EP 09: From Silicon Valley to Healthcare: A Cloud Computing Odyssey with Timothy Chou

David McKenney
Hello, everyone. Welcome to this episode of Cloud Currents. I'm Dave McKinney and today I'm joined by Tim Joe, and we're going to be talking about a few things today, most of which will be involving cloud computing and AI topics pretty near and dear to Tim. So Tim, how are you doing today?

Timothy Chou
I'm good.

00:32 - Introduction to Timothy Chou

David McKenney
That's great. So you have quite the career work with Oracle, multiple board member and chairman seats, lecturer at Stanford. Could you maybe unpack your career a little bit for us before we get into some detail?

Timothy Chou
Yeah, well, it's that you're in Nebraska. I went to graduate school at the University of Illinois and lived through lots of midwest winters. Came out to California to work for one of the original Silicon Valley startups called Tandem computers. Some of your listeners may know I went from there, ended up doing, I call it, two tours of duty at Oracle, with the last one being the president of the cloud computing business. In parallel. I taught at Stanford University. I still do. I started teaching in 1982, computer architecture, actually, at a time when you couldn't get a bachelor's degree in computer science at Stanford, which is a weird thing to think about. When I retired from Oracle, I was hanging around the department, they said, oh, come back and teach. I was like, that's a lot of work.

So they said, how about a seminar class? I went, well, I've been a manager for a lot of years, why don't I do that? And so we started this class on cloud computing. Any of your listeners cs? Three hundred and nine a Stanford edu, and you'll get a little glimpse into it. I do the first and last lectures, which you can guess know pretty much the same. And in between I've guest lectures who are all ceos of public companies, starting out with people that I knew, like the ceos of Salesforce and Webex. But over the years, realizing that no one was offended to be asked, I brought in out.

I mean, last year we had the ceos of HubSpot, Informatica, Franklin Templeton, and we ended with the CEO of Intel that a couple of days you just said, I've been sitting on public company boards for 20 years, became the chairman of the Alchemist accelerator, and about three years ago basically decided to come out of retirement to work on the pediatric moonshot, which I can talk about later, but in essence it's a spin out of the class. We have engineered a next generation, privacy preserving, real time in the building edge cloud service in pursuit of the objectives of the moonshot mission.

David McKenney
Yeah, that's amazing. Yeah. I'm looking forward to talk more about that. For sure. I'm curious. You've got such a history in technology. We like to play this game at my household. Like dad, if you weren't in technology, what would you do? So if you weren't in technology, what would you do?

Timothy Chou
Well, first of all, it's really hard to think about because I think I wanted to be playing with computers from the age of twelve. So I don't know that I ever had anything else. But the one thing which I know everybody will find OD is I always wanted to design perfume bottles.

David McKenney
Oh, that's awesome. I find that in our field you get the most expected responses from being a chef to mowing lawns to whatever. Your journey in technology will definitely tell you how far you're willing to go to get away from it. Yeah, so many decades in technology you've seen a lot of milestones in the industry that probably have affected and formed your career. But what were some of those unexpected turns over these decades? Obviously you just mentioned the lecturer position at Stanford was a little unexpected, but seems to have turned out for all the better for probably your students, certainly. But what other expected turns along the way?

Timothy Chou
Well, I'll tell a story which is a good example of oh, we love serendipity or what expectation. So I'll take the listener back to it's 1999, if you remember. I mean, we are in the heart of the first Internet boom. Web, Van E, toys, everything. I am talking to a company who is about to raise $50 million if they name the CEO and they're going to focus on a website for korean teenagers. And I'm sitting there going, okay, that's kind of interesting. In parallel, I had already been at Oracle before. Some of the folks at Oracle said, well, we'd like you to come back and talk to Larry about running this new business. In fact, if you go back in history, we called it business online for a while. So why don't you come talk to Larry about that?

So on the very same day I literally had a conversation in a garage in Palo Alto with the founder of a company building websites for korean teenagers and that afternoon went up in the giant oracle Towers and met with Larry about running this brand new business.

David McKenney
What is it about garages and good ideas?

