On this episode of The Six Five – On The Road, sponsored by Intel, hosts Daniel Newman and Patrick Moorhead welcome ai.io’s Chief Product Officer Jonathan Lee, for a conversation on how AI is revolutionizing sports around the world and how ai.io leverages Intel’s AI solutions to democratize athletic scouting, development and insights for athletes across the globe.
Their discussion covers:
- ai.io AI application overview, poised to revolutionize the sports industry, including a concise explanation of its functionality
- AI’s transformative influence on sports video analysis practices
- The process involved in optimizing ai.io’s software platform for Intel silicon
- Compelling athlete success stories from traditionally underrepresented areas, showcasing opportunities for sports advancement.
- The ongoing exploration of intriguing applications for ai.io’s technology
Be sure to subscribe to The Six Five Webcast, so you never miss an episode.
Watch the video here:
Or listen to the audio here:
Disclaimer: The Six Five webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded, and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors, and we ask that you do not treat us as such.
Patrick Moorhead: Hi, there. The Six Five is back, and we are on the road at Intel Innovation 2023 in San Jose Convention Center. Dan, a lot of energy, a lot of discussion about AI. Shocking, actually, not shocking, because pretty much we’ve been discussing that, researching that for a long time, but really put it into overdrive since this November. Great to see you.
Daniel Newman: Yeah, it’s been a great couple of days here in San Jose. It’s always nice to hear from CEO Gelsinger. Hear and see, if you missed it, pushups on stage, because what other CEO does that?
Patrick Moorhead: No, I know. And by the way, when he does that, I’m pretty sure he’s mocking me because I probably can’t do as many pushups as he can, but I’m working on that, Dan.
Daniel Newman: Well, I’ll tell you what though, this event has been full of not only where are we at with our process, where are we at with our innovation, but it’s also been where are our partners at? This is a developer-centric conference too. What are our developers doing? What are they building? As we talk about AI, it’s not just about AI for the sake of AI, it’s about how Silicon plus developers plus software create applications that change the world? As he was doing those pushups, I said to myself, there could be an app for that. It’s the CEO health tracking app of some type, like there isn’t one…
Patrick Moorhead: Gosh, imagine that.
Daniel Newman: But at the same time, I wonder if any developers are doing cool things with app-based technology that could enable us to better measure maybe an athlete’s performance.
Patrick Moorhead: What a coincidence, we have the Chief Product Officer from ai.io. Jonathan, welcome to the Six Five.
Jonathan Lee: Thanks so much for having me.
Patrick Moorhead: Yeah, first time. I love that Daniel pointed out a belief that I think you have to have, which says, hey, the technology’s great. It’s very difficult. But in the end, it’s about what technology can do for consumers, for business, and society as a whole. Welcome to Six Five.
Jonathan Lee: Thanks so much.
Daniel Newman: It’s really interesting, Jonathan, the event’s keynote focused on Pat Gelsinger, who’s kind of a known… He’s a known athlete. He likes to run the halls at events. He’s known for that, the pushup thing. This isn’t a surprise, but then there was this cool video that showed him running around HQ or some campus doing soccer drills. You know me, soccer guy. And then also we found out it was all part of this plan to bring ai.io on stage. The company’s doing something really interesting.
Having had a daughter that played Division I soccer, went through the recruiting process, I was watching this and I go, oh my gosh, this is really pretty amazing. It’s like matchmaking for athletes and potential teams, clubs. You guys are really taking AI to the level that could change recruiting. It could change athlete development. I mean, talk a little bit about the idea and what you’re doing at ai.io.
Jonathan Lee: Yeah, absolutely. Our mission at ai.io is really to democratize opportunity in sports. That’s an opportunity to be signed by a pro team, but also an opportunity to be trained and developed with the cutting edge technology that these teams have. At the center of that is our aiScout platform. We partner with pro teams, universities, leagues, federations to provide opportunities to anyone in the world to try out for and try to get signed by these clubs and organizations.
The aiScout app is available to anyone in the world, and it consists of a series of drills that have been carefully designed by our sports science team. These drills are able to help us in an automated fashion, obviously using AI, analyze key components of athleticism, things like speed, coordination, agility, but also technical skills as well. For example, for soccer, which is our first sport, it would be dribbling, shooting, passing. Really any drill that a scout might want to see, we can put that into the app.
A player downloads the app. When they use the app, they would get instructions on how to set up the drill. It could be a ball here, a cone here, then how to perform the drill, and also how to film the drill. They would get a friend or a parent to film them, do the drill. The video gets streamed up to the cloud, and that’s where our AI takes over. Our AI models really do an advanced human analysis or human movement analysis starting with skeletal tracking, inverse kinematics, biomechanics.
And from there, we can get a really good comprehensive understanding of how an athlete moves and how they perform. We provide that feedback back to the athlete in the form of a score, as well as feedback in, hey, you did well here. Not so well here. Here’s how you could improve. But then of course, importantly, we take those scores and then we send those to those clubs and organizations with rankings as well. Those clubs, they can take a look and say, “Hey, I’m looking for an attacking midfielder, left-footed, good speed.”
