The Machines Will See You Now: Anmol Madan's Human Roadmap to AI Health Care

00:00:31 Andrew Keen: Hello, everybody. Very interesting piece in The Wall Street Journal a few days ago about how, patients are using AI with or without the permission of their doctors. It comes from a an interesting new poll put together by KFF, which suggests that one in three adults are actually using AI for health information. In other words, the machines are already seeing us, which is the title of a new book. The exact title is The Machines Will See You Now, by my guest, Anmol Madan — A Human Roadmap to Autonomous Health Care. Anmol is a health care AI health care entrepreneur, living just up the road from me in San Francisco. Anmol, the book is out the beginning of September. Congratulations.

00:01:26 Anmol Madan: Thank you, Andrew. I'm excited.

00:01:28 Andrew Keen: Did you use by the way, Anmol, before we get into, AI and health care, did you use AI to write the book? My friend and I, Keith Teare, we have a weekly tech show. He lives in Palo Alto. Yeah. Seasoned entrepreneur, and he's boasting about using AI to write a book in about twenty minutes. Did you did you use Claude or OpenAI for this new book?

00:01:52 Anmol Madan: I did not, Andrew. And some of that is because the manuscript was written a couple of years ago now and went through a series of revisions and clinical reviews and peer reviews and things as part of, the publishing process. So I don't know if I would use it more aggressively today. I think when I was writing this a couple of years ago, it was just, you know, the AI tools are just starting to get there, and they were, you know, I think it's a great it was certainly a great thinking, and sort of collaborative thinking process, but the writing quality just wasn't there. And I also think if you're gonna write a book, there's something about the authenticity about kind of putting your own ideas and pen on paper, which, I think is important if you're able to write something. So I did not, but who knows? Maybe the next step I do this. I might—

00:02:38 Andrew Keen: Maybe, yeah, you can write a follow-up. Do you think there's the same sniffiness within the medical community about AI as there is within the publishing business. Publishers are very wary of books authored by AI. In fact, some publishers don't want their authors to even use AI for grammatical or second readings, critical readings. There are, of course, a lot of doctors who are wary of AI, seeing it perhaps as looking over their shoulder, doing away with their jobs. Is there, a similar kind of nervousness, antipathy within the medical community towards AI?

00:03:17 Anmol Madan: Yeah. And and for context, I mean, I'm a PhD, not a medical doctor, so we'd have to just, you know, use that asterisk as I caveat as I comment. But I do think the dialogue is broken, Andrew. And I see it broken really in two ways. I think on one side, you've got just incredible exuberance and excitement. And, you know, this is the AI community and computer scientists like myself. We're out there and saying, like, oh, this is all works great, and we're gonna automate health care, and that's one end. And on the other end, rightly so, a lot of the medical community has really good, thoughtful questions about safety, validity, testing, rigor. And I would you know, as in my personal view on this, again, I can't speak for the entire community here, but my personal view here is just because we're using AI or we have a new kind of technology doesn't mean we draw on [ed.: likely "throw out"] how we think of medical devices or health care specific software applications and the guardrails and the safeguards that need to exist. So I think the right answer is actually somewhere in between. And I think part of my reason for writing this book is I felt I was getting frustrated with the broken dialogue. It's it's like a one side you had people who were just really excited about it to no degree and weren't thinking through what the important steps were, what the nuances were. And then on the other side, you had people who were just really anxious and sort of not care where this would go, but also had to come to terms with what you just shared a minute ago, which is a meaningful number of consumers that are out there looking for this information. And if they're not looking for it on ChatGPT, they're probably looking for it on TikTok. And so consumer appetite and behaviors are there, and, we kinda have to go in and meet them where we are. So I think that's what we have.

00:05:03 Andrew Keen: And from your point of view, as a AI entrepreneur, you're the CEO of, RadiantGraph, a, an AI platform for large health care.

00:05:16 Anmol Madan: Yeah.

00:05:16 Andrew Keen: Do you celebrate, or are you nervous about this KFF report that suggests that one in three adults are already using AI for health information with or without their doctors? Presumably, they're going to the machines before they go to the doctors or in place of their doctors. Does that suggest good news? If not from the point of view of Rad— of RadiantGraph, as a start up, but more in terms of how this new technology, which is a new way of thinking and doing, is changing the medical industry, the health care industry.