Timothy Chou
Yeah, well, just to tell the rest of the story, for a variety of reasons, it was nice to be able to say no to Larry three times, by the way, but after the third time I'm going, well, this is kind of stupid. What do I know korean teenagers and the media business. And so I made a fateful decision to say yes to Oracle. Going back to Oracle. And if everybody remembers the story within about a year, that's when the whole world, and in essence I was in there at the beginning of what we now refer to as cloud computing, meaning letting the builder of the software in essence deliver it as a service, ultimately hardware as well. But yeah, that was an unexpected twist.

David McKenney
That's a good story. I think it's one of just the number of story or number of ideas that have come out of people's garages and turn into billion dollar industries. So you started out a little bit there about what you're doing at Stanford. At looks very intriguing with your guest speakers. How do you get the list? I love what you said about it. Don't be shy to ask. I can't remember what you said exactly, but at some point after you started to get a few of your acquaintances to come speak, did the doors just flood open? Has anybody ever said no to you? Like as in they don't want to.

Timothy Chou
Just give you a sense of scale of this. We probably had at this point 150 unique speakers, public company ceos, which means I've probably talked to 450 to 600 different ones because I'll get to how we end up with the, we do eight a year or eight a quarter out of that. I would tell you I've only seen two cases where they said no, I just assume not.

David McKenney
And they probably have legitimate reasons. The list is, honestly, if somebody put it in front of me, be like, there's no way somebody gets these types of speakers. But it's pretty incredible, the people that you've gotten in your door to speak to students.

Timothy Chou
Yeah, obviously it's an interesting opportunity for them because I say, look, this is not a shareholder, an employee or a customer, which is who you're always talking to. This is an opportunity to talk to a different group of people. And so while we do have a technology conversation, there's also, I say, hey, you've got 150 bright young minds. If you'd like to do some career advice, go for it. Yeah, well, and recruiting. But I have to say I do that class as much for me as for the kids.

David McKenney
I can believe it.

Timothy Chou
90% of these people, I meet them on the day that we do the lecture, so I don't know what they're going to say. I mean, I give them an outline of what to talk about. But yeah, so it's always interesting stories, ideas coming out of that.

09:42 - The Pediatric Moonshot Initiative

David McKenney
Yeah. That's awesome. Well, I could definitely go into some more about the course. I'm really intrigued. Let's talk about pediatric moonshot and give a kind of a synopsis. You talked about it a little bit earlier. Real time health information. No matter where you are, I can already tell where the inspiration is probably coming from. But let's hear it from you and talk about where this project originated.

Timothy Chou
Yeah, well, the class is the origination. So a student shows up now, I don't know, seven years ago, and he sends me an email and says, I'm a little old school. I like to meet the professor. I went, okay. So we arranged to have breakfast at Joni's cafe on California Avenue for those locals around. I'm sitting there, he comes walking in. I'm looking at him going, doesn't look like a regular student. Turns out he has an MD, an MPH, an MBA. He's chief of pediatric cardiology at the children's hospital in Orange county. And so I'm looking at him know, why are you talking to me? There's nothing I know that could help you, right? And he goes, well, actually, I've been watching jeopardy. And I think that it's time for big data, cloud computing and AI to meet medicine.

And so he is actually enrolled in the bioinformatics program at Stanford. And that's why he is in my class. And for those of your listeners, check out Chang. He's a major voice in this transition in AI and medicine. The other thing that happened was Anthony is a major league networker. And so he kind of introduced me to their tribe. Just to give you a sense of this, there's only about 500 children's hospitals in the world. He runs a meeting called Peeds 2040. The very first time I went there, he asked me to do a keynote. I'm pretty sure there were 100 of those hospitals in the room, and it wasn't like a hospital administrator. It was this strange character who's a pediatric endocrinologist who wants to talk about graph databases. Right?

David McKenney
Wow. So multidisciplinary people.

Timothy Chou
Yeah. I remind people, the kids that decide to go to medical school or premed in high school, they were pretty much the smart kids, right? That was kind of the thing the smart kids did. Well, smart kids turned to be smart adults, too. It's not too surprising that some group of them is interested in graph databases. So anyway, so I got introduced to their world and I learned things like, they're still using cdroms to pass data around. In fact, there's a really sad story. It's a friend of mine. The kid went in for optional shoulder surgery at a regional hospital. There were complications. They decided to airlift him into the children's hospital in California, and he arrived, but the CT scan did not, and he died later that day.

