AI in Spine Surgery: Medtronic Fireside | Thomas Anglero

Please welcome to the stage guest speaker Thomas Anglero and chairmen Dr.

Christopher Ames and Dr. Jeffrey Gumm. Well, it's, uh, it's my sincere pleasure to welcome Thomas Anglero to the stage. He did a great job last night. The cranial team was bragging about him, and I told him it's a spine night tonight, so he needs to speak at a slightly higher level because the spine surgeons are here, not the brain surgeons tonight. So Thomas has had a great experience in the startup world and in big data. I think published a book in AI in 2022 and also captained the championship basketball team, I believe. Is that right?

Keep on going. Keep on going.

So I think it would be best actually, because your background in data is so detailed and long, that you tell us a little bit about how you got so involved in data and how it became a passion of yours. To the point that now you're so sought after to give talks on data. Before we dive into it, so maybe a little background on yourself would be great.

Should we sit down?

Yeah, let's sit down.

I feel like I'm in Broadway doing a show. So the answer is real simple. How I got into data was, as most of the men would understand, I fell in love with a 6-foot blonde. That's the short answer. So I was working as a mechanical engineer in New York City, and I started dating a Norwegian woman. We're now married, kids and all that stuff. Anyway, as all Norwegian women do, they love their country. She moved back to Norway. We had a long-distance relationship. It was $15 a minute to call Norway back then. That's a lot of money. And I could— this could be very long, but I'll keep it short. Anyway, because I wanted to save money and because I wanted to stay in touch with the woman that I love, one of my first claims to fame was that I'm the 10th person in the world ever to use the internet as a phone call. And that started me on this whole nerd IT data thing, and I was just an engineer that played with natural gas pipes in Brooklyn. So 6-foot blondes is the reason why I am where I am today. I'm just honest about it.

It's a good reason.

Yeah, I think Jeff and I are going to stay quiet on that, but we can probably both understand that. So let me ask you, as you envision the future of healthcare, being such an expert in AI and machine learning, and those are terms that are just thrown around so, you know, routinely these days.

Way too much.

Yeah. Give us your, give us your perspective on how we can leverage those tools, because they're basically statistical tools, to optimize the care of our patients. And what do you think are the biggest barriers, the biggest opportunities and the biggest barriers?

So listening to being here the last day and a half, 2 days, I got so much input from the inferior surgeons that were here the last day. Does that make you feel better?

Yeah, I know you had to speak at a slightly lower level for the I don't understand what's going on between you guys. I'm a neurosurgeon, so I can—

If I get beat up in the hallways afterwards, I'm coming after you. So let me tell you the picture that I see, and I cannot do— you were brutal, by the way, to the last bunch of people, so you should be beat up in the hallway.

They all got A's at the end.

I hope I fail. And I'm being dead honest because We— you want to fail. In your world, you can't fail because you're dealing with a human body, right? In our world, we don't want to know what the hell we're doing. We want to be blown away. For example, what's happening now is that the AI— if everybody knows about ChatGPT, everybody play with ChatGPT? Give me a yeah. Yeah. Who has never touched ChatGPT? Say no.

No.

You can leave. I'm kidding. Big hug. Google. So ChatGPT is the one that got most of the media attention. You know all about that. But then you have companies, and that's run by a startup company called OpenAI. But then you have Google who's in the race, right? And they have theirs called Gemini. And then Apple's coming out with theirs. Facebook has theirs. So the example I used yesterday, but I think is relevant, is that Google Gemini, which was called Google Bard, they launched Google Gemini 2 weeks ago, version 1.0, and it was better than ChatGPT. And they said version 1.5 will be ready in 1 year. It's gonna be fantastic. 2 nights ago, they launched version 1.5.

Google.

They had no idea that they could do it in 2 weeks. Why? Because the damn AI is writing its own code. It's going at an accelerated rate that we have no idea what's going on. It is doing things that we cannot predict. So that's why I hope I fail. We are failing all the time because we're succeeding at creating something that is— I won't say has a mind of his own. But it is accelerating at a rate that we don't even comprehend. Yesterday we talked about the IBM competition where the computer played against the Chinese game Go. Was it Go?

