Dan Greenleaf: Hi everybody and welcome back to the AI Rising Podcast. I'm joined here today with Chris Slee, one of the first folks I've actually got involved with in the AI community, as well as an AI expert in his own right. Chris actually has the distinction of being the first person to answer a question for me using an article that came out the very same day. He is up to speed on all the goings on, and more importantly, he's doing some really cool work at a company called Laina. With that, Chris, you wanna tell us a little bit about yourself and then the company you work with?
Chris Slee: Yeah. Yeah. So the company's been around for a long time, and in fact, we've been in the market for about eight years. And it was Include Health, and Include Health focused on virtual physical therapy. So we built a bunch of machine learning that helps people recover from surgeries and injury — anywhere you would have PT in the loop — being able to actually perform their PT plans in front of a web camera. The webcam tracks their body position, counts reps, grabs range of motion off of whatever they are — whether it's shoulder or knees or whatever — gives that back to a PT. The PT can then adjust their plans, and then the person's outcomes get better and reduce pain. PT is very important when it comes to surgeries and injuries and those kind of things.
Include Health recently has rebranded, and so now we're Laina Health. And Laina Health is, a couple of reasons to do that under the hood — not only did we have all of this machine learning that's there, that does this interactions with the patient, and the software itself will have the patient move away from the camera more and turn to the left and turn to the right and talks to them while they're doing it — but it was always robotic and pretty manual. And so we wrote an underlying mechanism that uses a persona called Laina that has a more conversational approach to being able to help that person actually do their PT.
Then we took that another step further, and now when you're done with your workout, Laina will start asking you questions about your workout and how did it feel, and understands your schedule — when you're gonna talk to your PT next and who your PT is — and just have a conversation with you about your recovery. And all that goes back to your PT. But for you, on the other end, it is very conversational. You're just having a conversation with Laina. And then if you call in and need help getting your link to your plan or reschedule with your PT, Laina actually handles all of that using conversational AI.
So it was a good inflection point for us to switch from Include Health, which was very focused on PT, to Laina Health, which is where we're putting our AI forward a little bit. But our PTs and our clinical team is right there. So we're really saying that AI is assisting our clinical team and helping you recover. And that's an exciting change for us because now it's interleaved through the entire experience with us, rather than just some machine learning that's doing some body tracking kind of things.
Dan Greenleaf: That's really amazing stuff. So I personally needed physical therapy a couple times in my life, and I could attest that out of the gate it was easy for the first few weeks to commit. It was easy to go in and have the sessions and maybe do a few exercises at home. And then it slowly tapered off until it just stopped. And I know that there's more that I need to do, and I know I'm probably in the majority of folks who don't follow their plan, unfortunately. And I do know the benefits of PT — still don't do it.
Chris Slee: As is everyone. In fact, the numbers are staggering that out of people that are prescribed PT, only 13% actually complete PT.
Dan Greenleaf: Wow.
Chris Slee: And so you see this — everyone knows they need it. It's like any other thing that we know, like we need to lose weight, we need to stop drinking as much, we need to drink more water. Even positive things — I should go exercise more. Yeah. We don't. Right? So what we found, and there's been a bunch of clinical studies, is that use of conversational AI, how we're starting to use it, actually drives our engagement and our motivation up, because humans do have this need that if there is some accountability, they continue doing it. So if you know someone's paying attention to you, you tend to do your PT. And not only the fact that we have a PT actually in the loop who's getting all this information back and reaching back out to you, but you have this conversational persona, Laina, interacting with you on a daily basis and will ask you how your day was. And if you miss a couple days, she will ask you like, what's been going on? Is there some reason you haven't done your PT in a couple days?
And so there’s been studies that have been done by Harvard and MIT on human interaction with AI and how that drives engagement. And so we're trying to use that in the physical therapy space to increase engagement. And we know that if people do their PT, they will get better. So that's a great sort of tie-in to that side.
Dan Greenleaf: I know I've read — done a little bit of research on — there's programs where they have a nurse call patients that are at higher risk of not following their plans and having higher impacts. And they get better compliance by just calling once a week and speaking to a patient for 15 minutes. Of course, paying for a nurse or a trained professional is pretty expensive, whereas an AI model isn't quite free, but comparatively it's a lot cheaper. Right? If you get similar results, that'd be amazing. But it's amazing to me — you guys started Include Health what, eight years ago? How old is it?