Now, all of us obviously don't know whether having the CT scan would have made a difference, but it sure wouldn't have hurt, right? And so today, all the time, people are still using primitive technology to share data, right? Or share images. And then at the other end, and this is where you have to think, oh, in pediatrics, 60% of the rural counties in the United States have no pediatric expertise. I'm not talking about cardiology, orthopedics, like subspecialty. Like, none. Zero. Three states have no pediatric emergency physicians. Like, zero. If you go global, you go, oh. There's actually only 300 pediatric cardiologists india. And we all know how big that population is. There's one guy in Rwanda. So if there was ever a place in which AI could make a difference, you're looking at it.

Because we can't grow enough pediatric cardiologists or whatever, build enough medical schools, right? So the second observation was, shoot, we could use this technology to build AI applications, but AI applications require access to a large amount of diverse data. Otherwise you can't train them accurately. So that all came together when Covid kind of began. And I'm sitting around thinking to myself, oh, we're going to sit around watch Netflix all day. Maybe we could do something better with our time. So I decided to come out of retirement, launch the pediatric moonshot. So what is our mission? Our mission is to reduce health care inequity, lower cost, and improve patient outcomes for children locally, rurally, and globally. Now, how are we going to do that?

We're going to do that by creating privacy preserving real time applications based on access to data in all 1 million healthcare machines, in all 500 children's hospitals in the world. And by healthcare machine, I mean everything from ct, x ray, ultrasound, to gene sequencer, blood analyzer, ventilator, et cetera. So that's the mission, as I kind of alluded to, and this is cloud currents. We looked at, and we said, like the original moonshot, went, we need to build a new rocket to do this. And so we engineered a real time, privacy preserving in the building. When I'm talking to the Clinicians, I go, it's in the building edge cloud service. So that servers are in the building in children's hospital of Orange county. Why do they need to be in the building?

Well, the only way you can talk to the ultrasound or know gene sequencer or the blood analyzer is you have to be on the network in the building with that healthcare machine. So you could guess. We've had to engineer a lot of, we engineer 30 security and privacy features. One of the core team is a former student, has 15 years of privacy law experience. So we knew that was job one about a year ago, a year and a half ago, we said, hey, it's time to leave the lab and let's go into the world. So we deployed what we refer to as edge zones in eight children's hospitals on three. So we're way out of just hypothesis about doing this. Our focus recently is on what we call the Mercury program and the Gemini program. Right, in keeping with the moonshot.

 

17:18 - The Role of Technology in Healthcare Equity

Timothy Chou
So mercury is to build a global image sharing network to allow non children's hospitals, for example, rural hospitals, to share images with experts at children's hospitals. So primary first use case is emergency medicine. So just give you a simple example of this. There is a hospital in Willets, California, population 5000 north of San Francisco, by about 4 hours to this very day. When they have an emergency, they ship a CDRoM to UCSF or to Stanford or to UC Davis or to shriner's burn unit. And the future we're creating is, no, you don't need to do that. The application looks remarkably like Instagram. You'll be able to select the image off the x ray machine, no training required, and share with an expert, right in the speed of light, not four and a half hours later, an image.

So mercury is all about sharing with a human expert. And Gemini is our program to build AI experts, right? It's an AI research lab for children's medicine. The big innovation there is we have to figure out how to do decentralized learning. So chat, GPT, which obviously is the most visible example of this, all, basically assumes a centralized architecture. I'm going to take all the data, I'm going to move it to AWS or azure or whatever, I'm going to learn on it, and then I'm going to deploy the application there, as we all have seen. Well, that whole idea doesn't work in medicine. I mean, number one, the data sizes are enormous, so you're going to soak up a lot of network bandwidth if you're moving MRI or ultrasound images. Number two, what about privacy?

I mean, aggregation of large quantities of data is not privacy management. And in fact, when you look globally, countries like Norway are saying Norwegian data is not leaving Norway. Right. And then the last point is. Yeah, are you going to do real time applications from a server that's a thousand miles away? No. You want to have that deployed at the point of care, which frankly, could be on an ultrasound machine in an ambulance that's headed to an emergency room. Right. So the trick technology problem, which I think your technology listeners will be interested in, is, so how do you learn in a decentralized way? And so there's been a lot of work on consumer side in so called federated learning. So Siri actually works this way, which is, can I learn?