Right?

And there's a famous move. Anybody ever play Go? Anybody know this story? So there's a famous move in Go, and the IBM machine played against the Chinese, the world champion is Chinese, doesn't matter. And it was move, I think it was 37, that the computer did. And in the history of the game Go, no one's ever done that move. And that move, the announcer said, "That move was alien." Everybody freaked out saying, "What was that move? We've never seen that move before." And it was a defining move. When the competition is over and the machine beat the human being, they went to the developers and they said, "What did you do?" Because at that time, this story was 5 years ago. At that time, AI, what we did with AI, we gave it all the historical information. These are all the moves human beings have ever done so that AI should only duplicate or replicate. Move 37 was a move that it was never taught. That's where we are today. It is super exciting. Now to get to your point, just give me some background context. For surgeons and what do I see, I'm so excited for you. So I hope you guys get excited. I was going to say a joke about the guys yesterday, but they were really cool because a whole bunch of them are here. So what's really exciting is everybody's familiar with the Apple headset, right? How many people have it? How many people bought it for their kids? Who are the great papas and mamas? Nobody? Worst parents I've ever seen in my life. Go buy your kids some expensive— I don't even think it's $5,000. No big deal. So what's really exciting, when I was listening to the last panel, I was seeing this thought of— because one of them was saying, you know, I don't want all the data, I want the solution. So let me offer you this. With the Apple headset, what you can do is when you put the headset on, we can take all the knowledge of the best surgeons, all of you. If you spend time in training the headset, the software, we can replicate your knowledge and make you accessible to every surgeon in the world. Simultaneously. Wouldn't it be wonderful on a Friday night? I don't know what you guys do on a Friday night, but I make believe that you guys are so dedicated Friday night, you guys are practicing at home. No? Yes?

No.

Okay, say yes to make me impressed, all right? No.

Sure.

She's like, hell no. Okay, so imagine, and this is going to one of the questions we'll ask, imagine putting the headset on and as soon as you put it on, you have all the surgeons that you want to work with, and you can have a virtual patient, and at that moment, that— and I'm not saying they're online, they're offline, but it's been digitized. Your knowledge has been digitalized. And it doesn't matter that that one surgeon is Japanese and he only speaks Japanese. He's— it can be translated into English, and you can work on that patient. And we could actually do that during a surgery where you have all the knowledge of all the best surgeons in the world, whether they're alive, they're dead, available to you at all time. It's almost like— anybody seen Pinocchio and that little Jiminy Cricket? Everybody remember Jiminy Cricket?

Yeah.

You guys never— what's wrong with this group? Nobody's seen Pinocchio.

I'm sorry about the audience.

Okay, we'll just talk to you two guys. So Jiminy Cricket, right? So Jiminy Cricket, his job was to tell Pinocchio what's right and what's wrong. We can do that today. And your knowledge, your capabilities can be shared with every surgeon in the world. You can live— forever. Your knowledge goes on forever. It goes to, like, my son, he's in his third year of medical school. You can influence him. You can influence kids in grammar school. You could influence senior surgeons. This is where we're at. And this is not that your 20-year question is going, why this guy's caught up on 20 years? I have trouble with 2.

Yeah.

Anything beyond 2 is— if anybody ever answers you, say, I'm a futurist. In 5 years from now, just tell them to shut up.

2 years.

2 years is hard as hell. 3 months is comfortable. 6 months, I start sweating. 2 years, I am like so full of it at that point. Because again, 2 weeks ago, the world's most state-of-the-art generative AI was launched. 2 weeks later, when it was supposed to give birth a year ago, it was birthed 2 weeks later. This thing is going so fast. 2 years? If a year is 2 weeks, somebody do the math for me. You guys are smarter than I am, right? 2 times 52, that's 26. 26 times 2, 52. So in 2 years, this is the equivalent of 52 calendar years. Dude, that's half a century of knowledge. And matter of fact, it's not linear, it's logarithmic. So do the math on that.