00:05:54 Anmol Madan: Yeah. It's it's a little bit of both, and I give you both sides of the argument. So I think it's exciting that people are finding value, and it's exciting and, like, let's start with the bigger question. Right? I mean, health care in our society is fundamentally broken. And I talk about this in our in the book where we spend where, you know, we have one of the highest, spends in sort of in health care expenditure as a country. We don't have, you know, great outcome.

00:06:21 Andrew Keen: And we, of course, is the United States of America.

00:06:24 Anmol Madan: United States. Yes. And, we don't have the best outcomes as a country compared to other 12 countries for what we spend. We have, you know, massive access shortages and everything from mental health to, you know, we talk about things like Medicaid and Medicare. So I think overall, health care, despite the amount of money we spend on it and whether, you know, we spend over as employers, we also spend money on it as taxpayers. Right? And even we're talking about San Francisco and we're talking with the Bay Area and, you know, the amount that we spend on programs in supporting our local populations as well. And so I think, fundamentally, we're not you know, when I look at the last couple of decades of innovation in health care, my kind of frustration, and I've been part of some amazing companies, is, you know, we haven't really seen the impact of that on the economics of health care, perhaps even the quality and access of health care. And, you know, we've you know, and this is a big risk with this area. So what is exciting is that people are finding value, that they're going off, they're looking for help. I think if we create the right experience and the right safeguards, it is a mechanism to make health care more accessible. Right? I'd like to live in a universe, in this bad [ed.: likely "a parallel"] universe, where health care doesn't cost us $4.7 trillion. It maybe cost us half of that. And twice we have twice as much access, and we know that the AI systems that are delivering that access are safe and tested and rigorous. That's the parallel universe I'd like to live in. But there's a lot of steps between where we are, the reality of where we are, and that parallel universe. Right? So so it's promising in that, okay, great people are finding value. They're looking at these things, but we've gotta create the right support mechanism for society to transition to that. Some of that is technology. Some of that is, helping the medical community and the health care community get comfortable with this change that's coming. Right? The the various elements of the workforce. But some of that is also how we structure health care, how we pay for health care, how we're incentivizing innovation in health care. So a lot of those things have to move and move together. So, again, it's a long winded answer to an easy question. I think it's exciting, but there's a lot of work to be done for the real vision to emerge here.

00:08:33 Andrew Keen: You've said earlier, Anmol, that American health care system is fundamentally broken. You're not the first or the last person to say that. You wrote an interesting piece recently on "Health Equity 2.0: Bridging Health Equity Gaps with AI Agents". Is the main reason why it's broken or the main consequence of it being broken that has a great division between access for wealthy people and everybody else? Is that the biggest reason it's broken, or is it broken because of the insurance system, because of the unhappiness of doctors, because of the state of hospitals? How would you define this brokenness?