I don't take David or Tim's voice print and send it to the apple cloud. Why? Privacy network bandwidth. Right. So can I learn on what you said locally and only transmit model weights, which are just a bunch of floating point numbers to an aggregation server? And so Google keyboard actually works this way. And so we have done early experiments that indicate the same thing could occur in medical imaging. And so the lab is to bring this up and perfect the techniques for doing decentralized or federated learning in cardiology, orthopedics, emergency medicine, neurogradiology, etcetera.

David McKenney
Yeah, so that's fascinating. Just prior to that, you were talking about the need for almost like regulatory reasons as well as just overall, the size of the data. It really inhibits bringing it all to a centralized place that you're going after this edge solution. But now you've also brought in that sort of personal touch to it too, that because this has privacy rights beyond regulatory, that this is something that's personal to that whoever the owner is keeping it right there with them, but still taking the benefits of what these models, weights and parameters from AI bring. Yes, you mentioned, though initially, because I see that there's multiple things that you're solving or having to solve for here, the connectivity need to get past USB sticks and cds. And by the way, that story is awful, but I have personally had to do that.

I know that at one point for an MRI, I had to go to the bottom floor of where I was at. I actually had to wait in the lobby 30 minutes to get a CD burned at the time and then drive to a different clinic, I don't recall where, and had to do all that, but it was multiple hours. Thankfully, it wasn't a life or death thing for me, so I can only imagine situations where that's just got to be so defeating for people. But connectivity, that's obviously huge. But then the connectivity and the real time nature begets your next problem, where it's great I can get connectivity and outreach through or to a centralized set of high demand professionals. But now you're met with the problem that there's not enough of these people.

If you put these people, you book them every 15 minutes on a screen share to just to go around the clock, there's just not enough of them. So now you talk about bringing artificial intelligence to learn from these people. And all this is getting to one of my questions here. Around the AI side is humans around the world. And I'm going to say this, relatively speaking, humans around the world are still humans. We all kind of have the same makeup, right? I swear. Hopefully that sounds smarter than it did in my head there. But when it comes to the systems in first world, all the way to third world countries, they're very different. In America especially, we are very fortunate to have some very bleeding edge medical systems.

So I'm kind of intrigued by the data set that you're bringing into AI and what AI might be picking up and what it might have thought it knew simply by looking at a data set, say, in the Americas. But then when coupled with data that it's coming from, maybe third world systems that we're talking about, like MRI or imaging machines that are probably 30 plus years old and otherwise would have made incorrect assumptions. Hopefully you understand where I'm going here, but that's got to be a technical challenge here, maybe even a cultural one, too, with these systems being so different around the world.

Timothy Chou
Yeah. Well, let me say how we're trying to address that. So that's why when we said the mission is a million healthcare machines in all 500 children's hospitals, we're not trying to solve the problem right now about, well, what does a machine look like in a clinic in Kenya? We are. So just to use that one as an example, Gertrude's children's, which is one of only, I think, five children's hospitals on the entire continent of Africa, just to give a sense of this. Yeah, we're working with them. So the type of equipment that we would see at a Gertrude's or bambino Jesu, which is one of the largest children's hospitals in Europe, is not too dissimilar to what we're seeing here. Right.

So we have, let's call it at the same generational level of machines, at the same generational level of networking infrastructure, et cetera. So that's step one, is let's get the primary things connected and be able to leverage the data. I'll just give you a simple example of this. There's a condition called focal cortical dysplasia. If left untreated, the child has epileptic seizures. There's a kid in Florida right now, he's, I think, 13 years old. For the past twelve years, he has had seizures two to three times a day. At night he wakes up screaming. They MRI imaged him very early on, didn't see anything. He has been on a multitude of drugs, some that they could afford, some they couldn't. They were getting ready to put in electrical implants. They MRI imaged him again.

They now believe he has this condition, which is a brain lesion. And if you discover it, and I've actually seen the MRI images, I mean, frankly, David, if you and I saw it, we go, I don't know how that fuzzy thing is different than this fuzzy thing.