What's the—

You raise a really interesting point. When you look at AI playing games like Go or, you know, playing chess, for example.

Yeah.

I'd always thought that the AI would just wear the opponent down with, like, a series of perfect logical moves. But actually, if you want to detect whether AI is playing the game, you need to detect the most creative moves, because AI is going to make the craziest, the move 37 from Go. And we saw in our own data that AI actually, or machine learning, taught us, gave us information about spine patients that we didn't know before. For example, we always thought there was a very intimate relationship between alignment and outcome. When we used AI to predict outcome, we actually asked our scientists, can you predict, since you can predict the outcome, can you predict the ideal alignment for that patient that would optimize the outcome? And he couldn't. And the reason was because it wasn't anywhere— the post-operative alignment wasn't anywhere in the top 10. And we wouldn't have known that without using more sophisticated data techniques to try to predict something. So I really like your, your, your thought about AI and even generative AI teaching us things that we didn't realize before. I think a lot of people look at AI almost oppositionally. Traditionally. But really, when you go through using the process of AI, it's going to— I feel like it's going to teach us things. Do you think the same about things we already considered that we knew in terms of looking at it in different ways?

I think a lot of guys in this room, they were using ChatGPT. I think it saved their marriages come this past Valentine's Day. Anybody use ChatGPT? As a matter of fact, there is a doctor in this room who told me that he actually on Valentine's Day used ChatGPT to write a poem for his wife. And he gave it to his wife. He printed out, gave it to his wife. She said she started to cry. Where are you, Dave?

Now he's in trouble, man.

Now he's in trouble.

Come on. She thought that was—

she's not here. What does it matter? Right? And then he says— I forgot. I think he said, I didn't tell her or whatever.

Too late.

So, so, 2 points for you, right? Will it teach us? Absolutely. But is the word— where I'm struggling is the word teach. It's— I don't think that's appropriate. I think it's more of will it assist us to have more epiphanies? That's where I think is really interesting. Will it help me to have more what-if moments. Oh wow, I didn't see that. Okay. And the beauty of AI is that you can go, okay, and you could have a conversation with it, and you could talk to ChatGPT on your mobile phone right now. You could go, that's interesting. Now what if you try this and this? And then it'll correlate in a few seconds later I'll give you an answer. And wow, okay, but what about that? That's where we're at. So is that teach— I'm not trying to shoot down the word teaching like he was shooting everybody else down a moment ago. It's more than teaching. It's— I look at it like being back in kindergarten when you were playing in the sand and everything was so much fun. As an adult, we got bills, we got a job, we have, you know, you know, all these things. Isn't it? To me, it's so exciting to have this refreshness of like, oh my God, it's so much fun to be alive again. It's unlimited possibilities. Woo! Yeah, that's what, you know, that's the level that I'm at.

But on the flip side of that, Mr.

Negativity, go for it.

Hey, hey, you said you want to fail. Yeah, right?

Please.

In this type of technology and advancement. I mean, what you're talking about is the scale of data use, the scale of data incorporation and analysis is not— exponential is a soft description of that, right? What we're doing with it now. And it's going to allow us to fail less. We're inherently going to avoid hurdles that our society has always gone through. You said you want to fail. Yeah, right. So as I incorporate this into my life, I'm going to fail less. Yeah, I'm going to be better at what I do.

Yeah.

What's downstream of that?

Well, that is a very spiritual question. It is. If all the answers to the questions we have are answered, And the robots. So I will talk about robots.

Thank you. Thank you. Finally.

I'm trying to take care of you.

Emily brought up robots.

This is me paying you under the table.

Yeah.

Yeah. So, and the robots are coming in 5 months from now exactly, actually. And I'll elaborate that in a moment. So if there's every— if it could help me, what is our purpose then?

Me?

If all my questions are answered and everything's taken care of me, what is my purpose? Who am I? That's why you're asking me a very spiritual question. That's a beautiful question in itself, isn't it?

Yeah.

