October 2, 2023: Cameron Lewellen, Director of Healthcare ISVs at VMware, joins Bill for the news. They discuss the intricacies of VMware's approach to healthcare and the use of AI in the industry. Why is interoperability not yet a front-of-mind concern in healthcare? How is Cameron's friend Larry Ellison, Chief Executive Director at Oracle Corporation, reshaping our perception of healthcare through his innovative approach to data? With topics ranging from the role of AI in medical imaging to the concept of Intel creating large language models on laptops, this episode gives listeners a deep dive into the current and future applications of AI and data in healthcare. What could this mean for patients who need to modify their lifestyle as they age?
Join us for our webinar "Interoperability Outcomes: A Discussion of Whatβs Possible" on October 5th at 1 PM ET/10 AM PT, discussing challenges in healthcare interoperability. We'll tackle key issues like fragmented technology systems, data privacy, and cost-effectiveness. Engage with top-tier experts to understand the current landscape of healthcare IT, learn data-driven strategies for patient-centered care, and discover best practices for ensuring system security and stakeholder trust. Register Here -
https://thisweekhealth.com/interoperability-outcomes/
Alexβs Lemonade Stand: Foundation for Childhood Cancer Donate
This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.
Today on This Week Health.
The future is really bright because you're going to have the ability to have imaging systems that are literally tailored for the thing that they're doing. You individually are going to be able to go and to now take care of your health in a more transparent way
Welcome to Newsday A this week Health Newsroom Show. My name is Bill Russell. I'm a former C I O for a 16 hospital system and creator of this week health, A set of channels dedicated to keeping health IT staff current and engaged. For five years we've been making podcasts that amplify great thinking to propel healthcare forward.
Special thanks to our Newsday show partners and we have a lot of 'em this year, which I am really excited about. Cedar Sinai Accelerator. Clearsense, CrowdStrike,. Digital scientists, Optimum Healthcare IT, Pure Storage, SureTest, Tausight,, Lumeon and VMware. We appreciate them investing in our mission to develop the next generation of health leaders.
Now onto the show.
(Main) β π all right. It's Newsday and I'm excited. I am joined by Cameron Lewellen, with VMware one of my favorite nerds. Is it okay if I call you a nerd? Oh please. You and I can geek out. You're director Healthcare ISVs for VMware, it means you work with.
The ISVs for VMware. You look at a lot of really cool stuff, I would imagine. You look at a lot of these startups, and I would imagine spending a lot of time with AI type solutions and whatnot. We're gonna hit on some imaging AI stuff today and we're gonna talk about your friend, Larry Ellison.
Your friend. Yeah, my friend.
Personal
friend, now. Personal friend. Let's start there, actually, because you're coming back from Vegas. So you were out at Oracle's conference. And Larry Ellison is now, in your words and mine too, he's an influencer in health care.
He owns one of the major EHRs in the space. what transpired out there? What'd you hear from Larry?
Yeah, so I went to Las Vegas to represent VMware. I had a speaking session to talk about VMware's Aria product and extensibility. And of course I went to the keynote to hear Larry Ellison of this influencer that I, again, I'm getting familiar with as we all are in terms of what he plans to do with Oracle, but then also specifically in healthcare.
And what was crazy is that the benefit of genius like his is that if. If you're looking at healthcare from a database perspective, that is in some ways very new, right? We've been looking at patients and we've been looking at patient records and we've thought about how we improve the clinical environment, but interoperability, how applications really access that, the speed, some of those things haven't been front of mind.
And so when he started talking about what the database is going to do for healthcare, removing the walls between clouds, your data actually being your data. It was really cool because again, he is approaching healthcare from the perspective of someone who sees data as a superhighway and wants to have everybody travel on it.
Yeah it's interesting. , when you talk about that, what I hear is patients, so the patient's going to benefit from this in that it's going to be a lot easier for me to get my record. I'm not going to have to go in and, do all the things I normally have to do. Now, people will say it's gotten easier.
You could just go to MyChart and get it, but you can go to MyChart and just get your MyChart data. You can't get anything else. it turns out that over the years I've seen a lot of hospitals that aren't EPIC only hospitals. So, help me understand, and I don't know if he shared this or if he sort of looked at this.
I assume he's talking a lot about data and interoperability, which is what you shared. Is he talking about AI as well?
