could talk about those more specifically, but first is to recognize that

Right now, this technology is not a panacea. It's still in its early stages of development and deployment. And we have to go in with eyes wide open about its limitations, about some of the challenges associated with its deployment. I know you play a big role in the co-leader of Raise Health, so responsible AI for safe and equitable health.

And judging by your background, it's not like you needed more work to do. So I'm curious, like, what is it about that specific idea, right? AI using it responsible in medicine. What is it about that?

that grabbed you and said, this is an area I want to focus on? Chris, we started the Raise Health Initiative during the summer of 2023. And this was in the early days of generative AI and its deployment. And it was clear for those of us that were

using chat GPT at the time. Others have other models. Other systems have come online, of course, since. But using it just casually to see what type of information it provided us, it was clear that this was an inflection point in technology. It wasn't an incremental advance. It was a real inflection point. And it was going to have implications for healthcare, for biomedical research, and for everything we do.

And related to something we talked about before, Stanford for decades has been exceptionally strong in information sciences, information technology, computer science. And a lot of the computer science faculty at Stanford are looking at healthcare biomedical topics because that's where the interesting challenges and problems are. And great faculty are drawn to problems and challenges.

And so during the summer of 2023, Professor Fei-Fei Li, who is one of the leaders in generative AI technologies, and I decided this was the time to start an initiative. We, Stanford, should be a leader in developing and deploying AI, but we should be doing it in a responsible way. And so, yeah.

responsible AI for safe and equitable health came out of those conversations. It's not a building. It's not an institute. It's a convener. It's a convener of people and ideas. And also one of the functions of race health is to disseminate information about how AI is being deployed and how it may, you know, how it's responsible deployment and development can be impartialized

influenced in the future. So those are the goals of Race Health. We had our first conference in May of 24. It's actually a series of conferences during a sort of AI and healthcare week that we hosted here at Stanford. And I think the

the thought convening component of race health is going really strongly, as well as we've had some seed grant programs to enable faculty to do some really creative work. And we're looking, constantly looking at how we learn about

the deployment of AI both within our system and then in other systems to make sure that it's being responsibly used and constantly improved. Yeah. And with the experts you have, I know Dr. Fei-Fei Li, is that right? Yes. Okay. I know world-renowned computer scientists, one of the first really in generative AI. What do you all recognize as the biggest challenges to

AI, AI growth in this space? Like, why can't we just go, all right, we have this technology. It's fantastic. It has all of the world's knowledge. Let's use it. I think there are three broad categories of challenges, Chris, that we have to approach as

directly and with our eyes wide open about how these challenges have to be recognized and then overcome. First is privacy, and we'll talk more about that. The second is that these systems can

can and do make errors. And we have to recognize that and always be a bit skeptical about what we're seeing and learning from generative AI approaches. And then the third is that

AI, generative AI is no better than the data on which it's trained. And a lot of that data has bias in it. And therefore, the results we're going to get back from generative AI are going to be biased. So really those three categories, first, the category of privacy, we have to be aware that particularly as information is brought in from multiple different sources, we have to be

be very aware that privacy can be compromised even in a completely quote innocent and unintended way and let me give you an example um if there's an ai if there's a generative ai approach that is bringing in social media information as well as as well as information about demographics

you could see where a simple query that a healthcare provider might enter, say that there's a question about a 35-year-old man who just traveled to Costa Rica, who has high fevers and also has a history of kidney failure and diabetes. That question,

query doesn't have any protected information in it per se. Okay. It's not a name. There's not a birth date. There's nothing that we would consider to be

as we call in healthcare, HIPAA-protected information. Yet that query, if it's linked to social media and that person happens to be on social media about their recent trip to the location, all of a sudden you've identified the patient from the query, even though you weren't intending to do so.