Right? Will we get there? No, because human beings, we're so crazy, we're causing war. How many wars do we have now? I mean, it's just, we take something beautiful and we destroy it. We're destroying this planet. And then during the cranial people who you don't like, so both of you guys are negative in your own way. The cranial guys, they asked me, you know, AI, you know, it can be used for bad. Bad people will do bad things, good people will do good. Is human beings, we are the floor. The AI is just gonna do what we tell it to do. Yeah, you have your hand up. I can barely see you, but please go right ahead. Does anybody have a microphone for the gentleman?

Or no.

ChatGPT is based upon the information we feed it. Yes or no? What if it is wrong?

In advance.

So it is wrong. So let's, it is wrong by the way, not all of it. It is getting smarter. So AI in itself is about 40, 50 years old. Generative AI was using neural networks and large language models and all that. That's only 16 months old and it's a baby and it's learning and it's getting smarter. What's amazing about the announcement two nights ago with Google Gemini version 1.5, they then, they now said that they are pretty confident that it's correct 88% of the time. When 1.0 was launched, they were at 70%. That's insane. It's getting better extremely fast. Is it based on the information we give it? Yes, but it's correlating. It's like when you guys right now, what you're doing is you're listening, you have your own historical experience, you're listening to me, And you're going, "Ah, okay, that's what the damn algorithm is doing, but we didn't tell it to do that." This is what's freaking us out, and we like it. And one of the scary things is, you know how just because you can, you shouldn't do, right? A lot of you guys in college, just because you can, that was a joke, too. Go be guys.

I got it.

But if you guys have fun in college, you know what I'm talking about. Just because the AI AI can do that, should we allow it to do that. We're approaching that precipice where we have that as a serious question and a dilemma. It's getting that advanced, that fast. It is wonderful because to your point, society issues with sustainability, pollution, the forest, climate change, a lot of that we're going to find new solutions that we never thought of before, and that's fantastic, but it's a slippery slope to other things as well, right? And human beings, we are so strange that we do weird things when we— when the absolute— how many people been in a perfect relationship and you just messed it all up? Raise your hand. Not one person's going to raise their hand.

Look at you.

Yeah.

Yeah.

Okay. Right. Exactly. This is the problem. We could have a perfect planet probably. I'm, you know, I'm this optimist. But human beings, we are the flaw in the whole thing. Even if we get the perfect answer, we get it to 99.99%, somebody's going to do something stupid with it.

Let me have you stay on that point. So like, if we have the best AI and the AI predictive accuracy increases all the time, we get better, more and more data, better and better data, we can get really high predictive accuracy.

Yeah.

At least for the short term before the AI and the machines take over. We're still trying to utilize that AI for the benefit of humans.

Always.

And so humans are limited by our own cognitive capacity, and yet we still have to process all the predictions that the AI can make. So, for example, if you think about the simplest predictions, maybe they're financial predictions. You want to invest. It gives you all these different scenarios and it can predict accurately. Or you're thinking about having a medical treatment. And it can predict very accurately the outcomes, the complications, the costs. How do, how do, how do we work with the human mind to, in the decision capacity, the decision science of it, so that humans are using the AI to make the best decisions? Because we're still limited in our processing capacity. So if we can predict everything, but we still as humans have to decide what to do, do, how do we translate that into decision science? I mean, Medtronic's trying to do it. They can predict everything, but how do the patients digest it to make the best choice for themselves? You understand what I'm asking?

To add on to that real quick before you answer, that's the reason why when you ask who's used ChatGPT, there's not a single person in the room that's like, oh, that's cool, but how do I use it? How can it Where in my day-to-day life—

everything—

does it make it beneficial?

Everything, right?

But that's what Chris is getting to, is, but do I ask my phone?

Yes.

Like, is it the day-to-day clinic evaluation? But yes, that's the barrier that we're stuck with.

Everything.

How?

