What Impact Will AI Have on Training Software?

AI training software is here to stay. We talk with Dr. Paul Laursen about where it may go and how it may impact training and coaching in both good and potentially bad ways.

FT EP 292 with Dr. Paul Laursen

A lot of fear surrounds the development of Artificial Intelligence Training Software, especially among coaches. For athletes, the concern is regarding how much they can trust a training plan built by a computer algorithm. Coaches, on the other hand, fear that the software will take their jobs.  

Both of these are valid fears. To be frank, coaches who have made a living building templated training plans should be concerned. And since AI training software is here to stay, the question we need to ask is not whether it’s a good idea, but how we’re going to respond to it.  

Joining us on the show is renowned physiologist and owner of Athletica.ai, Dr. Paul Laursen. As a physiologist, Dr. Laursen has been fascinated by the potential of AI training software and has started working with Dr. Stephen Seiler to explore its potential in research. He talks with us about both the exciting and dangerous ways the software could change training. The question we’ll ask upfront is whether this technology has opened Pandora’s box.  

With this technology, there’s the potential for bad advice – or garbage-in, garbage-out – and the danger of losing that all-too-important communication between coach and athlete. It’s not all doom and gloom, though. There’s potential for good, too: the ability to deal with data overload, give novel observations, and adjust training day-to-day for the best results. Finally, we’ll talk about the place for coaches in this new training paradigm.  

Joining Dr. Laursen, we’ll also hear from ex-World Tour cyclists turned gravel riders, Kiel Reijnen and Alex Howes; inventor of the modern bike fit, Dr. Andy Pruitt; owner of Build PT and Performance Center, Larry Meyer; professional cyclist with L39ION, Robin Carpenter; and host of the Sonya Looney Show, of course, Sonya Looney.  

So, log into ChatGPT and ask how we can make you fast.  

RELATED: Fast Talk Episode 291 – Neural Networks: Possibly the Most Important Training Tech You’ve Never Heard of—with Alan Couzens 

RELATED: The Future of Coaching Has Never Looked So Good by Joe Friel 

Episode Transcript

Rob Pickels  00:04

Hello, and welcome to Fast Talk, your source for the science of endurance performance. I’m your host Rob Pickels here with Coach Connor.

Rob Pickels  00:13

Both athletes and coaches have questions surrounding the development of AI training software. Athletes want to know if they can trust the recommendations and coaches want to know if they’re going to have a job. AI training software is no longer a hypothetical, it’s here to stay.

Rob Pickels  00:27

The question we need to ask is not whether it’s a good idea, but how we’re going to respond to it. Joining us on the show is renowned physiologist and owner of Athletica.ai Dr. Paul Laursen, as a physiologist Dr. Laursen has been fascinated by the potential of AI training software, and has even started working with Dr. Stephen Seiler to explore its potentials and research. He’ll talk with us today about both the exciting and dangerous ways the software could change training.

Rob Pickels  00:56

We’ll ask upfront whether we have opened a Pandora’s box and we’ll also talk directly about a coach’s place in a world with AI training software. Joining Dr. Laursen we’ll also hear from professional cyclists Keil Reijnen, Alex Howes, and Robin Carpenter. We also hear from renowned bike fitters Dr. Andy Pruitt and Dr. Larry Meyer. Professional cyclist coach and podcaster Sonya Looney also gives us valuable input. So it’s time to log into chat GPT and ask it how we can make you faster.

Trevor Connor  01:28

For nearly two years fast doc Laboratories has brought you the craft of coaching with Joe Friel, the ultimate resource to become a better, more successful and happier coach. We’ve bundled some of the most popular pieces of content from all 14 craft to coaching modules to reshare and what we’re calling the craft of coaching with Joe Friel coaches picks, which includes the star power panel of featured experts like Dr. Stacey Sims, Dr. Andy Kirkland, Jim Miller, Victoria Brumfield and Jim Ruppert, this incredible library will provide a lasting legacy and guiding life for endurance coaches for many years to come. Check out the craft of coaching with Joe Farrell coach’s choice at fast talk. labs.com. Well, Dr. Ralston, welcome back to the show. Pleasure to have you. I think my levels are high because Rob’s had to shop back when I said that.

Rob Pickels  02:15

We’re gonna turn down the bass here, Skid Row. And

Trevor Connor  02:18

so how are you doing on this fine, fall day? And how’s weather up in British Columbia?

Dr Paul Laursen  02:23

I’m doing well. The weather is it’s Kanan down with rain, seven degrees, six, seven degrees and rain. You know how that is for writing. So it’s as a triathlete. I’m moving indoors a little bit. So doing a bit more, more swimming and sauna, which I’m I’m enjoying. So yeah, that’s I’m doing pretty good. Thanks, man.