00:09:17 Anmol Madan: Yeah. You know, you're asking me the trillion dollar question. Like, what's the causality? What's the underlying cause? And, you know, it's probably some combination of all of those. Right? I think are there economic structures of how we pay for health care and, you know, how we incentivize certain behaviors, that could evolve and would have a better impact? Yes. Right? And so as an industry, we've spoken at length about value based care. And so as a historical, you know, kind of background here, most of the care that we receive in the US is fee for service, which is a doctor sees the patient, gets paid for that. Right? And with value based care, there's an incentive to take on financial risk. Value based care is still a small percentage of the [unclear] spending — [unclear] care that's delivered. Right? So it's still an emerging pocket, and there are different areas where you see more of it. Some parts of Medicare are not in other pockets of it. So I think there's some structural issues. I think the thing that I, you know, sort of have the humility to say is I'm not with the book or as a computer scientist building these systems in the real world. I'm not gonna solve the broader, regulation, economic incentive problems. I mean, there's a lot of debate on both sides of the political aisle. You know, there's lots of kind of arguments for how health care payments would work and what the incentives are. So I think there are some structural issues that are, I'd say, outside the purview, and then sort of cheating and sort of sidestepping those issues. Right? But the thing that we do control, the thing that is in the hands of people building these systems and building these technologies, as well as, I would say, pioneering, medical experts who are participating in helping and shaping these technologies. The thing that we do absolutely control and we don't have any excuse for is thinking through the broader road map of what these technologies can do well, what are the ways in which you think about clinical safety and guardrails, And then how do we find the right relationship with regulators, you know, whether it's folks at the state medical board, it's the FDA, and other places, to create safe environments or boxes for these systems to start to help take on the access burden in our society. Right? And so I think there is this question of if people aren't getting access to care, and if we can have AI systems come in and start doing helping with, you know, screening information, helping them connect to first order support, If you can create opportunities for them to get coaching on things like behavioral health or managing their diabetes in the absence of human resources that are missing in today's health care economy, I mean, are those things that you know, how do we design for those? How do we make sure they're safe? How do we make sure we kinda bring those to market? I think that's the part that we do control as technologists. And there's enough market incentive, and there's enough need in society to solve these things for us to go build those systems. Right? There is a, like, you know, the equity argument that you pulled up there for a second, you know, go back to health equity. My my thesis on that is, you know, we've the health equity is sort of it used to be a really promising word, and let's say in the last couple of years, it's gonna take on a different, you know, less popular kind of viewpoint. Some of that depends on the administration that's in charge and their viewpoint. It's sort of appropriate to the system. But if you think about the fundamental level, right, the one of the big issues in health care is, you know, there's a lot of personal context in how health care works. Right? So let's say you and I, maybe we speak different languages. Maybe we have different cultural viewpoints. Maybe we live in different parts of San Francisco. Right? And, you know, with human providers, you know, even with the best intent and the best training, you're kinda always sort of dependent on the human to adapt and evolve and interact. With AI systems, you unlock kinda massive new opportunities for these systems to communicate with each one of us in a way that's personalized to us. Right? So that's what's really magical and powerful. Right? So I'll give you an example. Right? There's, a lot of literature about things like eating disorders. And, you know, eating disorders, people have one view of what that is, but there's a whole bunch of other subtypes. Right? Different, communities, right, have different forms of and so, like, the system's adapting and saying, here's I'm gonna speak to a person who's South Asian, kinda like me, and translate how I'm gonna help them overcome the lifestyle changes they need to make or the behavior changes they need to make. Right? That stuff you can do with AI because you can adapt the tone. You can adapt the empathy. You can adapt the language. You can you can bring in a lot of information about the person, which was very hard to do with human providers. So there's things that we can do here which actually make health care more personalized, more effective in some ways, and then define the right safeguard. I think that's part of the equation that I made out in the book.

00:14:04 Andrew Keen: Well, isn't that what a doctor should do? I mean, it's no secret that, for example, you're of Indian heritage from your name or your looks.

00:14:13 Anmol Madan: Yeah.

00:14:13 Andrew Keen: You go and see your doctor. They would pick up on that. They would have that information in your file. How are how are doctors dealing with this? How should they deal with this? We've done many shows on this subject, Anmol [ed.: transcribed as "Daniel"]. One of my most frequent guests is, Robert Pearl, who used to run, a major insurance network here, Kaiser. Teaches now at Stanford. I'm sure you've come across him. He wrote a book, a best selling book, about how miserable—

00:14:44 Anmol Madan: Yeah.

00:15:03 Andrew Keen: So—

00:15:04 Anmol Madan: Yeah. So and, Andrew, there's a lot of nuance here. Right? And so I will, you know, one of the things I talk about in the book is actually a road map for increasing autonomy in health care.

00:15:19 Andrew Keen: Yeah. And the subtitle of the book is A Human Roadmap to Autonomous Health Care, and maybe we'll define what autonomous health care means too.

00:15:28 Anmol Madan: And so and so I kinda talk about you know, and I use self driving cars. And, you know, self driving cars, were, you know, there was this issue of maybe a decade ago with self driving cars is which, you know, somebody had sort of radars, and somebody was using lidars. You know, all these manufacturers, and you didn't know what was what. Right? And then, you know, the industry came up with standards. It kind of laid out the different levels of automation, in that industry. And it went all the way from kind of cruise control and, you know, little help with the driver all the way to fully automated cars. Right? It kind of laid out where you were playing in different companies and chose a different path to play in there. Right? A lot of the traditional manufacturers focused on getting the supporting the driver, and then other companies like Waymo decided they were gonna go all the way to the fully automated version. Right? But it brought structure and it brought clarity to how these different levels are defined, how they ought to be tested, potentially how they ought to be regulated by local governments and cities. So I think that's one of the things I lay out. And I do kinda an equivalent version for the increasing levels of potential autonomy in health care. And so most of health care that I kinda call out in society today is kinda like a level zero. Right? It's like the base level where, you know, 99% of the decisions are human made. There's a nurse or a doctor that's making a treatment decision or diagnosis decision. They're prescribing a medication or perhaps a clinical treatment. And I sort of talk about how we could think of multiple levels of that to be supported by machines or perhaps even augmented, you know, operated by machines. And so the next immediate level, which we're sort of getting to, is machines reading, you know, kinda call this, like, doctor assistance. Right? So kind of level one, which is machines kinda coming in and reading X rays and reading [unclear] reports and reading imaging and arming you with the right information as a human provider, as a human doctor or nurse to kinda use that information and make better decisions, right, or make the actual decisions. And that's, you know, taking away some of the cognitive burden that exists in the work of a doctor. And I think we're seeing a lot of that being deployed in society today. And especially when you look at fields like radiology, they've come a long way, where, you know, I've had doctors describe this to me as, you know, I'd have to carry, like, backpacks of reports to read, and now the system's kinda, like, telling them where to focus. And so they're spending their time understanding and reviewing the data that's analyzed by machines versus not. So that's kinda like level one. That's kinda like the cruise control.