Pick up your damn phone and talk. It's that easy. I mean, you could down— ChatGPT is free, and you've just put your thumb down where it says record and you start talking. ChatGPT, I forgot to buy my wife roses for Valentine's Day. She's Norwegian and she has a very bad temper. What should I do? Right? Yeah. Use it for everything. Listen, this This technology, generative AI, and the training that is being put into it is being done by— Everybody know, do you guys have any idea the tsunami that's coming at you? My heart has been bleeding listening to the administrators. There've been two wonderful ladies on stage explaining the pain of being of hospitals and it was a razor thin margin, she said before, right? And I'm thinking, my God, in the IT world, we have infinite money. It's insane. And I'm like, "Regis and Margins." So Apple just reported the quarterly earnings and the quarterly, it wasn't even net revenue, its profit for Apple this past quarter was $157 billion. One quarter profit. That's one quarter. Apple—

Now you got Medtronic's attention.

There we go, baby. So Apple's total gross revenue on a yearly basis is equivalent to the GDP of the 34th largest nation state in the world. That's one corporation. Google has almost as much. These are the companies that are throwing insane amount of money into this to make this happen. Those are your partners. I think somebody said in your panel, who should we be partnering with? Stick to your core business and what you do, but partner with the other entities that know sensors, software, algorithms, and create an ecosystem. And that's how you get this thing done.

So AI and predictive models are, are only as good as the data that's input into them.

Yes.

And we have to be very careful in healthcare when we interpret predictive models because it's very subject to inclusion bias. On the front end of the data that's, like, for example, in the study groups, in our research groups, there's certain criteria that we use to enroll. You can't generalize really beyond— you can't generalize predictions beyond the subgroup of patients that met the enrollment criteria. Large language models are based on language, and to some extent that might be limiting. Where do you see the future of these, of generative AI? Do you think it's going to start including multidimensional information? It's going to include like raw visual data? Like how can we overcome some of this bias where it includes everything in an unstructured way?

So the question, how do you overcome bias?

And how you ask your question is how you Right, we ask our question using language, which is itself a bit limiting, but that's how we as human— language is the software of our minds. Even when people who speak different languages probably think slightly differently because they have different software running there in terms of the structure of their thought processes is limited by the language in which they think.

So how else do we— we're human beings, we communicate through language. That's the tool we have. I started laughing because the ultimate answer, of course, is to put a chip inside your brain, and then you don't have to use your mouth. And you're directly connected, right? You have an AI supercomputer chip in your brain. It has wireless or Bluetooth— probably you want to use wireless so that you have maximum bandwidth capability— and you connect right into the neural network of Google's neural net, Facebook's neural net, Apple's neural net, all simultaneously. And as you think, you're connected to the hive mind, that's probably where you're going. But that's technically doable. But do you want to actually have a society of that?

Wait, did we miss that yesterday? Is that what you guys— I'm sorry, is that what you guys talked about yesterday with the cranial guys?

Is that what we talked— we talked a little bit about that.

Yes, he was organizing the implant technology with the brain surgeons yesterday.

That's what you guys started with.

No, but I'm being honest. I mean, you guys know about Elon Musk's Neuralink, right? They make They make big hype about everything. But they've opened it for that. You go to the webpage, put in your— what is it? First name, last name, email address, and you get on the waiting list. It's here. Right? And the cost is to get it down to the price of a LASIK surgery where you're in and out 2 hours, under $1,000, you're chipped. This is reality. Where you're going is someplace really interesting, but as human beings, our form of communication is talking and we have built-in bias. So today with what the way we talk to the software is you prompt it, which is basically talking. I want, I'd like to make an investment in the stock market, or sorry, let's keep it clinical. I want you to look at this X-ray, I want, or this research report, but I want you to put greater emphasis on the, I'm not I can't make stuff up. But put greater emphasis on this, put less emphasis on this. That's how we— you can modify the bias or influence the bias because developers are writing code and they're biased. You have to remember a developer is a human being. And that morning he or she may have— I keep on doing these wife jokes. I'm not mean to. He or she may have gotten in an argument with the dog. The dog may have did something on the carpet. And he or she went to work and they're coding in a very aggressive aggressive way, and that's built into the code. We're all human. Human beings are the flaw. What scares us is that when we tell the AI, write your own code, mature yourself, we don't like that because we don't understand what it's doing. Actually, an interesting story happened was that they took two generative AI systems and they told it to talk to each other and to ask these other questions and then rewrite the source code based on the answer. When they ran it for 2 days, after 2 days they looked at the outcome and what had happened was the 2 AI systems had created their own language. And the developers said, "We don't understand what they were saying to each other. We didn't teach it to create its own language. Where the hell did that come from?" That's where we're at. This is a most super exciting, We are going to be able to solve in a very short time. I'm talking about like 2024, problems and situations we never could think of. So to all of you, start playing. Be playful in your mind. Pick up your damn phone and talk. Everybody know how to talk?

