NVIDIA: The hospital itself should be a robot | Kimberly Powell, VP of Healthcare transcript

Think about how the hospital itself should be a robot. The hospital itself should have eyes and ears sensing everything that's going on and using the modern world foundation models and spatial intelligence to operate. >> We have declining birth rates everywhere. We have populations getting older and and sicker and just more medically complex. We we are not adding net new clinicians quickly enough. So those are the realities that that we face.

AI is now hirable because it can do work. You can hire it. And so this isn't like a software system that you have to go buy a license for. This is it has to be thought of as somebody who's on payroll. >> Hi everybody. I'm Christina Far. I'm the founder of Second Opinion and we are so appreciative that you would tune in with us to this podcast. One small thing you can do that means the world to us is to just hit that subscribe button.

We'd also love to hear from you about what you liked about the show, what you didn't like about the show, what do you think of the guests, and who should we have on next. Hi everybody. Welcome to Lifers. I am here with Kimberly Powell from Nvidia who is very much um an ideal Lifers guest because she has been with Nvidia such a long time. How many years has it been now, Kimberly? >> Oh, you have to ask. 18 wonderful years. >> 18.

Wow. Okay. So, um, there's so much to talk to you about today, but I've got to hear the story, and I know you shared it before, but I love the story of how you first ended up at Nvidia. Um, and at the time, you had been working in the the more the medical device and imaging space. So, tell us about how you learned about the company and then how did you get the job? >> Yeah, thank you for having me, Christina. I love I love your podcast.

It's um it's an honor to be able to share this story. It's really fun. Um, I was working as you described for a 60 person small little startup in in the Boston area and we were taking radiology from an analog um filmbased practice into a digital world and actually at that time um this company was building their own graphics cards and um we were working on these new systems and um I moved from not only being being an engineer at the company, but I had this wonderful experience as an engineer to enter into our demo room and actually see a radiologist interact with the systems we were building.

And this radiologist just sort of described and said in his own words that I don't know if I would see this or see this this clearly without this technology. And it was this moment that kind of caught me to say, you know, I'll never be a doctor. I admire what health care professionals do every single day. We're all patients at the end of the day and I admire them. How could I use my engineering background and technology to serve them?

And and so with that sort of experience of seeing the real world impact of what technology can do um is where I sort of took the leap into more of a product based role where now it's was sort of up to my responsibility to say what are we building and why and that was the moment I said why are we building graphics cards? there's these companies out in Silicon Valley that are doing it with, you know, thousands of of engineers and I could see through my technical background that um if we asked Nvidia to make a few changes to their platform, we could use it for this this medical application.

And so called up Nvidia, got a hold of a gentleman, he's still here. I just saw him the other day Ian Williams um as a field application engineer and described what I was trying to do and asked if they'd be willing to make these these changes in the product and and in fact they said yes of course we'd love to this is a really important uh area of work and if we can be a contributor to it we would love to and um it just that's how I met the company and it was about six or so months later that um Nvidia was really coming to understand that they were evolving from a gaming company and a professional graphics company into an accelerated computing company and there was some early signals that um radiologists or scientists around the world were trying to use invidious technology for healthcare applications and so we wanted to um really discover what were those applications.

So they called me up and they said, "Hey, we've built we've built something. It's accelerated computing platform. There's researchers in the healthcare industry trying to take advantage of it. It's hard. Um, we need to go discover what's hard, what they want to do with it, and see if we can make a bigger contribution. Would you like to do that for us?" And I said, "This is a dream come true." Uh, and so the rest is history.

I guess that was back in April of 2008. And when you joined, roughly how many people were even at the company back then? >> I want to say it was probably between 1500 and 2,000. Somewhere in that ballpark. >> Wild just given how how big the company has has gotten. So you must have thought, okay, there is definitely potential here in healthcare, but they're not really an established player yet. So, what was your first couple of years like in the job when you know nowadays I'm sure if you're like hey I'm Kimberly from Nvidia people are like cool like let's chat but back then what was it like trying to infiltrate the health care sector which can be notoriously difficult with newcomers especially from the technology industry?

Yeah, it was it was not as hard as it might seem partially because there was um in the in the radiology world, it was digitizing and radiology and and diagnostic imaging companies like GE Healthcare, which was the first um partner that I called when I joined Nvidia, was leveraging our technology to do what's called image reconstruction. And so this is essentially using a lot of complicated math to take the sensor information that's coming from X-ray or CT or MR or ultrasound and turn that sensor information into what they call anatomical images, right?

Something that a human can understand of what the anatomy looks like. And so this complicated math and we built this new um accelerated computing platform and this way to program a GPU uh with a language or an API we call CUDA. And so it was it was a challenge in that um the math and the mapping of the math to the GPU was still in its infancy. CUDA was in its early early days and so that was challenging but the need was there.

Um, and so that was a, you know, somewhat, if you will, a a niche application, but actually not really. What the company came to understand is, um, you know, because other industries were trying to do the same things. If you if you think about taking a CT of a patient, it's kind of the same thing you do when you're looking for oil in the earth and doing oil and gas exploration. So we were actually helping to build new math libraries that had applicability in healthcare but it also had applicability in these other industries and um that was a a really cool experience too and and it's really how Nvidia has operated over the last um 20 years is what's happening in these different areas and how can we apply it into healthcare and so I think you know it wasn't as challenging as as it seemed there was early adopters s and it helped us build because we're tackling some challenging problems.