00:17:59 Andrew Keen: Yeah. I mean, you using the Waymo example, though, Anmol, that I know you use, you and I, if we were to visit one another, you're mile or two away in San Francisco, we'd catch a Waymo, automated car. The drivers have been taken out of the equation. So why would doctors be encouraged with that? I mean, I take your point that maybe doctors will do the face to face with patients, but I'm assuming that the algorithm also can be taught to do that in a more sophisticated way, in a way that won't be affected by moods. Why should doctors be happy about machines seeing all of this?

00:18:37 Anmol Madan: Yeah. And so that's, you know, that's where I start to ease into the level two and the level three where there's increasing levels of automation, right, where the machines are not taking on perhaps the lowest level of treatment diagnosis decisions, and they may still go to a human review. Right? So so I think from a technical perspective, there is a way for us to deal with this problem, and there's also a way for us to structure the right validation and sort of regulatory perspective to oversee that. To go back to your point, like, why do human doctors care? You know, why should a health care provider care? Right? Maybe even a nurse. Right? Beyond doctors, maybe this also impacts how nurses work. It impacts how other professionals work in a health care setting. You know, I kinda go back to, and I'm not a doctor, so it's sort of not my place to say this, but I'll sort of draw this comparison anyways, which is, you know, I think just because, you know, you think of, like, how we travel, you look at other industries that have gone through this transition. Right? I use crossing oceans, for example. Right? Just because, you know, you have a pilot who's in control where autopilot's doing seventy, eighty, 90% of the flying on a transatlantic flight doesn't mean, right, that person sort of no longer important. Right? So it's sort of like, you know, you don't have to be pulling the sails. You don't have to be pulling the rudder to cross the ocean. Right? You've seen this in other—

00:19:54 Andrew Keen: So, [unclear], so I take your point. And and, of course, you're right. We still have pilots sitting on, especially transatlantic, jets even if they don't spend all the time, driving, so to speak, the plane. What do you think humans can do that machines will never be able to do? I mean, you talk about this human road map to autonomous health care, which to some might sound like a contradiction. Of course, it's meant to sound like a contradiction, which it isn't necessarily. But what will the machines never be able to do when it comes to health care?

00:20:30 Anmol Madan: Yeah. So I think the—

00:20:32 Andrew Keen: Never, of course, is—

00:20:32 Anmol Madan: Yeah.

00:20:34 Andrew Keen: Is is a dramatic word. I mean, who knows? But let's say in the next twenty or thirty years.

00:20:40 Anmol Madan: Yeah. I mean, I it's look. I think the machines today are kinda like toddlers walking around, learning how to walk. Right? So they stumble. They're not very good. They make mistakes. Right? So I actually I think the risk that machines take over everything's everything that human providers do or health care professionals do, I think is actually quite low. And I talk about it as, like, that's kind of the end state. And by the way, we may never get to the end state. Right? However, the framing I would kinda encourage doctors to have or health care professionals to have, including folks who work with us at RadiantGraph, our chief medical officer, our medical directors, our clinical experts, our coaching staff, right, who are associated with the company, who are guiding and overseeing the training of these systems, is think of it as, like, where could we put these systems in a way that they can actually be trusted? And and so, really, I would think of these systems today, Andrew, as sort of little toddlers that are learning how to walk, and they're not really good and they stumble and they're gonna make some mistakes. You're gonna have to have layers of protection and safety to make sure that whatever you're trusting them. But having them do that frees up time and frees up resources and frees up capacity in a constrained health care system, which allows our health care providers who work incredibly hard to have better point of view of life [ed.: likely "a better quality of life"]. So I think for us, I think the mistake would be, you know, kind of falling in the trap of, like, these systems are not perfect, so we shouldn't try them at all. I think—

00:22:09 Andrew Keen: I mean, that but that wasn't my question. I'm not sure you're really answering my question. I said to you, what can humans or what will machines never be able to do that humans can do? And you told me that currently, these machines are like toddlers, which I don't think is particularly controversial.