Yes.

It's almost an insulting question, but it's free to do this.

Play.

And the best version, so Google Gemini, if you want the very, very best, is $20 a month. I think you guys are surgeons, you got— everybody got Netflix, you can afford $20 a month, right? So if you want the access to the best generative AI tools today, it's the price of a Netflix subscription.

But, but still, I want to get back to my, my original question, which is, for example, you can predict everything, and in the financial world, Bloomberg made a lot of, you know, it was very successful by making interfaces that traders could use to digest all of those predictions and all that information in real time to make the best decision. Do you think that decision interfaces is going to be an industry that will grow out of the, the ability to predict? So once you can predict a lot of different things accurately, You have to consider your own values and then digest the information. It has to be placed in a digestible form to make the best decision that you can make. Do you see that decision science or user interfaces as an industry that's going to grow in the world of AI? I don't see how it cannot. Otherwise, it can be a bit uninterpretable. You can't digest the information.

Yes. So you, in your example, you have an individual who's digesting the, all the information. And then the way I see it is that individual is only one of a million people that I want to know what is their outcome of their analysis. We could do that today. So it's infinite, almost infinite amount of information. So Warren Buffett, I today, or you today, when you guys are done here, you could use ChatGPT or Google Gemini or Bard or whatever you want. You can say, I want you to analyze this S&P 500 for the last 3 months from a Warren Buffett perspective. What do you see? You could do that today. So what it does is really interesting. It will look at everything, every transaction, trade Warren Buffett has done. It will go back and it'll read every— it has already read every single book every single article, every single blog post Warren Buffett ever written. It will then simulate Warren Buffett's methodology, his attitude, his perspective. It will take all of that and then look with the S&P 500 for the last— since January 1st, and we'll analyze it, and then it will give you a conclusion. We could do that today, and that takes it about 15 seconds. And the thing is, I just gave in that example, that's Warren Buffett. You can say, I want to do Warren Buffett, and then what Who's his partner? Charlie somebody, Mogler. He passed away recently. Munger. Thank you very much. He's an investor. And you could say, you could have a multitude, a multitude of people and you could get all of their perspectives now. And when you get that conclusion, you're not done. Interesting. Now I want you to blah, blah, blah. And you could go back and forth, back and forth. I call it iterations of innovation. What I'm talking about now, a lot of you are going, "What is this guy talking about? Is this possible?" What's the most amazing point is not only is it possible, is that when you get that conclusion, you can ask a question, 'cause remember, we're human beings. We always go, "Hmm, but what if?" And a few seconds later, you have the next question. Iterations of innovation, the level of innovation we're gonna be at, or we are at, this is not future. You got a mobile phone, you have it with you. Where we're at now, we've never been been here before. This is beyond exciting. It is. And when it comes to healthcare, yes, we, you know, right now there's so many famous stories of AI having found protein combinations that we've never seen before, and pharmaceutical companies are going to make billions because, you know, they have new products coming to market. When I was with IBM several years ago, we had a solution where we guaranteed a pharmaceutical company we could take that— usually we took them 10 years from when they started researching or something, when they brought a drug to market, we said we could take you from 10 years to 2. By the time we actually did our second project, we had it 10 years to 2 months. And that was a project from 4 years ago. I guess today we could do it in probably a week. It's just accelerating and accelerating and accelerating and accelerating.