So protecting these systems, walling them off so that they don't, you know, so they don't get outside of a very protected environment.

is really important to protect privacy, as well as the concerns that we've had even before generative AI related to genetic information and related to how information about our health gets out beyond ourselves and our healthcare providers

related now to the ability of generative AI to bring in far more understanding about bits and pieces of our health and their implications for the future than were present in the past. So privacy is a big concern and protecting it is of absolute paramount importance. Second concern with regard to errors, one feature of generative AI that certainly generative AI experts can explain better than I, but

For traditional algorithms, we know exactly how the outcome of the query we give to the algorithm is being derived. Let me give you an example. We now use algorithms to calculate the dosages of medicines. So a physician will enter in the patient's age, their height, their weight, their kidney function. And then there is an algorithm that will calculate

based upon a lot of evidence, exactly what the dosage of that drug needs to be for that patient. You feed that information into the algorithm a hundred times, you're going to get the same answer every time. If you don't, then there's a problem with the algorithm that needs to be fixed. Generative AI, because of the way it's designed and the way it's worked,

you can ask the same query or maybe just a slight variant on the query and you may not get the same response back every time you ask it. So,

And the exact mechanism for how the mapping is done to bring in the response to your query is so complex that someone designing an algorithm is not going to be able to take it apart as they could in the medication dosing algorithm I just described to you.

Well, we have to take every result we get from a query in generative AI related to a scientific point, related to an aspect of healthcare delivery. We have to take that result with eyes wide open that it may have a hallucination, it may not be wholly accurate, and we have to be really, have a high degree of skepticism about how it impacts the treatment decisions we made.

Now, that's not necessarily that's not to say we shouldn't be using it or we shouldn't be learning from it. But we do have to maintain you have to go in with our eyes wide open and maintain a degree of skepticism. And then the third area of bias, we need to do a lot better job and that improvements have been made, making sure that clinical trials are broadly representative of the demographics in our country or more broadly demographics in the world.

Because when we're training a generative AI algorithm with data that just contains a narrow slice of the population, the recommendations coming out from that data are really going to be no better than the data used to train the algorithm.

Awareness of that, and the solution comes partly from the design of the algorithms, but also more fundamentally from how clinical data is generated that trains the algorithms. The solution is to be very aware of these biases and to constantly be revising the data sets or at least knowing specifically where the bias is and

we have for example in our emergency department at Stanford we have one of our faculty members who's really studied and has been able to modify algorithms based upon someone coming in with a complaint of chest pain and how

that the race, ethnicity of that patient may have a significant effect on whether that chest pain, as it's described by the patient, is indicative of a cardiac event or of something less severe. So if we're aware that biases can and do exist in the way

the algorithms deliver recommendations and steps, then we're oftentimes able to overcome those biases and design methods that actually are quite accurate and appropriately make recommendations based upon the information that the algorithm's receiving. So those are three broad categories of concern that we need to really keep at the front of our minds as we move forward.

As you think about Stanford being a leader in this and the coalition that you have created, aside from awareness, how does the problem get fixed so that the patient will eventually trust the technology? I think number one is communication. I'm so glad that we could have this conversation today. I'm excited about the potential for AI, but all

also want to try to have eyes wide open about the complexities and the challenges. So that communication is job number one, and a lot of that responsibility is on our shoulders as healthcare professionals, healthcare providers, scientists, and engagement of the public in what we know the opportunities are for the improvement of health and healthcare delivery, and also what we know are the challenges.

Rarely works in any technology to say, well, we're just not going to use this or we're going to shut it down completely. What works much better is to understand the limitations and to proactively correct them and make sure that we're responsibly deploying the technology. Okay.

I love it. Well, Lloyd, thank you so much for being on. It's very helpful and eye-opening. I'm excited about the future. For those listening, if they're interested in learning more or Stanford's recommendations or what's happening, do you have anywhere where we can go? Is there a website or anything that we can kind of learn more about?

Sure, our Stanford Medicine website has links to a lot of the initiatives we've talked about today. The RAISE Health Initiative, some of the other initiatives in deployment of AI and the interpretation of radiology images and other medical information. So sleuthing around our website or searching for various things that Stanford is engaged with or other institutions,

is a good way to start. I love it. Well, Lloyd, thank you so much for being on the show. Thank you, Chris. It's been wonderful to be with you. A thank you to this week's guest, Dr. Lloyd Miner. The episode was hosted, as always, by Chris Stemp and produced by yours truly, John Rojas.

And now for some quick housekeeping options. If you'd ever like to reach out to the show, you can email us at smartpeoplepodcast at gmail.com or message us on Twitter at smartpeoplepod. And of course, if you want to stay up to date with all things Smart People Podcast, head over to the website smartpeoplepodcast.com and sign up for the newsletter. All right, that's it for us this week. Make sure you stay tuned because we've got a lot of great interviews coming up and we'll see you all next episode.