00:22:25 Anmol Madan: Yeah.

00:22:26 Andrew Keen: They're gonna as using that metaphor, they're gonna grow up. They're gonna become children, teenagers, middle aged, old. What will they never be able to do?

00:22:37 Anmol Madan: I'll tell you where they seem to struggle. And so the issue with the never is that thirty year axis. As a computer scientist, you kinda learn never to predict thirty year predictions, but I'll do it anyways. So I'll tell you what seems easy for them to do and what seems hard for them to do that they may never learn to do as well as humans do. What seems easy for them is to solve with especially with large language models and different classes of machine learning models prior to that, is for them to solve kinda the puzzle of taking a set of inputs and guessing the likely diagnosis or treatment actions. Right? So the mathematical formulation, the kinda ability of these machines to navigate that is exceptionally good. Now there's a little bit of an asterisk on the exceptional good because I think we actually aren't testing— one of the things we talk about: are we testing these systems the right way? I think we're applying human benchmarks to how we measure the accuracy. And I think these machines are gonna fail differently than humans do, and so I think there's a conversation on how we're testing the right things. But I think regardless of that asterisk, I think, generally, you can say they're very good and they only get better because we are just starting to kinda scratch the surface how fast we can innovate. The one area that I do call out as a place where we're underestimating, you know, or we're sort of overestimating machines capabilities and underestimating how hard it is, is this idea of human perception. Right? So when you and I are communicating, even though we're on video, you know, my body language, the way I'm interacting— You know, if I walk in and you're, doing an in person visit with me in a clinic, the way you assess me, I think those things are actually really still quite hard for machines. There are fields of computer science, so things like machine vision, a variety of kind of sensing, you know, pervasive sensing techniques that are improving, but they're not improving at the same rate as sort of human providers are. So when you think of kinda, like, classic telemedicine use cases, outpatient use cases. Right? So I'm not talking about robotic surgery and some of those kind of examples, but I'm talking about the kinda general everyday care access to kind of basic everyday care, chronic condition care, primary care in our society, mental health care in our society. I think these machines are getting good at the just the problem solving part. They're still not that great at the, assessment of the situation. And they may never get the human capacity. I think that's one area that may never happen. Right? The other one that is still very early is it's not it's clear that humans are forming relationships with these, you know, LLMs, agents, companions. Right? And so, like, we're seeing that obviously outside of health care tremendously and, you know, people have these, you know especially when with the loneliness epidemic, we're gonna form these relationships. But I don't think there's a ton of literature that shows that those relationships make people better in things like mental health. Right? So I think that literature is still super early where you having a relationship with a chatbot that you like or trust or whatever. Is that actually clinically the same as the person working with a therapist? Right? A human therapist that might spend the same thirty minutes. So I think just because you're having an interaction, does that interaction translate into better outcomes in case of mental health? It's also kinda early. So I would say those are things where, you know, the there are places where the technology curve is going like this. And if you project forward twelve months, twenty four months, five years, you're like, yeah. We're gonna need it. Like, diagnostic, you know, treatment, kinda decision making, decision tree solving feels like one of those. And then there are places like the sensing element, the interaction element where the curve is a lot more kinda, you know, early. Right? And or the ability of people, to form relationships with, you know, these AI therapists or coaches as they emerge and actually see better outcomes. That's also pretty early. So those are things there. If I had to guess what couldn't happen in twenty years, we may discover that those things actually never been as good as you want them to be. And you always—

00:26:25 Andrew Keen: Yeah. And, we've done some shows recently on the failure of therapy or perhaps an overmedicalized society full of anxious people who believe that they're suffering some sort of mental illness when, in fact, they're not and spend too much time with therapists. So maybe a crisis of human therapists in the face of AI isn't such a bad thing. What should we be most fearful of? The Wall Street Journal wrote a piece recently on privacy, piece about uploading blood work to AI and oversharing. Should we be concerned that these AI platforms, like perhaps RadiantGraph, aren't very secure and that all our most intimate information is gonna get known by the government, by our employers, by our friends?