David McKenney
Probably looks like a white spot to us, right?

Timothy Chou
Yeah, it's just like. But okay, if it is, you can surgically remove it and the kids cured for life. Isn't that incredible?

David McKenney
It is absolutely incredible.

Timothy Chou
So the good news is, in the United States, this condition only happens 2500 times a year. Okay, that's good news. Of course, the bad news is no one pediatric neuroradiologist sees enough cases. So the idea that even the expert can miss is like obvious because you can't see this that often. Now, on the other hand, what if we had all the data from all the MRI machines in all 500 children's hospitals? We could build a friggin ultra accurate focal cortical dysplasia diagnostic that now every kid, whether it's an MRI machine in Kenya or here or in rural California, the machine could, you know, red, green, yellow, red. You really ought to go take this kid in yellow. Somebody ought to look at this again. Right?

David McKenney
So is this stuff that would otherwise typically number of medical dramas that you might watch on tv, but is this stuff that would otherwise surface, like in medical journals that somebody might know? Oh, I read about maybe one physician who had a case like this and let me talk to them. Is that really how it plays out today, prior to.

Timothy Chou
Absolutely, that's how it plays out. It's an accidental walk through the system that if you're a parent and you're persistent, maybe you gradually worm your way through to someone who goes, yeah, I know what that is. Increasingly in their community, they use the word rare diseases. And when you and I hear it, we focus on the rare. Oh, it's rare, but I'll tell you, my observation of this is everything is a rare disease. There's tons of this stuff out there that could be genetically diagnosed, could be Image diagnosed, and talk about lowering cost. We could severely lower cost. If you start diagnosing this stuff, early detection and early diagnostic is a way better answer to how do you lower cost? And by the way, improve patient outcomes? Yeah. In orthopedics, there are plenty of situations that they will tell you.

If they could have detected this early, they would have saved spending $100,000 on a surgery and the kid would walk better. We have the technology. It's so clear at this point, the technology exists to pull this off. That's why, while we call it a moonshot, at some level, at another level, it's just execution. We can do this.

David McKenney
Who do you think, in your mind stands to benefit the most? Obviously, there's the global connectivity of bringing these types of services to, as you've pointed out, places that don't have anywhere near this ability. There's also the one percenters, maybe I was, what I'll call it these very rare or not really well known diseases, or what you want to call it, there's no doubt an economical benefit. There's a lot of money here, but let's stick with those first two. Where do you see the biggest change that this would provide? Is it to bringing answers to those very rare diseases that don't get as much exposure? Or is it just bringing the overall service?

Timothy Chou
Yeah, I mean, what we're building, if it's not clear, I should say it again, we are fundamentally building a platform that could build applications from the very rare to the very mundane. Just to be clear about it, I think at the end of the day, where is the impact of what we're doing? And we can just stay in the United States for. Right. Know, if your mom works for Google and you live in Palo Alto, life's okay. I mean, what we're doing is going to make some benefit, but the impact is really the kid that lives in rural Nebraska whose mom does not work for. And, you know, that's where the gap is right now. This health care and equity you when you realize it's even here. I was just talking to pediatric cardiologists over at Stanford. You drive 1 hour south to Salinas.

There isn't a pediatric cardiologist to be seen. I mean, I'm just talking about the Bay area, right?

David McKenney
It's. Yeah, I guess it's when you've grown up near the facilities, you don't recognize that what you have so many, don't.

Timothy Chou
Well, and also you're know your mom works at Google. Well, her sister's friend is a pediatric know. That's, that's the world we all live in. Right? We're connected together. We're a social means, right?

David McKenney
Yeah. It strikes me that this moonshot almost presents as like the ultimate second opinion. Right. People always recommend, get a second opinion. 3rd, 4th, whatever. It's almost like this is going to be your first and last second opinion.

Timothy Chou
Well, in some cases it'll end up being the first opinion.

David McKenney
Right. That's incredible.

Timothy Chou
Yeah.