Something I've always wondered is, you know, I remember reading about the first, you know, usages of AI probably 10 years ago and healthcare. And yet today we see a lot of, you know, AI startups, AI companies, AI products that are starting to come through. I mean, I think the FDA is going to see 1,000 applications this year for AI-enabled, you know, medical devices. But today, even in 2024, we don't touch AI that much.

Lawyers.

In our daily work as lawyers, as doctors and surgeons. You think it's the regulatory process?

It's damn lawyers.

By that do you mean the regulatory process, or do you mean the practice, or what do you—

What I specifically mean is that the regulatory process that have been put in place by lawyers to protect individual rights, to protect the clinician, to protect— has gone way too far. GDPR, in that it is handcuffed innovation. Yes, last night I spoke about it, I'll speak again tonight, and that is in Western Europe, European countries as well as North America, way overregulated. You have GDPR, people's privacy rights, data that is created in a nation state cannot leave the nation state. So if data is AI's food, and that means your AI is only good as the data you have, and because of regulation, you can only use the data that your country generates, then which country has the biggest population? Shouldn't they have the best AI? Yes, right? You guys, that's the way it works. Well, in Asia, China has no regulations whatsoever, but all of Asia has no regulations. And China has an agreement with India and they have an agreement with Pakistan. Those 3 countries each have 1.2, 1.4 billion, and the Netherlands a million, I think a billion. They're able to take 3 billion people worth of data. As a person's crossing the street, The AI is watching them. As a person, matter of fact, I didn't talk about this last night, but on lamp posts in certain cities in Asia, they actually have microphones. And they say they use that to triangulate when somebody shoots, they can triangulate where the bullet comes from. But what we actually found out was they're also using it when it's quiet enough, or even when it's noisy, they can filter out traffic. They're listening to you breathe. They're creating data files on your breath. They're able to distinguish who you are based on how you breathe when you cross the street. They can go, "Oh, we know that tempo of breathing, that person is this in the database." That AI that's being built in countries without regulations has infinite amount of food. Infinite amount of data, it is, that AI is gonna be so much smarter than, I live in Norway, I've been living in Norway now for 28 years. I still can't ski, but let's not get into that. But Norway has a population of 5 million people. How smart is our AI gonna be? We're too small. But regulation has said you're not allowed to leave that, the individual data must stay in Norway. Keep everything in Norway. Yes, regulation is in the world of tech and specifically AI, it is killing it. There is a war going on right now. There's an absolute war, AI war going on now. Countries like the United States, all of the Western European countries are all losing and we know it. Again, I'm not crazy. For us, we're going, "This is unfair." The Asian countries don't have this. We have too many damn lawyers that are so proud to legalize and regulate everything. So from— and but it's good, right, for society to protect your privacy. I'm not stupid. But at the same time, from the tech side, the best AI being created today, even though all the hype comes with ChatGPT and all that, the stuff being made by Alibaba, Tencent, and other companies in Asia is phenomenal.

But do you think that, you know, okay, China has the most most people.

And—

but to some extent, you can't— you can't generalize predictive models outside of certain ethnicities and geographies. I mean, not— what the predictive models that might apply in China, for example, for purchasing decisions or even for value— values that are driving certain healthcare decisions might be different from the values in more maybe heterogeneous societies like the United States or Europe, like, do you think you can really generalize all those models out of China and therefore you can call it the best AI? Or is it the best AI for China and other countries, even smaller countries that have 5 million people, like Norway, for example? Yeah. I mean, we know from our own work that it's not just the numbers, it's the depth and complexity of the data that can create better, even predictive models and models that are based on a billion people, for example.

Some of the models that are coming out of Europe from other countries can still be utilized against the data that's in China as well. So it is applicable. Understand this, in healthcare, you know, I don't— I'm not sure how your culture is, whether or not— do all surgeons share all their secrets? I'm not sure, right?

Jeff doesn't share his. But most of us try to publish our secrets and share them amongst our peers.

So in your world, you share by publishing. In our world, we have lunch with each other. So a secret that's known in the AI world— so the Googles, the Facebooks, the Tencents, the Alabamas, and all that— when there are conferences enough, we all are having lunch with each other and we share everything. Because in our world, we're one family. We're doing this together.