00:27:14 Anmol Madan: Yeah. So I think, you know, I'll sort of talk about this in the general sense. I'm also happy to talk about RadiantGraph specifically if you want. But I think, generally, folks have to be mindful and thoughtful about how we're thinking about data and not just, by the way, health care data. I think all data in this era of AI. Right? And I think you started with the, you know, the open comments on kinda, like, how the publishing community is thinking about AI and any AI generated content. I do think they're one of the things that kinda spooked people early on was all the data, all the books and the articles of the Internet kind of being processed by AI systems and that knowledge going inside AI systems. So I think the same thing is going to happen for health care data as well. Right? So and I think it's a it's kinda like a tricky problem because on one end, in a purely kind of benevolent universe, right, if you imagine again, like, a purely kinda optimistic, benevolent sort of scenario, you know, the more data these systems have, the better they're gonna get at learning things about us and also driving scientific discoveries. Right? We may discover completely new understandings of what it is to be a person living with a chronic condition or dealing with mental health. We may understand completely different phenotypes of, say, depression and anxiety if we were able to get all this data into some systems. Right? So that's a benevolent kind of universe view. The reality on the other side, though, is sort of often there is individual incentives, there's security issues, there's, like, you know, data that you kinda have to make sure is this a HIPAA compliant platform? Is this a PHI, PII compliant platform? Right? So I think if you're going to work, you know and I believe health care needs its kinda own AI companies and own AI platforms because you have to comply with all the things that matter when you're dealing with sensitive health care data, whether it's biometric data like you just shared or lab data, or individual PHI. And and so I think that's I think I don't know if the mass market at a lab is, like, pulling this in a ChatGPT or something like that is kinda like frame right now. Now that is changing because, you know, everybody from OpenAI to all a lot of the smaller peers have all created their version of the, you know, the health care specific one that maybe has a higher set of safeguards. But I think I would urge the average consumer personally, I mean, at least for me, you know, being a little bit cautious of where we're sending our data and how we think of it, because I think we're still kinda defining a lot of this in the industry. And I think if it's a health care specific platform, like the one that we were calling at RadiantGraph, right, it's important that you are secure. You're designed for PHI. You're very clear on how the data is used or required for health care operations. Right? So those are there's certain guidelines that have to apply, if you're gonna be using this data in any kind of API.

00:29:57 Andrew Keen: Why, because you mentioned OpenAI, Sam Altman is another of our San Francisco

00:30:05 Anmol Madan: Yeah.

00:30:06 Andrew Keen: Neighbors is not very trusted by most people, Slippery Sam. Same is true of, Anthropic and many of these other AI startups. People just don't trust them. These people are in the business of making money. Why should we trust entrepreneurs in AI, health care, Anmol, like yourself? Why why would we trust a company like RadiantGraph? And I don't wanna turn this into a conversation about RadiantGraph any more than any other platforms that are making entrepreneurs like you incredibly wealthy overnight and, in fact, driving up the price of real estate in San Francisco.

00:30:45 Anmol Madan: Yeah. Yeah. So, so we're a tiny start up. So just for projects [ed.: likely "perspective"], we're not OpenAI and— and Anthropic. So thank you for putting us in the same category, but we have some work to do before we get there. But I look. I think, you know, maybe the I do see where you're coming from. Right? I do see this kind of like, hey. You know, these people are starting these companies, and do they have the right incentives, and, you know, how are they benefiting from our data? And I think those are genuine concerns, and those are things that we have to think through. You know, I tend to have a bit of the other perspective, which is, you know, I kinda started a company, started a grad school. You know, it's so for me, it's like entrepreneurship or building these companies, whether it's in health care or in other domains. You know, part of what brings me to San Francisco, what I love about kind of being here, is, you know, to me, that's kind of part of creating the future. Right? That's the world that we wanna live in. And so, yes, you're right.

00:31:47 Andrew Keen: Although I'm not sure people wanna live. I mean, if San Francisco is indeed the future when you walk around this city, which on the one hand, has, you know, multimillion dollar mansions everywhere, on the other hand, armies of homeless people on the street, I'm not sure anyone really wants this as a future.