Chris Slee: Yeah, Include actually started — the brief history of Include is — it started about 12 to 15 years ago designing — we had a 900-pound functional trainer that was geared to go into PT locations. And so we had this piece of physical equipment and then we built out a sensor platform, and then we connected to all kinds of rowers and bikes and all those kind of things. And then we started getting into 3D vision with the Cincinnati Children's Hospital and the Air Force. And then the pandemic happened and we went all in on 2D vision models, and we ended up partnering with Google Research and built a brand new one specifically aimed at the complexities of tracking human motion through physical therapy. And then that's what we're in market with right now. So we've been live with patients since '18. Yeah. And then, but yeah, now we've switched over to running our own clinicians, treating our own patients.
Dan Greenleaf: That's pretty amazing — you said 2018 — so seven years ago, you were using computer vision to analyze if a person was performing their physical therapy correctly and giving them live feedback.
Chris Slee: Yep. So our machine learning is not generative, right? So we don't do — it's not GPT, it's not tensor-based. Ours is — I'll say old-school math machine learning, right? And so yeah, that was the catalyst for — we were ready when generative AI came in. It was just a different type of model that we were using. But yeah, we've been building our own models for almost a decade now.
Dan Greenleaf: Wow. So you guys started from scratch before there were all the models out there and the tools to build up from nothing. Yeah, that's pretty amazing. You don't realize just how far back some of that stuff goes.
Chris Slee: That's funny. People go, oh, it's AI, that's just in the last couple years. It's like, no, that's been here for 20 years. It was just harder.
Dan Greenleaf: Yeah. Now it's just more accessible, right?
Chris Slee: Right.
Dan Greenleaf: Yeah. It used to be when computers first came out, you had to build your own. Now it's like you just go to Walmart or Amazon and get one sent to your house. But yeah, that's amazing. You guys have been doing it, working in the AI space for so long. Tell me a little bit about Laina and just — are you seeing results that you can share already or is this early in the process of introducing a new agent?
Chris Slee: So our catalog of about 7,000 activities that our PTs used to prescribe for care — we used to call it robo voice because what it would do is the instructions for what the patient needed to do and how they needed to do it essentially came across as text, and then whatever browser you were using read it to you. So it kind of came across with this very robotic kind of voice. Being able to inject Laina underneath that changes that whole experience.
Even me, who uses it on a daily basis — both in development and use and production and test and those kind of things — it is a completely different experience. You really do feel like someone is having a conversation with you about your recovery, and you very quickly forget that it is a generative AI because it's very conversational. Especially at the end of a workout: “Hey, good job, Dan. You did what you needed to do today. I know you have a scheduled appointment next Tuesday with Thanos, who’s your PT. Is there anything you wanna tell 'em? How is your pain level today? I know you did your pain scores, but tell me how you felt about that.” And you just start having this conversation and all that, of course, gets recorded and goes back to your PT. So that way, next time when you're on a Zoom call with your PT, they're able to go, “Hey Dan, I know you had some knee pain last week when you were doing your squats. I've adjusted the squat a little bit. I've reduced your range of motion that I want you to get. So it should be a little bit easier.” But that loop — right — between AI and human and your PT on the other side, and the PT actually paying attention to what you're saying through AI, it changes the behavior of our patients.
Dan Greenleaf: I didn't even think about it, but all of my PT has been in person. I'll get take-home exercises, but I've never done a Zoom call with PT and I've never had a robot observe me or a program or a bot or something. Are most of your physical therapists, your PTs, meeting with their patients online now or are they still…
Chris Slee: We are a hundred percent virtual. So if you think of that 13%, or the pass rate — right — everybody else fails PT that gets prescribed PT. One of the reasons is, it's a pain in the butt, right? You have to — if you're going to your PT — you're spending some prep time getting in your car, driving to where that PT is, doing whatever you have to do there, and then driving home. So you're blocking out hours and hours of your day, typically during the day, because PT is normally not running at night when it's convenient. Or there's a proximity problem.
So the VA is one of our clients. We support the rural initiative. So this is accessing PT to veterans who are not within a hundred miles of a PT location. So if you start thinking about just the logistics around in-person PT versus the number of people that need it, versus the reasons why they don't do it...
Plus, you know, when they handed you your folder of papers and said, “Here you go, Dan. Do all these things every day,” what's the likelihood that that is gonna happen is like almost zero, right? Whereas the same PT that you're dealing with to do your assessments and range of motion, and they're like, “Let me see you move your arm, Dan. Let me see your shoulder. Okay. We wanna work on this.” And then here's your plan. And you get it.
Then Laina starts talking to you. And you see your reps being counted and you know that your PT on the other side is gonna watch you do them and know that you did them well. And it just changes that whole dynamic.