They can compare those and benchmark those against players they’ve already signed. Players are already in their academy or against a need, and they can do that all within our control center and our aiScout app. All in one place, they can see these players coming in in real time and all automated using AI. We’ve already partnered with, well, Premier League teams actually.
Chelsea Football Club, Burnley Football Club. We have a partnership that we just announced with all of Major League Soccer, national teams. Even in a relatively early stage, for us, we’ve had really good traction and success.
Daniel Newman: Congratulations on the success.
Jonathan Lee: Thank you.
Daniel Newman: I remember years ago where video-based sports analysis, at least on the swing of a golf club I’ve seen that, even on quarterback drills and even in equestrian. I’m curious, how is this different from what we saw four or five years ago?
Jonathan Lee: Sure. I mean, AI is really the key to revolutionizing how we analyze athletes through video. If you think about how it’s done even currently where… Say you work with a lot of track and field athletes. We’ll take a track as an example. Say you have a long jumper and you film her jumping. Typically, what’s happening is when she jumps, freeze-frame, and you’ll draw some angles. You say, “Hey, your launch angle is 23 degrees,” and maybe try to shout it out a little bit. That’s fine, except that you’ve lost a lot of information.
You’ve lost the run-up. You lost what happens afterwards. What we then look to do is then add information. So then the coach might put sensors or markers on the athlete, or they might have them do the movement inside a lab, to your golf comment, which is fine. It gives you more information, but then you’re taking the athlete out of their natural training environment, or you’re putting things on that will impede how they move. For us at ai.io, what our goal is is to really take things that you can do in the lab and then move them onto the mobile phone.
Our pipeline, as you go into the cloud, again, we want to extract as much information as we can from the video. Our tech is camera agnostic, really video agnostic. We can run things on historical video as well. It starts with 2D pose estimation, which is commonplace now, although our tech really has been trained on a wide variety of athlete movements, a wide variety of scenes, backgrounds, lighting, all that. But 2D pose estimation, if you’re not familiar with it, obviously using AI computer vision to identify different key points on the body.
There’s 22 key points. And then you form all those key points into a skeleton. The last several years, that’s where it stopped. For us, we use AI to add more information. A third dimension, for example. We take the 2D image and then we infer a third dimension. Now we have a 3D skeleton that we run through our next part of our pipeline, which is called our inverse kinematics. Take that skeletal movement and fit it to how an athlete or human actually moves, because we don’t move like stick figures.
Daniel Newman: Except for me really. I don’t know.
Patrick Moorhead: You haven’t seen me run, have you?
Jonathan Lee: With that information, we can extract over 3,000 bomb mechanics and metrics, really, really analyzing in a way that we haven’t been able to before. You get all this new information, but we also distill it for you. For example, we go back to the app, we take all that information, but it’s a lot to digest. We distill it down to scores, to objective measures. And then when you simplify that process of capturing data, then you’re able to get data from athletes over time, which is really exciting for us.
We’re definitely working in this space as well, which is you can look at things like, how does a player mature? If you see a player at 12, how will that player look at 21? Or maybe even another way to look at it is you can baseline a player and look for things like precursors to injury.
If you are running, if I have say a team in the preseason, I have an athlete run or do a functional movement screen, I can look for changes in things like asymmetries, like lingering or precursors to injuries, or the other way around, which is if I have a baseline for an athlete from a video that I shot then, or just even just historical video, I can say, all right, if they did get hurt, how do I know when they’ve actually returned to their baselines? How do I know when it’s safe for them to return to play?
Daniel Newman: That’s really interesting. There’s a lot there.
Patrick Moorhead: I know.
Daniel Newman: I think you’ve got a fairly substantial opportunity and you also have a challenge to try to take this from one sport to the next because every sport is going to require lots of partnerships, lots of data, but at the same time that complexity is where opportunity lives. Jonathan, we are here at the Intel Innovation event.
Intel is one player in a crowded space right now of companies that are all saying, “We will be the AI champion when it comes to frameworks, software, Silicon.” We’re on stage here though, and you’ve made it very clear that you’re committed to partnering with Intel. I think the world would love to understand why you went that route, why did you choose Intel as a core technology partner for ai.io?
Jonathan Lee: Yeah, yeah, absolutely. We’ve been partnering with Intel since 2021, a couple years now, and it really is because we have AI and compute everywhere, so in the cloud, but we also have our AI labs, our gold standard equipment and testing all in one place. We can talk more about that later as well. When you have compute everywhere, it’s really helpful to know that you have a partner that is in all those places, cloud, edge, and of course, the software and the ways to optimize those models as well.