Um it helped us build um the next generation of Nvidia that had applicability in a lot of other industries. So it was it's it's really kind of wild when you can, you know, look what's happening over here in self-driving cars and say, well, a self-driving car and a surgical robot are actually quite similar. How can we take everything we've done in the car industry and apply it over into surgical robotics? And so that's that's kind of what we do every single day is take the most important technological breakthroughs of our lifetime right now.

And how do we make it more um approachable to the healthcare industry? So we're not a healthcare company, right? We're not trying to be a healthcare company, but we are building a lot of tools and computational platforms that can help companies achieve new things. And that's that's really what um our mission statement is. How do we take all these amazing technological breakthroughs and make it more approachable and accelerate its utility inside these different healthcare domains? >> Yeah, I think it's an interesting example with with self-driving and you see this example getting made a lot with the surgical space where we're talking about using, you know, AI and and robotics to guide surgical processes.

And then we're also talking about it in the broader care delivery concept of how autonomous do we see these tools becoming. And I think there are lots of schools of thought on this. We've seen prominent VCs come out and say we're going to replace physicians that this is inevitable and and maybe it's not going to happen on on the the timeline. Um Venode Kosa if you remember that. um it's not going to happen on that timeline, but he still maintains that that is the path that we're going down.

Similar to self-driving cars or if we want to achieve the the real safety gains at some point, it's just going to be the technology, right? Um and not the human beings that are prone to making mistakes. With medicine, you know, it's obviously complicated. There may be some areas and especially you mentioned radiology, um areas like pathology where aspects of the job get replaced. Um, but I it's just hard for me to imagine as somebody who's been in the industry a long time that we are going down this path of like fully ripping out and replacing positions.

So, what's your school of thought on this Kimberly? Um, especially as you've, you know, looked at areas like surgery. Um, you've also looked all the way through to that through to things like the partnership that you have with with a bridge um which is more about um supporting um physicians with documentation. Um, so where do you view view things going in terms of this idea of augmentation versus um whole who wholesale replacement? >> Yeah, I think uh you know Jensen says it really well and so I'll kind of describe it in the way he does because it's super approachable is you know a radiologist for example is trained to help diagnose patients and figure out what is the next u critical path for them to get get well.

Um what AI and radiology has done is maybe it's automated a particular task of being able to find the anomaly in an image or segmented but the task is not the job. Okay. So that's one way to think about this. The other way to think that we think about this um is with the more we're deploying AI the more it's increasing the need for healthcare services. And we are in a global crisis right now with aging population with managing you know um chronic disease with a huge health care professional shortage.

We're in that mode right now never mind what the next you know 5 years looks like and the fact that by using AI systems we're discovering more things we can fix and will need more healthare services. we're actually exacerbating the need for health care professionals. So in my mind there's a huge amount of work that is going to be coming and the only way that we can sort of meet this pent up and growing demand is through automation.

So that that's one way to think about it. I also think that, you know, as you said, a bridge and and open evidence and working with these um these digital health companies who are being super thoughtful of the day in the life of a clinician. And how can we take away any of the mundane administrative annoying um timeconsuming aspects of the job and open up their whole physical being, their mental capacity, the trade and the experience that they've built over the many years and operate at the top of their license in service of the patient. and it's not in service of you know making sure that the billing is correct or the revenue is getting to the to the hospital in the appropriate amount of time and so this is where we are creating a whole new experience and I think that as many of us even in you know the best parts of the world the clinician experience is very challenging and we have very high burnout um the patient experience also suffers be because of that we either don't have enough clinicians or the time they can spend with us is so so few um that this is an opportunity to put um digital and AI systems and agentic systems to work so that complete new experiences and maybe the way you imagine how medicine should operate can can get back to operating that way and and companies like Open Evidence and and a bridge and Heidi Health I mean they're all over the world are taking advantage of something that might sound simple and that is voice as essentially becoming the new interface of health care.

You know, the first thing you do even when you're not feeling well or you're just trying to take care of yourself is you're talking to somebody in the healthcare world. And so being able to use that voice to then automate or or or make a journey more navig, you know, everything is queued up about you into your, you know, your doctor's um potential phone. um because you know an agent was working in the background pulling all of your medical history and then knows what you're coming in here for and can really kind of get to the bottom of of anything that you're trying to solve.

I mean these are experiences that are well within reach that by working with these incredibly innovative companies who um really understand you know the crux of the problem that you know this is domain specific stuff. It's it it is complicated, but we now have technology that we can marry to that complexity and and create completely new experiences. And I'll I'll take you with with me on the ride of what do I also imagine, you know, the inside of the walls of hospitals to look like in the future.

Um to your point that surgical robotics and having you know right now we're in robotic assisted surgery and it's very helpful and it has incredible clinical outcomes and we will continue to be on this journey for decades to come to improve new you know surgical procedures that will be accomplished through robotics. But think about how the hospital itself should be a robot. The hospital itself should have eyes and ears sensing everything that's going on and using the modern world foundation models and spatial intelligence to operate hospitals.