Dan Greenleaf: Yeah, that's pretty amazing. I laugh when you talk about the amount of time commitment, 'cause for me, when I was really sick at one point, I would go to PT — which I wanna say was like 45 minutes — and then I was not able to drive home right after. Like, I needed another 45 minutes to recover. So I'd go sit down at a different spot, just wait 45 minutes, then go out to my car. And being able to do that from home and then go walk over to the couch would've been so much nicer. Yeah, that's crazy. I hadn't even thought of that.
So tell me a little bit about just the technology side. Do you have to have a specific camera, or are people just using their phones? Or how does this all work?
Chris Slee: Yeah, it works on your desktop and your tablets and your phones, and it's really anything with a sort of modern browser. So it doesn't really matter whether it's Chrome or Edge or Safari or those kind of things. We know that it doesn't work on an Amazon Fire tablet because that's just like too small.
But anything that can run a modern web browser — it works. And the patients interact with it. We do a lot of performance analytics on your particular device, in your particular browser. And then we tune our models to essentially the inference capability of your particular device.
So when you connect — and if you're on a beefy machine, or you're on a MacBook or a Windows laptop and you've got lots of processing power and great bandwidth — we can capture more information. So I can go from… Because if you think about what body positioning tracking is doing, it's every frame that we are seeing — right — in the video, it's just a picture. In each of those frames, we map your body. And then frame after frame after frame, we know what you're doing through a time series model.
And then knowing that, if you're on a weaker machine, I really only need 15 or 20 frames a second, which is pretty small. Whereas on most devices, I'm getting 120, 130 frames a second out of your video transport. I get lots more data. I can be a lot more accurate. Our outcomes are more accurate. But it really works on very old computers with very old browsers. So this isn't something that you have to have a GPU and an NVIDIA card and stuff to run. This is aimed at someone who is in their house. We've treated patients from 12 to 98.
Dan Greenleaf: Wow.
Chris Slee: Our average is in the sixties — late sixties. So yeah, this is not a technology that we have to have out there for high-performance kind of things.
Dan Greenleaf: That's pretty amazing. But it's funny — in a way, you've got some old-school AI — just eight years old — with the computer vision, and then you're integrating more new age… I don’t know if you quite call it generative with Laina, but—
Chris Slee: Oh, it is. Yeah. Yeah. Because the conversations you have — that's all generative AI under the hood. And what she's saying and what you're saying and her response to that and what she decides to sum up and send back to the clinician — yeah, that is all generative AI.
Dan Greenleaf: That's amazing. Crazy how fast things are changing. With that, any final thoughts you wanted to share with the audience just about Laina?
Chris Slee: I think that if you're ever unlucky enough to have to have PT — now, we don't sell direct to consumer — but you should seek it out. It's a pretty interesting thing. If people wanna talk about it, there's always the Columbus AI Meetup, and I host there. So if you ever wanna have these conversations, we should have them there. It's a fun time.
Dan Greenleaf: Yeah, I hope not to need PT again till I'm in my mid-sixties, but if I do, I know who I'm gonna call and figure out who your providers are.
Chris Slee: It's funny 'cause now people that know what I really do — and they're like, “Hey, I'm having some back surgery.” And the first thing I'll say is, “Do your PT.” I don't care who — like, just — 'cause I have seen people who don't do their PT versus people who do their PT. And we have all the clinical studies that go: if you wait too long or don't do it with enough cadence, you'll never regain range of motion. But if you do what you're supposed to do and follow your PT’s guidance, you will regain almost all — if not all — of your range of motion. So it's really like: future you will thank you for doing your PT today.
Dan Greenleaf: Yeah, I can't speak highly enough about PT. I had sciatica a year or two ago, and I tried a chiropractor. I tried drugs. I tried a massage therapist. I tried all these things. And then I tried PT — and it wasn’t overnight. It’s not like a drug you take and then all of a sudden it’s fixed, or a quick surgery. But definitely I noticed after the first week it started to get better. And as long as I kept doing my PT, it kept getting better until it was gone. And I'm like — this is what I should have started with.
Chris Slee: Yeah. It turns out it's not magic. It's muscles, ligaments, and joints — and the muscular system. And yeah, you make all those work better. It's amazing how much better your life is.
Dan Greenleaf: Amazing. Chris, I appreciate your time and thank you. Thank you for sharing with the rest of the community, and look forward to seeing you around at Columbus AI Meetup, as well as the other meetups around town.
Chris Slee: All right, I’ll talk to you later, Dan.
.png)
.png)