00:32:05 Anmol Madan: Sure. I think there is a massive amount of social inequity in San Francisco, and I would agree with that as a [unclear] of the city. And I think, by the way, we have a lot of work to do in public health in San Francisco as well. And so, you know, for anybody who's listening, I'm a adviser to, the Department of Public Health in New York City. I would love to do some work here and just kinda help out where I can. And I know there are lots of entrepreneurs like myself who won't [ed.: likely "want to"] do that. So I think, like, to agree with you on that. But I think what's interesting about the Bay Area is you look at if you look at the technologies that have changed our world in the last thirty years, right, starting from, like, the personal computer to the iPhone to what we're seeing what we saw with the web, right, in 2000, to what we're seeing today with AI, what we saw in, like, 2011, 2012. And so you look at all these companies. As a person who went to school in Boston, right, and sort of enjoyed his time in Boston. Right? You you look at San Francisco, and you're like, this is where—

00:33:00 Andrew Keen: Almost Boston, Anmol. You were at MIT, just over the river in Cambridge.

00:33:04 Anmol Madan: That's right. Yeah. And so—

00:33:06 Andrew Keen: You're gonna offend all the Cambridge people.

00:33:08 Anmol Madan: Sorry. Yeah. [unclear] Cambridge as I was there. Right? And and so yeah. So it's, like, that's what's special about this place. Right? And so you're right. There's there's bad sides to everything. There are things we should be doing better. We have social issues to work on for sure. You know, I think there's probably not a con we probably have a whole podcast episode on, like, San Francisco's political politics and how we are operating and things.

00:33:31 Andrew Keen: Well, we've done many of those. Yes. Of course. Right?

00:33:33 Anmol Madan: And, like, there's a lot of work for us to do and see. I don't— I do not think— and I say this [unclear], so I'll just say, I think we— while we have a great tech community and we have a great, you know, kinda, social [unclear], I think we would wanna work with you as a city, and I think all of us need to, like, help out and all of us need to do that. So with that said, but what's interesting about the Bay Area is these people have pioneered, you know, kind of different stages of our society. Right? Just this morning, I don't know if you caught it, as we're recording, but, you know, Jeff Dean just announced he's starting something new and Jeff Dean is—

00:34:04 Andrew Keen: Yeah. So Google, one of the first 20 people at Google now is—

00:34:07 Anmol Madan: Like, you know and for those of us who grew up in computer science, right, in the last decade, if you look at the innovations that have come out of that small group, the stuff that he himself has done, sure that Google has created in that era. And I think that's I mean, a lot of what society is today, a lot of what the world is today. Right? You travel to Asia. You travel to Europe. A lot of what the world takes for granted actually kinda happened here. And I think, again, I there's I'm not saying everything's fixed. I think there are a lot of work to do. I think we have a lot of stuff that's broken in San Francisco, in California, both in the US health care system. But the thing we do have is incredibly smart, talented people who are, you know, a lot of start ups today are 996 or whatever version we call, who are pushing with some vision because at least some part of them wants to live in a better society, like, wants to invent the thing that's gonna make life better for us. And that's why I go back to your part, like, what you know, I can't talk about the other companies or whatever. I think for me, what drives me and what I'd like to exist in this universe is a place where we have, you know, twice as much access in health care. Our providers have a much better quality of life. Our nurses and doctors are not doing the little mechanical task. We're kinda automating a lot of kind of, you know, simpler care use cases, situations with the right card [ed.: likely "guardrails"] or with the right supervisions. And health care in our society is preventative, right, versus waiting for [unclear] and cost half as much. And if we can take the next decade or two and take advantage of the shift and get there, that's totally worth it. And that's what we should be marching towards.

00:35:43 Andrew Keen: And one of the things that and I've had this conversation with, Robert Pearl as well on this is you don't need to be a nuclear scientist. You don't have to have an MIT degree, Anmol, or a Stanford degree to understand that, you should eat healthy foods, that you shouldn't go to fast food too much, you shouldn't drink too much alcohol, you should exercise. So a lot of the preventative stuff that Robert Pearl or you talk about is out there in the culture. Maybe people don't wanna hear about it. Maybe they're switching off when they hear it. But why would a machine that tells someone that they need to lose a bit of weight or they need to stop smoking or they need to reduce their alcohol or drug intake, Why is that gonna convince them any more than what a doctor will tell them or what they see on television, the radio, or on the Internet?