So you'll, you know, the hospital itself will be a robot. There will be, you know, dozens and dozens of robots that are running around the walls of the hospital to make sure the patient has their medication on time or deliver the the surgical tools to the operating room exactly on time so that the the procedure can start. And again, here it's all about efficiency. And again, you want the patient to be under anesthesia the shortest amount of time. let's make sure anything that happens before the procedure and after the procedure helps facilitate that.

So this is coming through now a physical AI um world where spatial intelligence married with robotics can allow for yet another level of automation. So again nurses who have dedicated their life to the practice they aren't delivering sheets to an empty bed. Instead they're having that you know bedside time with a patient you know helping them get well as fast as possible. I think that's the future that we're going to see here not too far out. >> Wow, that's that's amazing. >> Hey everyone, we'll get right back to the conversation after a word from our sponsors.

If you've been following the show, you know that I've been using Granola for about a year now. And if you are anything like me, juggling backtoback meetings, you know how hard it can be to be present. You're in one call, you're trying to actively listen, but another part of your brain is still in the last one, and another part is jumping ahead. Granola changes that. It's an AI notepad that securely transcribes what's happening in the background.

So, I don't have to choose between taking notes and actually listening. The notes it produces are structured and clean. And I walk out knowing and being able to tell my team what the context was, what was decided, who's doing what, and what are the follow-ups from that meeting. So, if meeting up your day, Granola is a no-brainer, and you can try it out totally free. Just head to granola.ai/ipers. That's granola.ai/lifers to get your time back.

Try it out for free at granola.ai/lifers. For years, the most discerning people have quietly kept two healthcare lies. There's the traditional doctor they see when something is wrong and the holistic world they actually believe in for everything else. Two sets of advice that rarely speak to each other. Baker Health ends that double life. You no longer have to choose between your real doctor and your holistic one. Because at Baker Health, they are the same person.

This is longevity rooted in science where rigorous evidence-based medicine and genuine holistic care finally sit under one roof. Your doctor runs your labs on site, walks you through every result, and is just as likely to send you down the hall to the studio for yoga or meditation as to write you a prescription. Nutritionists and therapists work right beside them. And the technology is unlike anything else in medicine.

Your doctor, your labs, your results, and your same day reservations all live in a single app. Text that doctor at any hour and hear back in minutes through the only 24/7 live synchronous chat. Settling most small worries right there without an unnecessary visit. A real doctor always in your pocket. The proof is in how people feel about it. Net promoter score of 91 almost unheard of in medicine. This is excellence made personal.

Imagine healthcare reimagined. Baker Health. Um, so I one of the stories I'll share with with our listeners is um so for this podcast I actually reached out to Jensen um uh directly because uh I don't know if he told you this but I met him a number of years ago at an event um in San Francisco. It was a was a conference and we spent um a really fun dinner um with a with a group talking a lot about healthcare and he expressed a lot of passion enthusiasm for the sector but this was before a lot of the AI rush.

It was like in 2022 I believe so um certainly an important CEO in the room but it wasn't well nobody was really talking about AI all that much at that moment in time. Um, and I think I may have given him uh one of his first Uber rides home. Um, I'll tell you another fun thing about that story is the Uber driver asked if he should buy Nvidia stock. And I think about this guy all the time like [laughter] like imagine asking that question back in back in 2020.

Um, he's probably living in an on an island somewhere. Um, which really happy. But um I you know I think often of of that time because we must have spent you know two hours just talking about healthcare and all the all the potential and I I didn't see it then with AI like I see it now. Um, and I'm wondering if you can comment just on, you know, from your perspective, how important AI is to the company today when you've got or health, sorry, how important is healthcare to the company today when you've so much potential across, you know, you must be getting polls across every industry out there.

So, how how important is health care given just all the vast potential to be doing so much right now? Yeah, I think you know as as you described Jensen has been passionate about the healthcare industry going back you know 20 plus years. Um and that was really I think fueled by seeing just researchers who had dreams and aspirations to do something very very meaningful and impactful in the healthcare domain and us potentially having you know the tools to do it to achieve that goal.

Um some researchers have said things like I will actually be able to get my life's work done in my lifetime because the accelerated computing platform that Nvidia had created and now of course artificial intelligence can really just accelerate you know what people are are thinking about and early on it it also is the case as I talked about that um image reconstruction early application and use from of Nvidia technology when you tackle really hard problems it teaches you what your platform or your products aren't good at yet.

And it helps you see the future. Um, if you're kind of just tackling the easy stuff, you're not evolving what you're building. And when you can evolve what you're building, you can apply it in so many different areas. And so I think one, you know, whenever Nvidia goes to think about working in a particular area, we often ask ourselves three questions. Is it really, really hard? because if it's easy, let's let somebody else go do it.

We have worldclass engineers that want to work on really really hard problems because they have the talent to do so. The other is is it have an incredible impact and in healthcare that's an easy one, right? You're of course it has an impact. It has an impact on patients. It has an impact on the medical practice. It's an incredible um opportunity. And the third is does Nvidia have some unique capability to serve it? And so that's when I describe like all of the things that Nvidia does, our general robotics work, our self-driving car work.