Yeah. So let's let's kind of look at it like this. So have you ever had a suit tailored? Okay, well, one of the reasons why you have to go and get a suit tailor is because it is going to fit you.
It is something that's designed for you. Now, let's think about the same thing being a database or your data, right? Unfortunately, right now, the way the healthcare system is, your data is over here, your data is over there, you talked about MyChart, not being able to really get something that really speaks to you.
Larry Ellison talked about data, in a cultured way, almost like for medical imaging, right? He He wants to tailor the database to be able to find cancer earlier, or for you to be able to know changes in your growth pattern that maybe are distinct from anyone else in the world, and he's accomplishing that because of something that's actually really old, but again, we have not approached from a healthcare provider.
Perspective, which is the vectorization of databases, right? So when you think about medical imaging, we started with 2D flat X rays and we're looking at gray and then again, the density of a particular area is where you're trying to figure something out. Well, we already knew that machine learning could do a great job at like picking out density better than you, right?
Then with less errors, but now when we think about instead of going from 2D to 3D, and if we think about instead of really just going from gray into color. Then we start to get something where spatial reasoning becomes another aspect of the way that you would read an image. And again, machine learning is going to do even better than you can imagine.
So the future is really bright because you're going to have the ability to have imaging systems that are literally tailored for the thing that they're doing. And as a result of that, you're going to have better, clinician support, but really you're going to get a better journey as a consumer of health.
You individually are going to be able to go and to now take care of your health in a more transparent way. Yeah. And
One of the stories CBC News AI shows major promise in breast cancer detection. New study suggests. And every study that comes out, it's just more and more promising in the in the imaging space.
And I'll be honest. The customizing the database, customizing the approach. I'm going to have to research that a little bit. It's a new concept for me. It's a new way of thinking. And I'm really gonna have to explore that a little bit. That's interesting.
That the progress AI is making in imaging is interesting to me. I was talking to somebody. And they said that the thing they're most excited about is not, somebody comes in today and it does the imaging and it helps the radiologist to look at the image and say, hey, this might be that kind of stuff.
Sort of AI as the copilot. to the radiologist, but the thing that he was really excited about is he said, look, we have 15 years worth of images and we can now send this thing off into a room like it is a trained radiologist and it can then go look at all those images and say, did we miss anything?
That's right.
But again, it can say, did we miss anything? But again, that's a version of like supervised learning, right? Unsupervised learning is you're just going to send the images into the room and just say, do you see anything? And what it could come back with in terms of patterns is that is unbelievable, right?
Ezra was an organization that was got FDA clearance. For an algorithm that I believe looks at prostate cancer better than, again, so when the algorithms are getting approved by the FDA, the standards are going up. But we're talking about, again, things that they had been looking for.
The perspective of all these other organizations getting in, using all of their images, is that we have a faster rate of improvement around all aspects of any image that they've ever taken. And so that doesn't mean just because, like we were talking about breast cancer. If I take a picture of my ankle because it's broken, well guess what we're looking for?
We're looking at the ankle as it's broken, but now with this, you would have the ability to look for other things that were wrong that we weren't even thinking about at the time. It's a perspective change because it's not going to look for the thing that we humanly were looking to fix. It could see something before it's broken.
It's interesting when I saw the Gartner now has an AI hype curve, a specific AI hype curve. Then they have the different AI technologies on the hype curve. What you're talking about this computer vision and the ability to look at images and stuff. This is way past the trough of disillusionment.
I don't think people recognize this has been around for a long time. When we're talking about generative AI, that's at the peak of hype. And we're going to see some disillusionment, Oh, it can't do this, can't do that. But this has many years of going through and studies a point where in the imaging space.
This is practical, this is usable, this is going to deliver results in the near term for for patients and for health systems.
So, the big difference between machine learning and AI, to me, always comes down to the math. And we've been doing math a very long time. So if take two images of a person and I compare them, the math is going to show me the delta.
The math is going to tell me, essentially how something has grown or shrank or whatever. And then you compare that math to the other math that we've had with everyone else in all the other images. That's really different. from the sort of consciousness that we're trying to approach with generative AI.
And what generative means is once it generates this data, does it create something that's really actionable for you? Is it something that you can use and turn a return on your investment? And that's why we get so many perspectives on how large language, models interact in different situations or why you can't really write a great.