Well, that— okay, let me— we have 2 minutes left and there are 2 really important points I want to bring up.

Sure.

To you as an AI expert. So how, when we get a predictive model as surgeons or even as a people, as humans, how do we know that predictive model is accurate? What sort of processes should be in place so we can trust that that model is believable?

Great question. Okay, so I would tell you, this is my own personal belief system, is that I would— I have to be very careful. The model that's been trained by the clinician, or clinicians plural, is the model that should be trusted. You're— the healthcare world has not spent the time The partnership with the Googles of the world, if we keep it to American companies, there is, I have not heard of a deep enough, IBM partnered with Memorial Sloan Kettering for IBM Watson. That was a very good partnership until it fell apart. But that, when that begins to happen, and like was said before, who do we partner with? When the doctors, when you start spending time, when you start training, you put on the headset, When you have, the next time you have surgery, I mean a point, sorry, don't wanna be rude. The next time any of you have surgery, you should have partnered with an IT company. You should have haptic feedback gloves on so that as you are squeezing something, I'm recording that. You should have cameras everywhere so I'm 3D, I could create a 3D model and a 3D model of your actions You should have sensors on the patient. You should be monitoring the patient's brainwaves, is alpha, beta, right? You should be gathering everything. And we start taking that amount of information and putting that in together in software, then I believe that the trust kicks in. What we're seeing though is that because this marriage is not happening yet at a fast enough rate, it is sort of happening, the Googles of the world are gonna do it anyway.

5 years ago, or more, maybe more than 5 years ago, IBM announced this huge partnership that you mentioned where they're going to partner with Sloan Kettering and they're going to solve cancer through AI.

Yeah.

Massive failure because cancer is 100 or 1,000 different diseases, not one. How can we avoid that in spine surgery?

So it wasn't a failure because cancer— I was part of that, so I know the inside story. It was a failure of not managing the relationship between an IT company and a hospital. There's two different entities with two different cultures. They both were excited on day one, but there wasn't somebody sitting in the middle who was able to manage that relationship. And so it went— that went like that. That's actually what happened. It wasn't that cancer was impossible. Cancer is extremely difficult. You guys know more better than I do. How do you manage it? Going back to the very first thing I said, be open, be playful, train the software, gather data. Every ultimate example, how many of you guys think about when you talk to someone, do you watch when they blink? That's data, that's relevant. I was— somebody was talking about Google Tesla before, and they were saying how when a Tesla goes down the road, it's gathering data from the car. No, it's not. It's doing much more than that. Do you know every time if you're driving a Tesla and it hits a pothole, when it hits that pothole, it's not just gathering the GPS data of where the pothole is located, but it's also gathering the data of what is the humidity in the air. It also says if this is a smart city and the city is alive because it's digitalized, what's happening in that city? Was the light red or green? What was the traffic condition? What was— were there cars behind it or in front of it? It's gathering everything. What I want you all to do is be playful. I want you to begin gathering as much data as possible. Do not delete anything. And I want you to start talking into your telephones. It sounds really stupid. Use this. Start with simple little things. I use it for everything. Just simple little things. How do I make spaghetti tonight? I know it sounds stupid. It is so easy to use. You're gonna have so many epiphanies. Wait a minute, I could do this. I could do that. I could ask it this question. I could ask that. My daughter's having trouble with her economics exam. I told her, "Mina, this is what I want you to do. Take the question you're having any trouble with, the text, put it into it and say, 'Explain it to me like I'm in 5th grade.'" She called me back a half hour later and said, "Papa, thank you." She goes, "It didn't insult my intelligence. That's what I needed." That's what I want you to do. I want you to use it and say, talk to me like a 5th grader. I'm not insulting any of you guys. But that's what I want you to do.

Coming from California, I'd say that they're sitting on a huge repository of pothole data, and they're not yet doing anything about it. So I think that concludes our session. Thank you so much. That was tremendously enlightening. We really appreciate it.