00:36:42 Anmol Madan: Yeah. That's a great point. So, and I think you hit the nail right on the head. Right? Is if we could just solve the behavior change problem in health care, that's a trillion dollars out of the 4.7. Right? There's different people who've kind of come up with that statistic. And so that's kind of the, you know, the way to think of it. It's massive, massive issues. So let's go let's take a specific example. Andrew, and I'll tell you where I think machines can work. Right? So, a person goes to the doctor's office. Maybe they show up for, you know, exam of, you know, meeting their doctor once a month, once every few months, once a quarter. Right? They get to that advice and then they kinda forget and then they go off. Right? Where I think these systems can play a role, and I've seen them play a role in my previous companies and sort of, you know, in real world settings, is they then pick up the part that's kinda in between the doctor's visits. Right? So in exam and they cater that and they personalize accurately. So let's take a person who's living with, say, type two diabetes. Right? And let's say they're managing their diabetes, but they're at risk of getting unmanaged. You know, maybe they track their blood glucose, maybe they don't, and they're not kinda you know, they go to the doctor's office, like, three days after the visit. They're, like, super engaged, and then they kinda, like, fall off. Right? All of us do that. Right? All of us are probably guilty of that. Right? And so what these systems can do is then they come they can learn what's it like where to support and engage this person. Right? What's it like where to engage this person and select to this person who's less likely to change their diet? Maybe they're more likely to walk more. Maybe there's a comorbidity with mental health. They're stressed out. They're stressy. Right? Maybe they're, you know, they're certain kinds of coaching and support that they need. So what so instead of kinda like the generic one size fits all, eat healthy, you know, walk more like, the system actually personalizes every element of your experience to make it easier for you to eliminate that friction, eliminate that value that's getting in the way. Right? And so I've seen this at scale in commercial scale. And in some of the previous companies, we built these systems to help tens of millions of people kinda understand their patterns and learn those. Right? And so when I think of machines starting to help with the actual care delivery, the way we experience health care, I think there's an opportunity for that kind of personalization to exist. Right? And so if you think of a coach, for example, helping you manage your diabetes, your experience could be completely different from mine because the stuff that you would respond to might be completely different from what I respond to. Right? And and just in that specific context in managing, a type of— we may have the same A1C. We may have the same, you know, age group and the same demographic variables. We are different individuals, and what will engage us in our care is fundamentally different. And so I think that's where these things can shine. And because they're machines, you know, we're not we don't have to carve out human time. We don't have to carve out a person who go think about those, process that information, come up with the call, be prepared. It kinda happens programmatically and can happen at massive, massive scale for essentially—

00:39:38 Andrew Keen: Yeah. And I take your point, and I hope you're right. Although I have a feeling that with AI slop being increasingly ubiquitous, messages from the machine on getting more exercise might be seen as complete slop and be eliminated. Anyway, you may be right. Maybe the machines will see you now in terms of medicine, health care. Maybe that's a good thing. We are on the brink, for better or worse, probably, of a human road map to autonomous health care. Certainly, the autonomous health care stuff's for real. Whether there's a human road map is another issue. Finally, Anmol, I often do this when I we have AI conversations, especially with entrepreneurs because bots are becoming more and more ubiquitous. The Machines Will See You Now suggests that doctors maybe will become scarcer and scarcer or less and less seen. How would you convince me and our audience that you aren't a bot? What's human about you?

00:40:44 Anmol Madan: Well, I'm quite flawed, and I'm often wrong. And so but part of being a technologist and part of being an entrepreneur, in some of the starting companies, you kinda have to take the big swings. You know, I feel like, you know, you can spend eight hours a day, ten hours a day at work, and you might as well make the biggest thing you can and sort of create the world in what it is. So, in terms of, you know, if I was a machine or an AI avatar of Anmol, you know, I probably wouldn't be as wrong as I sometimes am. So, you know, we all we all, like, here, we're kinda, like, learning machines, and we figure it out. We try new ideas. We see how they work, and we get better. So that's kind of how I think of—

00:41:24 Andrew Keen: And that's a priceless response, which is not surprising given that Anmol in, Sanskrit means priceless. Anmol Madan, a San Francisco based entrepreneur and the author of a new book, The Machines Will See You Now: A Human Roadmap to Autonomous Health Care. Congratulations, Anmol, on the book, and con— and best of luck, and congratulations as well with your, AI platform for large health care. RadiantGraph might not be quite OpenAI or Anthropic yet, but you still can grow. Thank you.

00:41:59 Anmol Madan: Thank you so much for having me.