Um we that is unique capabilities that it's sort of my job, my team's job to say, how do we take that uniqueness and get it into the hands of healthcare and then it creates that virtuous cycle when you have this unique capability and they go try to use it and it doesn't quite work right. You improve the technology and so it goes, right? And so we we've recognized that working on really hard impactful problems with a unique lens on how to take take what we've built and apply it um magical things can happen and you know a lot of the a lot of the libraries that we built for healthcare have huge applicability in other areas.

So we don't really see it as you know a siloed or a completely um vertical effort. It has it's built on a ginormous if I could use that word a ginormous contribution of all over Nvidia. So all over Nvidia is contributing to healthcare and then healthcare is contributing all over Nvidia and that's really the the beautiful design of the company. >> Yeah. And you said it correctly that healthcare is hard and it's hard for lots of reasons.

It's it's hard to access the data which is oftentimes extremely siloed. Um things are very inefficient. We have lots of middlemen in the industry. there is inertia um and you see entrepreneurs just really struggling in this space more so than in other categories. Um you know one example we give a lot in the industry is you look at fintech we we have incredible ways of of accessing our financial information within seconds.

But in healthcare, it still continues to live in dozens of different systems that were each require their own login and information is just not seamlessly shared. Um, so as you think about the potential here and and you mentioned AI as as one crucial area, how does the industry need to adapt to be able to make that future that you described possible? Which by the way sounds incredible and we need this yesterday, but um, like I said, we just have all these challenges that we've had for a long time where, you know, even throwing kind of vast amounts of money at the problem hasn't hasn't really solved it because it seems to be rooted in the incentives.

Yeah, there's there's some of that for certain. Um, what I believe now in this current moment and why it's so exciting is the advent of agents. Okay, where people think you kind of have to what is an agent? An agent is a computer system that can actually do work. This is this is the moment where AI has actually become useful because it's it's doing work. It's not just generating a picture or you know writing an email.

It's it's able to um do very very complicated work and that's only really happened in this last six months to be honest. You what is an agent? It's it's more than a model. Yes, it is a model but it's also how do you support that model with all of the tools and the data necessary for it to get that work done. And so because that hadn't arrived before, you do think about all of these maybe they're antiquated and necessary systems that run the hospital, right?

There's scheduling systems, there's electronic health records, there's all sorts of systems that you would otherwise have to try to write software to interact with them. But these agents have the capability now through API calls to kind of go out and pluck just the amount of information it needs to give the agent the context to move on to the next piece of work to get done. And so I believe that we're at a very critical moment in the healthcare industry that this is a great unlock. you know, health records have gone into practice here in the United States for over two decades, and as you say, we've largely not been able to get any true value or benefit out of that digitized data.

All of that is changing now and you can see um the companies in this space are able to achieve completely again new experiences re-imagine the workflows that they've otherwise had to go through to get their everyday job done. And that is in essence because we have this capability to tap into these systems in an agentic way. Um, and it's it's also the capability that allows for there to be they're safer for health care, right?

Language models and and general intelligence is a is a fantastic thing for the world, but it's just that it's general. We need to have the capabilities to domain specialize these because as you just described, healthcare is super complicated. Not only is the processes complicated, the domains are complicated, the amount of medical knowledge, sheer medical knowledge that's being produced on a, you know, a minute-by-minute basis and staying up to that is complicated.

And so we have the opportunity here to use agents on on our behalf, synthesize that knowledge, you know, automatically schedule that appointment, make sure you're you're doing the appropriate prior authorization questioning during the patient visit so there's no delay in in future um you know medical care. So that that's what we're at this incredibly opportune moment to take advantage of. Agentic systems unlock in a way.

It's not it's not in a in a way that um it's it's going to um be challenging for the privacy because the data stays where it is. It's just a bit of context that's needed for these digital systems to operate. And so no longer do we have to think about how what if we could pull all of the healthc care data together to you know build and operate AI. It it's not it's not necessary anymore. These agentic systems can be deployed inside the walls of the hospital if necessary.

They can be deployed all over the world and as as you say you can just kind of give the model essentially what it needs to get very domain specific complicated work done. Um, and I think that's the true transformation. And it cuts across whether you're a health care system, whether you're an insurance company, um, whether you're, you know, a general practitioner's office, it it really, uh, cuts across all of healthcare. >> If, if you are a clinician listening to this pod and thinking, Kimberly, how do I even adapt to this future that you're talking about where the hospital is a is a robot?

Uh, maybe I'm I'm not super technical. um don't have a coding background using an interface of an LLM but um or using some of the tools that the health system um has procured for you but maybe not necessarily as sophisticated with um things like prompt engineering. How would you advise them to ensure that they stay as relevant as relevant as possible in their profession and then also provide really high quality patient care at a time when frankly I think a lot of people are fearful of AI um especially as you know this this concept of it's going to take your jobs like this is such a big refrain that that we read about today.

So, how would you advise the clinicians listening into this that want to use AI, maybe there's some aspects of it that that scare them and others that that create a lot of excitement and feel like a big opportunity? Like, how how do you adapt to stay, you know, um as cutting edge as possible in this moment? >> Yeah. I mean, it it's it's oftentimes said that um AI is not going to replace you, but the person using AI is going to replace you, right?