32:32 - The Evolution of AI and Its Ethical Implications

David McKenney
Well, taking the AI, let's kind of keep fast forwarding here and let's stay on the AI topic. So there's obviously a lot of ethics and regulations and things that you've started to mention in healthcare. But just in general, I guess what concerns you today with how AI progression has been handled, whether it's around ethics or just. I'll make analogy here. We all saw how social media and what it's done and how fast that launched onto the scene and some of the repercussions that we're still paying for because of very late regulations towards it, or even just awareness to what was some of the abuses and things that were happening. But do you think we're going to repeat that with artificial intelligence? Do you think we'll do better than some of these mistakes or are we? Buckle up. It only gets worse from here.

Timothy Chou
Well, if you can't guess, I'm an optimist by nature.

David McKenney
That's good. We need more of that, right?

Timothy Chou
I hope that we're better at it. One of the things just come back to the class. I mean, I reach out to the law department all the time. A couple of years ago, we had eight law school students in the class. Right. Because I think, number one, you can't have a conversation about policy and law and all these sorts of regulation in the absence of knowledge. And unfortunately, I think that's where we are. Mean, you know, you don't need to go too many Senate hearings and go, what does anybody on the other side know about any.

David McKenney
This point right here, you're spot on, because it leads to the people who create it, making the policy, which is very much a conflict unless somebody really is doing right by the world. It's a very interesting predicament.

Timothy Chou
Well, and I think at least in my history of tech used to be very much a backroom thing, right? Let's make payroll run or do our accounting and whatnot. I mean, obviously at this point, it's way out on the other side of this. So if the collective, we don't understand this, appreciate it, et cetera, then we will continue to, quote, make mistakes. So I hope, right, and I mean, I do try to do my part of let's get people smarter about what we're trying to do here, what works and what doesn't work. I mean, lately, just a simple example of this is I've been telling people, stop using the word AI, because all that happens is, I say most conversations, you could actually replace the word software that say AI, and it'd be the same thing because, oh, it's magical that it works.

Well, that's just software. Okay, fine. Right. Instead, the real technology break that's happened is these large language models of which I tell people, looks a lot like microprocessors to meaning there's not going to be a whole lot of people building them because they're not easy or cheap to build. So we're not going to have 3000 of them. We're probably not going to have three of them either. So then the question which, David, I'm not sure how old you are, but back in the day, we did have 1020 different microprocessor architectures. And the question then becomes, well, what are you going to do with it? Right? What are you going to build with this? And I think that's the question in front of us right now, is, yeah, llms are really cool. Chat, GPT, put it on my browser, all that. Yeah, it's very cool.

But how do llMs, how do we apply them? And obviously, I'm working in this, how do we apply it in medicine? How do we apply it in banking? How do we apply this technology becomes the question. And I think the hard part of this, which we all are going to have to figure out, is we have worked for, I'll call it 40 years, with deterministic software, debit and credit. If you're going to debit one dollars, you credit one dollars, not. We know how to do that kind of computing. Right. Which is deterministic. The world we are entering is not, it's probabilistic. Right. And how do you know that release two of your LLM application is more or less accurate than release one?

David McKenney
Yeah. Especially when they're being trained on previous data from previous models.

Timothy Chou
Well, and that you may choose to further train on all of these things. The broad we, I'm sure there's researchers out there who have the thoughts on it, but the broader we don't know how to deal with this. And I think it's in that pursuit of understanding how we're going to do this, educating ourselves, educating the consumer of these technologies, that's where this is. I say hyper important because we're trained that if it's printed, it's truth. Of course, we're progressively being trained that if it's printed, it could be anything.

38:04 - The Future of Personalized AI in Healthcare

David McKenney
Yeah, no kidding. So future state here, this is all very early in its days as far as the broad adoption for things. I mean, Chat GPT has done wonders as far as adoption goes, but you're right, most won't have the infrastructure to train a language model to that extent. But do you see, industry wise, that we'll see llms that are very specific to industries like banking and healthcare for a lot of reasons. I mean, there's probably intellectual property, there's probably privacy, regulatory reasons. But versus this idea that I'm going to build one ubiquitous LLM, that I can feed it a healthcare question, I can feed it a finance question, I guess, said a different way.

There will be a need for general llms like Chad GPT, but do you see that there will be a need for very specific, supervised, purpose built llms industries, or do you see it going a different direction?