Originally, our partnership started with a technology that Intel developed called 3D Athlete Tracking, or 3DAT for short. 3DAT was really tested, developed, and honed on some really demanding applications. Number one was the broadcast. It was used in the Tokyo Olympics broadcast, and I’m sure you know the broadcast, one of the demands there is latency. 3DAT was used for the sprinting events, analyzed the sprinters, produce graphics, and clips that were shown as replays afterwards.
That has to happen in 30 seconds or less. All the AI has to happen and the clip needs to be produced, because obviously otherwise the fan will lose interest. That’s the latency piece. But 3DAT was also used for training elite athletes, and there’s demands there as well. If you look at traditional markerless motion capture, pose estimation, it’s usually done on the general human population. But athletes, they move with a precision that obviously…
Daniel Newman: Superhuman.
Jonathan Lee: It’s superhuman. It really is. Having worked with quite a few Olympic level athletes and pro athletes, I mean, it is like fine-tuning a Ferrari, right? You need the tech to match. The accuracy there, it obviously meets our needs as well. The big piece beyond latency and accuracy and usability of that tech is the cost. 3DAT, trained on Intel Gaudi, running on Xeon processors and AWS. It is definitely a more cost-effective solution for us. Our whole mission is to democratize opportunity, and you can’t do that if it’s so expensive that we can’t actually scale.
It can’t be thousands or even hundreds of thousands of athletes. It’s got to be millions of athletes to really reach that goal. Running AI workloads in the cloud is not cheap. For us to be able to do it on Xeon processors and then knowing that as future generations of Xeons come out, we’re able to leverage those improvements in price performance pretty much for free. There’s not a lot of legwork required to port from one generation of Xeon to another.
Patrick Moorhead: Up to this point, we’ve talked kind of an abstract about the technology, what the company does, the benefits it brings. Man, the suggestion of democratizing it, and maybe finding athletes who never would’ve been discovered are discovered through your tools. Do you have any specific examples of where new athletes had been discovered or they got opportunities that they wouldn’t have gotten without you or would’ve had a lower chance, just to be fair?
Jonathan Lee: That’s the exciting part really is finding players where a scout would never go, let’s be honest. One example would be in India, we partner with the Reliance Foundation, and they have their Reliance Foundation Young Champs Academy, which is their academy that is a feeder into their pro soccer league. They approached us during the pandemic when obviously traditional in-person scouting was very difficult. They said, “Can you help us bring in the next cohort, next class into our academy?”
We said, “We can, but the app is in English and it’s not been optimized for the Indian market.” They said, “That’s totally fine.” We did the campaign with them. And that year, 19 players came into the academy, 17 of those players came through the aiScout app. Of the 17, four had never played organized soccer, and one used a phone that was shared by the entire village. Yeah, I know. It’s life-changing, right, because they’re moved out with their families for a five-year residential and educational contract, play soccer, get an education.
Really it’s life changing. And to date, we’ve trialed and signed 103 athletes across our partnership with RFYC, which we’re really proud of, but also with Chelsea Football Club, Burnley, different national teams. Even in call it our MVP state, we’re seeing some really amazing successes.
Patrick Moorhead: That’s amazing.
Daniel Newman: Well, congratulations on those. You heard me say, hey, how do we move beyond soccer, because ai.io sounds like it’s been designed to be a multi-sport, multi-function app, and that’s very exciting. Maybe how we end here is love to just get a what’s the road ahead, what’s the next year, and how do you really make that leap to the next sport.
Jonathan Lee: Yeah, absolutely. You hit it on the head. I mean, the challenges that we’ve seen and are solving in soccer obviously extend beyond soccer. Every sport pretty much has an analogous version of this problem. I can’t say too much today. Maybe next year when we do this again we can talk about the different sports. But if you can imagine a sport, either the team or league has approached us.
We’re excited to move beyond soccer, but we’ve also been approached by developers and companies that are interested in using our human movement analysis for even beyond sports, anything from health to fitness, industrial ergonomics, security, animation, retail, equipment fitting. It’s amazing how many ideas there are. We’ve solved a couple of really challenging problems, some around the human movement side of things, some around the deployment and the scale and the reach.
We love that. We’d love to have developers partner with us to use our tech to underpin the things that they’re doing as well.
Daniel Newman: Well, Jonathan, I want to thank you so much for joining us here on the Six Five here at Intel Innovation 2023 in San Jose. Very interesting. Like I said, for me personally, I’ve been through it. You’re building a technology I do believe can create a lot of efficiency, can help these athletes get identified.
I do think you’re probably going to find applications that you don’t expect because that’s what tends to happen when you build great tech, people figure out how to build on it. Let’s have you back in a year. Let’s talk more about it. Let’s hear more about what you’re all doing and much success in the meantime.
Jonathan Lee: Awesome. Dan, Pat, thank you.
Daniel Newman: Thank you. All right, everyone, hit that subscribe button and join us for all of our Intel Innovation coverage here in San Jose at the 2023 event. Patrick, for you and I though, we got to say goodbye for now. We’ll see you all later.