Because AI should be thought of as this incredible tool and every single day clinicians have to learn new tools. They have to upskill their medical knowledge. Every every single day they are they are inherently have to be learning machines as we call them, right? The and so if if that is the case, what we oftenimes also encourage is just start using it. And to your point, you don't have to know Python. You don't have to write code anymore.

You simply speak and you can interact and interface with artificial intelligence today. And that's what we we just love the idea that voice is this new interface. You don't have to learn any UIs or UX's and know what buttons to push. Um you simply speak and you can interact with this intelligent system. And then when you're when you're speaking and you're starting to see actual work get done, I think that's the experience that we I would recommend everybody have because just as you said until you try it, you just can't quite put your hands on how it's going to feel or what else could I do with this thing.

And it it's really quite magical. You know, the the uptake of systems like open evidence is very profound. Never has the health care industry been the fastest at adopting a technology than right now. And why is that? Because open evidence is solving a very critical need of, you know, putting the world's current medical knowledge in the hands of their doctors on a simple voice activated interface on their phone. There's literally no barrier to entry to using this technology.

And so my recommendation is to do just that. Just use it. Use perplexity, use codeex, use claude, use open evidence, use a bridge because you'll see what the potential is there. And as as we hear from A bridg's you know endusers how lifech changing it's been for these clinicians um to be able to just work and have a much more um personalized time with their patient eye contact instead of typing to be able to actually go home and have dinner at night with their family which they otherwise would be typing for the next three hours. or to say I I I recommend an MRI for my patient and I see that they are authorized to have it right there in that patient uh encounter.

So there's no delay to the future care that they're recommending for their patients. All of that comes with just incredible I think fulfillment and is going to enhance the practice of medicine because for the last you know couple of decades we've been putting a lot of undue pressure unnecessary pressure or maybe it was necessary but it was unfortunate pressure on the health care professionals to do so much administrative work that it it's it's changed the practice and and as I was describing before I mean this couldn't come at a more important time when we are entering into these, you know, population, you know, demand for health care with a slowing supply of health care professionals.

We've got to meet this need with these new tools and that's how they should be thought of as new tools. >> Okay. So, so Kimberly, I've been watching the self-driving space and we've mentioned it a few times. Um, and one of the things that we definitely saw with the advent of this technology and all the years of focus on R&D, um, was that whenever a self-driving technology was involved in any kind of accident, cue front page news, right?

Um, meanwhile, you see human beings um, creating fatalities on the road every single day. This is such a common occurrence that, you know, it might make the news, but it's certainly not front page news. So I think about that sometimes with health care as well and it did actually happen um where OpenAI is now facing a lawsuit from a patient that experienced a very severe pulmonary embolism and was talking to an LLM um and this LLM was was telling him that he could rest that you know the Lord would provide.

I believe this individual was um religious and so despite the advice of friends and family did not seek medical care. Now, um, OpenAI would say, well, we're not intended for diagnosis. We're not supposed to replace a physician. And the models have also significantly improved. This patient was using an older model. Um, but I I guess the crux of my question is, uh, Kimberly, that mistakes happen all the time with human beings in medicine.

Um, patients fall through the cracks. They're misdiagnosed, underdiagnosed. Now, we have this incredible technology that people are starting to use. You also see front page news showing that patients are getting diagnosed with rare diseases that had eluded their physicians for decades. Um and so how do we balance the um potential for harm, you know, with with this case that is all over the news at the moment. um and the potential to do incredible things like how do we balance that to ensure it doesn't go down the same path as self-driving where we almost have this I would call it techno perfectionism that the AI has to be 100% perfect there is no room for error in order for us to out in any meaningful way and yet we're willing to accept any number of of of human mistakes um so I I realize that that's probably the very complicated question maybe the question of the moment But where where do you land on all this? >> Yeah, it is super complicated and lives are at stake.

Um, this is true in self-driving cars. This is true in in medicine. Um, the couple ways I think about it is it is so important that you have health care professionals in the loop. And I mean that in several different ways. One is even in using the utility of the tools, there's still a health care professional in the loop. And you should these these systems that we're talking about they know their user and they are pro trying to provide their user which is a health care professional all of the information to help them get their job done better right and so human is still in the loop.

Not only that, companies that are just incredibly smart about this, you know, a bridge and shiv, you know, a clinician himself being able to understand the practice so deeply and the domain so deeply to build that into the systems in a way that it's going to become very apparent that it's, you know, going to be so important that it won't it will help the the clinicians will realize that without this I might have made a mistake.

It's it's it's essentially their second opinion or it's they have a million second opinions that doesn't that make sense but they essentially have a million doctors in their pocket perhaps right because they've they've studied all the medical knowledge so it's not only the doctor's experience own experience that it can draw from it's the experience of millions and millions of health care professionals it can draw from and so you know with with every new technology there comes risk and it you have to think about what is the risk reward just as you're describing and um these these systems I think we want to deploy them very safely.