Timothy Chou
So I say, I think as we today see enterprise applications of all ILPS, we will see LLM applications of all ILPS. And let me just give you an example that we're working on right now, which is, what if you took an LLM, trained it on all of the accepted research papers, et cetera, in cardiology, right? And then what if you took David's EMR records from the past two years and trained it on that now? I would actually have we'll call it chat David.

David McKenney
Personalized. Yeah, yeah.

Timothy Chou
Whether, by the way, that's a clinician who I've been through this recently, where you just laugh. I mean, you have the intern or the fellow asking you questions about your condition, staring at your electronic medical record. But once you see one of these electronic medical records, you realize why they're doing that, because there's no way. There's just tons of text and documents and whatnot sitting in there. They can't even read it, right? So even your history is lost, other than the context created by some doctor you've seen for three years. That's it. I'll say the context, the histories in their head, that doesn't need to be anymore, right? We could build chat David, where you could say, well, when was the last time that David had a CT scan. What's the relationship between his blood level and blah, blah.

Could you write me a 40 page summary of his medical condition? Could you write me a two word summary? I mean, there's so many ways you could think about this, but that's only going to happen if you can progressively train these things on progressively more. I'll call it fine grained data curated. Right. Et cetera.

David McKenney
But you're talking about something that a lot of people would love, right? It's almost like your own personalized AI assistant. It's like Iron Man's Jarvis. Like you've got somebody that you can. There's any number of analogies in the think that. I think most people would gravitate to a solution like that. The amount of things that it could do for you versus what you'd be giving up, it's amazing. But I can see a lot of people also being very concerned with the ethics of it. Right.

Timothy Chou
Well, you have to be conscious of how you're going to manage security and privacy. Back to why the infrastructure we're building has all been designed around security and privacy management. You have to have the infrastructure to do this sort of thing.

David McKenney
Yeah, it's really neat. Gosh, we've already bowled through an hour. I had a list of probably three times as many questions as we got there. We didn't even talk about really cloud computing and your outlook on. We talked about some futures state things. But I appreciate it, Tim. This has been amazing and I look forward to seeing. And I'm going to track the progress on the pediatric moonshot and see if I can find a way to snag a seat in your class. Hopefully that Stanford class is allowed for remote Stanford attendees. Somehow I can find a seat.

Timothy Chou
Yeah, well, the university knows how to monetize, so yes, it is possible.

David McKenney
Perfect. And right they should. It's all back to economics. Right. Economics. Back it all up.

42:48 - Final Thoughts & More on Pediatric Moonshot

Timothy Chou
David, can I do a final commercial message?

David McKenney
Absolutely, by all means.

Timothy Chou
So, the pediatric moonshot. Just to let everybody know, I've been funding our efforts over the past three years. The next step in what we're trying to do, we have estimated at 112,000,000. So I said, yeah, it's a little bit rich for me. We are not going to our venture capital friends who are not interested in pediatric health care, and we're not going to the children's hospitals because they don't have a whole lot of money. And in principle, we are building a network. And so one node in the network is not advantaged. So we are going down the twin paths of government sponsorship.

Some of your listeners may be aware of some new agency created called ARPA H, but we're also trying to talk to USDA, government of Singapore, government of the EU, because we fundamentally are building infrastructure, and that seems like something a government should fund. The other angle is corporate sponsorship. So I'll give you the simple shorthand. I was giving my friends at Juniper networks a little grief. And I've said, hey, you guys spent a million dollars to put a logo on a Formula one race car, right? And they go, well, actually was more than a million. I said, okay, you're just making my case for me, right? Why shouldn't corporations become sponsors of the moonshot? For any of your listeners who are interested in learning more, reach out WW dot pediatricmoonshot.com. You can register for a newsletter.

Timothy Chou
We've got podcasts that you can put in, show notes to show people about and for people to follow along or become part of what we call the moonshot crew. So appreciate you giving me that forum on cloud.

David McKenney
I went to the site. I was unaware of the project that you guys had been working on until now. And there's a lot of really great info on there, and it's a lot of stuff that really, people, I think, can relate. So that's fantastic, Tim. Thank you so much. This has been talk.

Timothy Chou
Thanks for having me, David.