Taylor Swift song with AI, or if you can, right? Some of it's so subjective, and we're still kind of
figuring it out.
The
National treasure, Bill. National treasure. I know, absolutely. I'm the problem, it's me,
okay? Absolutely. I, did you get tickets? It's impossible. Like, I heard the debacle, and then it just got to be ridiculous.
A buddy of mine is actually a promoter, and so I'm not going to say whether or not I got tickets, but...
So they have to watch the video to tell whether you got the tickets or not. curious, I've been doing a bunch of 229 project roundtables, and I now routinely ask this question, because when you bring up the word generative AI or just artificial intelligence, the problem is, Artificial intelligence now is getting lumped with generative AI.
And they're like, Oh, it's not ready. It's not ready. I'm like, hold on. There's a lot of different things around AI, and we have things baked into the EHR already that are AI based. So, it's not like, I mean, we've been playing with this for many years. I agree that generative AI is the thing that, you could go to any cocktail party now and it's like, Can you believe we can write a Taylor Swift song?
But I'm now asking this question, on a scale of 1 to 10. And one being... I don't know why we're talking about this. 10 being, oh my gosh, I think we're at the moment where we're really going to see transformation happen as a result of AI. in healthcare, as well as many other areas in our lives.
Where do you fall on that scale? Do you fall, just
specific to healthcare?
Yeah, let's talk healthcare. One to ten. One being, hey, we're not ready to talk about it yet. Ten being, yeah, this is this is really going to be a moment and it's really going to change things.
I'm going
to go 7.
why is that?
Yeah, so, in healthcare, we still have a long way to go because of social determinants of health, and so while I want to say that the technology that we're using for clinical support and the machine learning and AI for medical imaging is going to help, it will, but The system itself needs to be able to support the other people, the ones at the edges.
And so I don't want to act like an improvement in technology is going to all of a sudden, result in medical adherence or access to care for people in disillusioned parts of the country. I don't want to ever be that guy. At the same token it's... Phenomenal to think that new technologies that incorporate machine learning AI and specifically generative AI will be able to, provide on the spot translation services for someone who's receiving medical assistance outside of their home country or that generative AI is now going to be able to give suggestions for healthy behavior in the form of a text message or a TikTok or whatever for someone who's now managing a chronic illness.
And as a result of that becomes really powerful to me in a way that I don't think we can really even describe, because if you've ever helped someone who's elderly to write, to get care, and you're there as sort of a power of attorney in that particular case, you understanding it and being able to follow that is not going to be the same if that person can.
If they can adopt those changes. And as we continue to get older in life, we have to make modifications. AI can do that. AI can know your schedule and it can interact with you, looking at your actual reaction and start to create better human behavior. So the idea of say, a sandbox AQ, right?
Like the the company that's out there that's doing quantum magnetic sensing, of your of the body. The idea that I could go to Walmart and get a quick scan of my heart. That's incredible. And when that gets sent to me and translated and says, hey, for people your age, your height in this part of the country who have eaten your diet and blah, blah, blah, we see typically this kind of change in your ability your heart's ability to function.
And cardiovascular illness or 30 years. You put that in a meaningful way, Now, I'm a consumer of health. I went to Walmart, right? I buy one, get one free. Got some orange juice, got better on my heart.
You're a pragmatist. Who knew that No, but I like the way you, I like the way you're saying it.
It's like, yes, in some areas, this could be really transformative. And it's going to be transformative. pretty quickly. It's not like we're looking five, ten years out. It's going to be transformative. But the core of what health is isn't necessarily solved by technology alone. There's an awful lot of other things.
β π
β
We'll get back to our show in just a minute. Our monthly Leader Series webinars has been a huge success. We had close to 300 people sign up for our September webinar, and we are at it again in October. are going to talk about interoperability from a possibility standpoint. We talk a lot about what you need to do and that kind of stuff.
This time we're going to talk about, hey, what's the future look like in a world where... Interoperability, where data, where information flows freely. And we're going to do that on October 5th at 1 o'clock Eastern Time, 10 o'clock Pacific Time. We're going to talk about solutions, we're going to share experiences, we're going to talk about patient centric care.