I think you know even as we go into knowing the user and making sure these platforms know their user and what the intended use is um it has to be it has to be very thoughtful. Um and that's why you know these new technological breakthroughs allow for us to guard rail these systems to audit them to know who is using it to know what information it um it built its you know it its thesis or its work uh that it it produced on and and so I think that that the way I really enjoy thinking about this is the risk versus reward is just too strong.

Um, as I described, we are all patients. Um, I live in an incredible place to receive health care. It's still very difficult for me to receive good health care. And so, the reward of digital systems helping me schedule appointments, the reward of my health care professionals having agents working in the background saying, "Oh, that thing that Kimberly just said might be something more than than I realize." um that is a reward I think uh far too great uh to ignore and so the riskreward here seems much much higher um and I think you also see this in self-driving cars now um it's not that often that you're seeing it hit the front page it's become a more accepted um techn technological breakthrough I mean I drive a Tesla I love my Model S it has full self-driving got what, eight, 10 pairs of eyes.

I only have two. When I get into a rental car, I'm actually like, goodness, I'm not I'm not nearly as good as a driver as my FSD because it has so much more information, so much more sensor technology, the ability to be have zero distractions or tiredness, right? And so that's that's a an an incredible opportunity and and an unlock for the industry. And and so it's it has to be thought through. It has to be discussed.

It should be you know regulated in the appropriate ways and and that's why exposing how these systems work educating people on what is agentic systems education and then you know firsthand use is what's going to allow for uh the evolution of of people's thought process here. Do you do you think these agentic um you mentioned just the role of agents? I I one of my both concerns and um reasons for optimism when looking at healthcare is that I think everybody's going to get their own agent.

Um like the patient will get their own agent, the physician will get their own agent, but then even within industries I think you look at revenue cycle management um you know which I often refer to as as the underbelly of healthcare because it's the flow of payments. There are definitely going to be agents on the provider side and there will be agents on the payer side and they will be much more efficient than two human beings duking it out related to a medical bill and so people have postulated that we end up in stalemate in that scenario.

So, is there a world in which um you know, if you play this out to its natural end point, like is it going to force the industry to finally have to sit down and have these real conversations and start to collaborate? Because I just can't see a future ahead of us where it's just non-stop agents in battle mode um between the various segments of the industry. Um it to me it just blows up the model. So, um so what do you think?

What are you seeing? >> I I definitely agree with you. Um, and I went to a bridges event where they just had, you know, they launched a new platform that is actually, you know, a clinical intelligence platform that includes the providers and the payers and the life sciences companies. And I believe agents and the fact that they can operate will create a new level of transparency that these conversations have to happen.

And in fact on their stage they had a payer a and a provider kind of talking about what is going to have to be different. I think it was um the the uh executive from Sigma that said of course you should know um everything about your policy and what's going to be reimbursed and and everything before you even have a medical procedure. Right now today that doesn't exist right. So I think the whole industry is thinking about this.

I'm deeply encouraged um by what you know a bridge has accomplished in this this platform that they're creating because it's the flow of information that has been the the critical breakdown and agents are this new highway and this new flow of information in radical ways that I do think the whole industry is going to have to transform uh because of it in in in positive ways right there's there's too much waste in the system.

There's too much latency in the system. Um, you know, there's too much fraud in the system that really can, I think, be addressed in a very material ways because of the agentic system. So, I I I'm hopeful. I believe it should be a very different looking industry in not too long a time. Um, but it should all be in service of of patients. I think we will all benefit from it. And so, the conversations are happening now.

I think, you know, having that a bridge event in New York City where they have um these individuals, these executives from these establishments having the real conversations in public, it's this is real and it's happening. So, I'm I'm encouraged that it's going to really happen in, you know, in our lifetime, right? I I've been I talked to a lot of people about like, what does success look like for you, Kimberly? I'm like when my daughter's healthcare just feels way different than mine.

It feels easy and enjoy and not enjoyable. I mean maybe because you're like I'm able to look after myself in a way that isn't arduous and painful and scary, right? And I think that's that's I think the opportunity with agentic systems and digital health that is in the not too distant future. I'm very hopeful for that. >> Yeah. Uh, so I I love the theme of the not too distant future and we've talked about this a few times um that you have the thesis of the hospital becoming a robot in the next five years and I I like five years and not 10 because it's just more tangible to think about what's around the corner.

Um, so do you have any other predictions for the next five um that we should all expect to see as as patients, as providers, as really anybody who is using the healthcare system in in some way, shape or form? >> I believe in the next five access to health care is going to be dramatically improved because of these digital systems. You know, it takes the world a little bit of time to get over the fact that maybe they are talking to a digital system, but in fact, like hypocratic AI, the benefit of talking to this digital system is huge for both again improving the patient experience and the clinician experience.

And so there there should be a massive deployment of digital health agentic systems that are just tackling the non-clinical parts of healthcare but the very necessary parts of how it operates. creating new patient experiences, creating new clinician and health care professional experiences, and just being in a digital world with more access to data, more real time, with a a workflow that feels right. It doesn't feel broken.

Um the the other parts that we're really really excited about right now is in the area of drug discovery. Um we've made the the industry has made some critical breakthroughs. you know, you you heard about AlphaFold and that really taught the world that biology can start to be learned and understood better using these new AI methodologies. Um, and for the last 20 years or more, we've been building a lot of measurement systems, right?