And see what we can find out. We have three great leaders on this webinar. Mickey Tripathi with the ONC. Mary Ann Yeager, Sequoia Project. And Anish Chopra, who I'm just going to call an interoperability. evangelist, which is what he has been to me ever since I met him about 10 years ago. Don't miss this one.
Register today at ThisWeekHealth. com. Now back to our show. π β
think one of the most promising things I'm seeing is machines talking to machines. And I know this scares people, but the concept of layered AI models, just like your brain, and essentially I was talking to a health system that is putting in essentially all the logistics not logistics is the wrong word, but anyway, the logistics for a patient to go to the health system.
So, parking, location check in, all that stuff in. And one of the things that he was talking about is, we're probably going to be able to link that to drive times and traffic patterns and some other things. And so you can actually have machines talking to machines, creating a pretty good picture for somebody.
Because machines track how many parking spots are available in a... place, and that kind of stuff. And literally you can with a natural language front end say, hey, I need to get an MRI today. I'm thinking of going here, where should I go? And if the systems have been built to with that, if they've been trained, I guess it's the right word with that information, we're going to be able to create some really neat.
interfaces for patients. And you described one of saying, hey, here's the scan, how should I think about this? And it's going to give you give you a response. Two things. One, I'm just going to touch on mostly because I just want to touch on it for our viewers.
And that's The August flash report came out from Kaufman Hall, and it showed a slight degradation in margins. Now it's still in the positive territory. I mean, for most of last year and early this year, it's in the negative territory. It's still in the positive territory, but it went down from July, so just wanted to make note of that.
The story I want to talk to you about again, only because I don't think people know how much of a technologist you are. I think they're getting a picture the more we talk, but I want to talk about this Intel announcement. One, because I like Pat Gelsinger great guy. He went over to Intel.
He's now leading the charge over there. And they announced a chip. So in their announcement, a chip that essentially has space on it to run large language models and those kinds of things. And the concept is you're going to be able to have a large language model on your chip.
It'll run faster. And from a privacy and security standpoint. You're going to be able to keep the information right there on your PC instead of going out into the world. read the announcement. I mean, what are your thoughts on this?
, so Pat Gelsinger, a big fan, obviously, right?
And what did VMware do initially, right? We just gave you some sort of insight into your server. So, insight into your server really just translated into the ability for you to run applications better, right? And more efficiently. Well, fast forward. What is he doing now? He's just trying to run the large language model more efficiently.
So the idea of being able to run a large language model on a local machine a laptop, means that we can obviously get a lot better access. And then of course, if you can keep your data, we can also keep it more secure, right? One of the big announcements for VMware specifically was that we announced private AI a private cloud.
And so as data, in some ways has become more siloed with some of the major hyperscalers. We have people who are trying to open that up, and to give you a better option for the way that you access this new kind of generated content. But to be honest I think it was, this was like four or five months ago, there was, they were talking about when the Facebook large language model got leaked.
That someone was able to iterate on a new sort of large language models that was 8 billion parameters or less and do it in less than a day on 3D 100 NVIDIA chips. And those aren't state of the art by any... Shape, form, or fashion. Now that's a very large model still. But as we continue to really understand the value of the model, I think what you'll do is just in the same way you're talking about the organization sort of iterating and having machines talk to each other so that we can have better outcomes for the patient, you will essentially have the chip talking to the cloud, talking to the user that iterates in a better way so that the user itself gets a better experience for large language models on
laptops.
Talk to me about private AI. You said VMware announced private AI. We have this conversation a fair amount in terms of privacy, in terms of making sure that the doctors aren't putting information out there that they shouldn't put out there. We, but there's obviously a lot of information we have to protect in healthcare.
What is a private AI model? Talk to me about the VMware announcement. Let's talk about that specifically.
So, again, I don't want to wave our flag too much here, but let's go back to the friend of yours that was like, hey, we have a bunch of these images and we want to train, the model on it.
Okay, first, first question is, well, have you looked at the price of GPUs? When you were saying that margins have shrunk, let's be honest, it shouldn't be that big of a surprise. Health care doesn't exist in a vacuum, and so if the cost of money is high to buy a house, the cost of everything right now with inflation is going to be up, and we don't think at least no one that I've talked to, and I think JP Morgan backed this up in the last conference the cost of nurses has increased, right?
So. If everything else is increased in terms of cost then it makes sense that we would have reducing margins. Well, if the cost of GPUs is up, what do you think is going to happen to the cost of your data in the cloud once you start to try to shift massive amounts of images into the cloud to then start running models in it?
That's going to be a major cost. And have you, how do you account for that? No one's really dealt with that data set yet, right? So you, they could guess what it's going to cost you in terms of getting the model up there, running it, and how fast you're going to be able to iterate. But let me just throw a crazy idea.
Instead of having it in the public cloud, maybe you put it in the private cloud. Maybe you don't want to give your data to a hyperscaler, not because they would ever do anything wrong with it, wink, right? But because you feel better from a security standpoint of having that. So the idea of having private AI is that you could do this and run these models in a cloud of your choosing.
It sticks with the old VMware model of any device, any workload, anywhere. Except for the fact that now that we look at large language models and we really start to think about how they run, let's optimize it. I'm in favor of private cloud, but I'm also in favor, really, of you running it on prem because of those margins we just talked about, right?
on generative AI and, not get:You only get something far less. Joking, but I'm saying you're spending a lot of money without really knowing what the return on the investment is. And that's why it's located where it is in the hype cycle at Gartner.
No, it's interesting. A lot Healthcare has contracts with Microsoft, right?
So we all of us were using Microsoft. We had a contract and Microsoft's coming in and saying essentially, Hey, we can give you private access to open AI. But the reality of that is you do have those costs. You do have those costs associated with moving that kind of information or the amount of information you're talking about up there.
And one of the things I heard recently was just the the squeeze that was put on a lot of IT shops over the last year. There was a, inflation is real. And there was a significant increase in a majority of... the enterprise contracts, including the cloud contracts and the storage contracts and those kinds of things.
Yeah, I mean, I think a lot of organizations were so on the hype cycle and they started to think about training the models and then they said, wait, how long is it going to take to train? Now we've obviously moved past that. A lot of organizations are starting with a base level algorithm and then modifying it from there in terms of, tuning it.
We've got a great partnership with HuggingFace. Everyone loves HuggingFace and I think that's great. That's due to the way that you can train the model now, right? However, if we're really going to be honest, if we are looking at reducing or sorry decreasing margins, we really want to think about how we optimize that, and I would hate to be that CIO, that CTO that's throwing out this big plan for LLM.
Without really understanding what it's going to be in terms of cost, because we don't want to get this time next year and then say, okay, well, we weren't able to execute on other visions or optimizing the system for our current patients because we made a big outsource. And it's not like healthcare hasn't done that before but I would prefer to try to help as many organizations avoid that as possible.
Cameron Llewellyn, just amazing, great conversation as always. Coming off a conference, which I'm sure you got tons of sleep, was it pretty well attended? It really was.
It really was. So Oracle really can put on a show. I, again, I had not. I've been to an Oracle conference before, very familiar with Cerner.
Cerner was different, definitely slimmed down on the Cerner side just because of the fact that they're moving that into being part of the greater organization, but it was really good to connect with the people on the ground and talk about real life. use cases, understand the challenges that they have in using Cerner, what they want Cerner to do.
Oracle is really great about taking feedback and, from all of our, all the partners, but then really they were really invested in listening to the customers and what they kind of wanted to see. I'm hoping that, COVID doesn't have another iteration and that I get a chance to go again.
Maybe I'll see you there.
Yeah, we will see. Cameron, thanks for your time. Always great to talk to you.
Thank you so much, Bill.
β π And that is the news. If I were a CIO today, I think what I would do is I'd have every team member listening to a show just like this one, and trying to have conversations with them after the show about what they've learned.
and what we can apply to our health system. If you wanna support this week Health, one of the ways you can do that is you can recommend our channels to a peer or to one of your staff members. We have two channels this week, health Newsroom, and this week Health Conference. You can check them out anywhere you listen to podcasts, which is a lot of places apple, Google, , overcast, Spotify, you name it, you could find it there. You could also find us on. And of course you could go to our website this week, health.com, and we want to thank our new state partners again, a lot of 'em, and we appreciate their participation in this show.
Cedar Sinai Accelerator Clearsense, CrowdStrike, digital Scientists, optimum, Pure Storage, Suretest, tausight, Lumeon, and VMware who have π invested in our mission to develop the next generation of health leaders. Thanks for listening. That's all for now.