Whether that's liquid biopsy, new new ways of of measuring things in your blood or genomic sequencing or pathology system, new ways of measuring biology that we have this sort of opportunity that for the first time ever, we're going to be able to represent the world of of drugs and medicine in a computer. And when you can when you can bring biology in and you can bring molecules in and you can design molecules and and simulate them, you can uncover you can explore a much wider space and and we're really seeing that lift off now.

Um I would say in the area of molecular design, AI is now a standard. It's a standard use, but we still have some hard problems to tackle in the area of truly understanding biology. um building models like maybe you've heard uh efforts like the virtual cell. If we could truly model the cell, how could it help us, you know, understand and target, you know, come up with new targets um and and more causal understanding of how biology works?

Because in the end, we we do have a fundamental problem. 90% of drugs that go into the clinic fail. You know, we've got to improve that and we want to accelerate it. We want to provide the ability to get to medicines to the diseases that need them faster. And I think that this next couple of years, you're going to see huge advancements like we've just experienced in the language model world. We're going to see those advancements happen in the biology foundation model world and a representation of the world of drugs just massively accelerating uh our ability to have much better hopefully clinical um outcomes in in drug discovery.

So that's kind of like I I call that the two to fiveyear, you know, term. And and as we go here, you know, the delivery robots inside of hospitals are really starting to become a thing, right? That you don't have to build a new hospital to deploy delivery robots. Um NVIDIA's really helped the world be able to train robots in a computer for very different environments before they're ever deployed. and this new simulation capability um is really just come into fruition now.

So being able to kind of do delivery robots um I think that's in the near term and then it's like after that we're going to see some breakthroughs in sort of surgical robotic systems and and really automation in some of those areas. We're we're hard at work on there. But I see that as further out when you're kind of talking about clinical robotics, but but healthcare robotics at large, I mean, that's that's within the next several years.

There robotics is is the hardware itself is commoditizing in a in a very good way that we can put it to use. And then we've built all the AI computer and simulation systems to specialize these robots for really hard environments um like like hospitals and and healthcare delivery. Um, so I I think that's what I'm I'm super excited about. I mean, is it the here and now is digital health and putting agents to work. the up and cominging is being able to design molecules and medicines much much better more efficiently with much more success and then you know turning taking the physical AI the physical world of medicine and augmenting that with AI is something that we're going to need because as I said we're just we're exploding the demand for health care services and the only way we're going to meet it is through the deployment of these automated both digital and physical systems inside the healthcare practice. this. >> Yeah.

And and within the physical which you mentioned, I just see every country around the world has this need specifically around robotic caregivers just given you mentioned the demographic and the shift that we're starting to see where we have declining birth rates everywhere. We have populations getting older and and sicker and just more medically complex. We we are not adding net new clinicians quickly enough. So those are the realities that that we face.

Are you seeing that as well where you know anybody that is developing this kind of um robotics and and AI is um is really able to you know get funding from and support from governments around the world that I think are terrified about about this future and and our ability to support these individuals um especially you know without immense additional expense that that many countries can't afford. >> Yeah. I mean, we were just out at Computex and our our GPU technology G conference in Taiwan and they've really kind of created the blueprint for what this future of healthcare should look like with, you know, both digital and physical AI deployments. um companies like Foxcon and deploying them inside the environments of the hospitals, having their clinicians have access to these digital agent systems, using AI to read medical exams and um and just deliver a whole new set of efficiencies throughout the hospital.

Um I think it's a blueprint uh and they're they're adopting it. They're building it themselves and they're seeing uh absolute true benefit. Um, and so there's there's companies who are really starting to tackle this this challenge. And I I see it as a replplatforming of the healthcare industry. You know, you're going to when you think about what are the systems I need to run my hospital, robots are just going to become a natural part of it.

They'll and I and I don't even think they'll be seen as a hardware purchase. I'm not buying a robot. I'm employing, you know, an AI physical thing that is going to do work, right? And so I think that's also going to be a really interesting transition to go from sort of the capex to opex mindset in healthcare because for you know the last many many decades it's like I buy a CT machine and you know I pay for that once and then I surround it with a whole bunch of humans to operate that thing um to be able to charge for it.

Well, I I could see a future where you have digital agents and you're offering a service instead of offering a piece of hardware, right? And I think robots are going to be deployed like that too as just helpers and co-workers inside uh the walls of of the hospital. And so that's going to be a really interesting transition. But you know, I said it a I don't know what it's a a conference a couple of maybe two years ago or within the last year or two that AI is now hirable because it can do work.

You can hire it. And so this isn't like a software system that you have to go buy a license for. This is it has to be thought of as you know somebody who's on payroll. You know, my hope is that we don't just offset the costs like we say health systems you have to like pay for this technology and then they offset the costs by billing complexity goes up and everybody's just billing more and then this system becomes I think we're already starting to see that healthcare costs have gone up which is not what anybody wants because in theory AI should do the opposite.

So I think you hit the nail on the head. We've got to figure out a way in healthcare to pay for this technology so it doesn't just balloon the system even more. That's right. That's absolutely right. And so these are these are really hard and important conversations for health health systems to think through and and we're working with them on that and thinking about what is the infrastructure, what is the business model, how do you measure the ROI.

I mean, it's it's no different than every other industry, but but here to your point, this is a this is an industry that's really struggling. Um, and it's going to continue to get worse if we don't really embrace some of these modern technologies that can come to the rescue. >> Okay, a couple fun personal questions for you, Kimberly. One is, um, are you at all a health and wellness junkie? Like, do you use, you know, any of the wearable devices?

Are you out getting function tests ando tests? Like, what are some of the things that that you do in your own your own life? Yeah, I mean I kind of dabble in it, but then to the point of I don't feel like the tech is there yet, then I get tired of it. So I of course I have a smartwatch and I go into my Apple Health, but it's it's kind of useless. I did, you know, 23 and me and it's like, okay, it's telling me a little something, but it wasn't very helpful.

Um, and so I do go to DEXA scan and you know just partially because of my age now, you know, women's health is really complicated and I'm trying to learn more about it. Um, especially as I go into my later years. Um, and so knowing my bone density and things like that is actually really important. So I care about health and wellness a lot. Um, but I just actually feel like the tech isn't there yet. Um, Dexus get, you know, that's a very clear-cut thing.

It it measures something. I can tell, you know, if I need to do more lifting or high impact sports or things like that to really kind of keep up bone density, that's helpful. Or, you know, if my visceral fat is going in the wrong direction, it's like, okay, cut out the wine and cheese kind of thing. Like that I can actually I can take an action and see a change. But on some of these other apps, it's like it hasn't hasn't helped.

So, I'm hopeful, but I'm not overly junky about it because I just don't know how to make it improve my life. >> Yeah, I think that's been a historic problem is you get a lot of data and then what do you do with it? And then are the right people accessing it that that should be that can help you make sense of it? Um, I made some strides there, but I'm I'm with you that I I often still end up going down this path for a while and then I get bored of it and you know, it's not something that I sustain.

Um, and my other favorite question to ask anybody who's in AI and spends more time in AI than myself is, um, how polite are you when when you're using LLMs? Like I do you say please and thank you like um I'm just worried that I try to just because I'm worried that someday it's like gonna be my boss. Um, >> yeah, that's funny. What's >> your personal philosophy on that or are you just very efficient like hey get me the st? >> I am I am all over the map.

Um, there are times when it depends on what I'm working on that I'm like, if I'm asking for it to do work on my behalf, there's oftentimes a please there when it comes back and it did it wrong, it's like, why did you do that? And I get I get a little short with it. And so I'm kind of all over the map. Um, I don't I don't overthink it and I and I now much more just using my voice because when you're typing things, you're so intentional about what you're typing.

And I find now using voice and just speaking more freely actually helps the agent get to know you better and works better on your behalf. And then you don't have to be so thoughtful about your words. You just like talk talk talk and it goes and it kind of figures it out. So when you speak it versus type it to you're less rude I think you know because you don't want to hear yourself being a jerk. Uh, so I think in that regard, I'm just more cordial to it because I'm just kind of using it as having a conversation like we're having right now. >> Yeah.

And I was going to ask you my my last question is what is your favorite AI productivity hack that you use in your own day-to-day work? It sounds like you may have answered it, which is voice, but is there anything else that you would recommend to people that are starting to use AI at work? Maybe they use it a ton, but maybe it's something that you do that they may not know about. Yeah, it's probably not that novel, but I just love to, you know, I'm going anywhere, I'm talking to it, and I'm and I'm saying, "Prep me for this next meeting." And then I get the individual's name, and I I get to go learn about them really, really efficiently.

Where did they go to school? What did they study? You know, what what have they said in the news recently? And it allows you to be a lot more personal and present in the given conversation that it would take a long time to do that without like these agents going and just recalling all the information for you. So I feel like that's my personal favorite hack other than you know doing other administrative work. But it just it makes me feel like a more present and personal person in the meeting and who I'm meeting with.

So I I enjoy that part about it. I this is how I prep for this podcast. I was like, give me some of Kimberly's favorite and best uh moments and um it really gave me a sense of of who you were. And by the way, I love the talk you gave with Fortune where you mentioned the lack of funding for science and and that out. So um thank you for that and thank you so much for the time today. It was so enjoyable to speak with you.

If if anybody's listening that wants to get in touch with Nvidia or with you specifically, what's a good next step? >> Oh, you can find me on LinkedIn and it comes right to my work email. >> Amazing. All right. Well, thank you so much. >> Well, thank you so much. This is super fun and I love your pod. So, keep up the great work and all the awareness that we can build around AI and healthcare is this is the right conversation to have at the right moment.

And so, I really appreciate um you pulling on me to be a part of it. Um, we're super enthusiastic about it and I see a future where my daughter is going to have a way better healthcare than I have and that's a good thing. >> I I have two daughters and and one son, so I'm also rallying for that future and we will have to have you on again. >> All right, wonderful. Thank you so much. >> That's a wrap [music] on Lifers.

If you know someone else grinding it out in healthcare, send them this episode. And if you want more unfiltered takes on digital [music] health, check out the second opinion newsletter link in the show notes.