Trevor Connor  02:43

I remember that weather in British Columbia really? Well, you got there beginning of October, and then it was rain and cold. That’s six to 10 degrees.

Rob Pickels  02:53

Fortunately, we live in Colorado now. And that means we have to remember that weather because we never get weather. Yeah, I actually think it’s gonna hit at 79. But it’s snowing next week. So we have that going for us. Holy cow. That’s incredible. We like our drastic changes.

Dr Paul Laursen  03:10

It just changes in the instant.

Trevor Connor  03:12

I talked to people all over like, oh, you wouldn’t believe how much the weather changes here. And I’m like, come to Colorado.

Rob Pickels  03:17

I know everybody says that. I do kind of remember one time where I think it was close to 80 during the day, and then this cold front moved in and you could just feel the temperature drop out from underneath you. And it was like snowing that night. So we

Trevor Connor  03:30

hit last year in Denver, the worldwide record for the biggest temperature drop it went. So in Fahrenheit, it went from 80 degrees Fahrenheit to negative 17 in a day.

Rob Pickels  03:43

That’s right. That’s amazing. So we got that going for us great place to live. Let me tell you.

Dr Paul Laursen  03:48

I love where you guys live. Like honestly, like there’s no better place to go into a training camp, honestly. So it’s a lot of fun. Yeah,

Rob Pickels  03:56

but we’re not talking about weather today, gentlemen, maybe we should do an episode on weather. But today, let’s talk about something a little more futuristic.

Rob Pickels  04:04

All right, well, so Dr. Larson, you are here to talk with us about something you’ve got very involved in, which is artificial intelligence training software, AI software. And I’m going to start just kind of setting some context here. We don’t want to have that standard conversation of this this kind of hypothetical, what could it do all that sort of stuff. I really want to just just start out by saying it is here. There are a lot of companies now that are developing AI training software. It’s here to stay and I know there’s a lot of fear associated with this. We hear this from coaches. But the thing that we’re going to bring up is, it is here this isn’t a question of can we avoid this. This is much more a question of how are you going to respond to this so this is like the back when it was horse and buggies and the car was invented. Cars were here to stay you had to respond, you had to figure out how you are going to adapt, or you’re going to get left behind. And I think that’s what we’re seeing with with AI training software. But Dr. Larson, what’s your thoughts on this?

Dr Paul Laursen  05:10

Oh, my, my thoughts are exactly that, too. So a little bit my history is, is I was working with high performance sport in New Zealand. And my job was really to kind of make a difference in the program in in the New Zealand Olympic program, and one of the things was putting monitoring systems in place. And again, for context, you have to remember that this was when training peaks was like, you know, just kind of emerging more on the scene, and not all coaches wanted to use training peaks, right, they would much rather use pen and paper, the coach, you’re talking about, you know, started before it was just everything was started in the head and, and go with this, and now we’re, we’re coming in with his monitoring system, you know, putting sensors in place. And that was a, that was a really big shift, to actually get your head around doing that. And it wasn’t easy, there was a lot of resistance. And I think we’re sort of in that same sort of level that I kind of reflect on there, when when I was trying to put training peaks and and monitoring systems like like Garmin and polar and stuff on on athletes and getting coaches to buy in to the value of it. So I think Yeah, same same sort of time, as well. But just like back then, I mean, look, look where we are now, like, it’s not a big deal to think about monitoring yourself using training peaks, and all the intelligent tools that are on that that program, you know, will put training peaks out there is the gold standard, right. So even before the software was here, we started with, really the book, the science and application of high intensity interval training, you know, and really looking at the bulk of research in terms of training science, that we believed in, we wrote that that book with my colleague, Martin keshite. And then again, the story is that secondly, we wanted to teach that in a course. But then thirdly, we said we again, at the same time that the AI and intelligence was was kind of emerging and just becoming to me very evident that that we will get to today eventually, and that it would become a tool to use alongside that. We say we have to have an app or a software pieces. Well, that’s harnessing the principles that we believe in. Ultimately, it comes down to the the answer the same question that we that we try to solve within the book, and course, what session do I do today? And that’s what we try to do with the technology as well, we make sure we’re trying to answer that, that key human question or coach question, what training session do I do today, and that’s kind of where it evolved. But to your point, Robin, it started with the human first and the fundamental principles of science that we learn through the research.

Rob Pickels  07:53

Before we dive deeper into this topic, let’s hear from two ex pros, Alex house and kill Rajan with some of their reservations that are fairly common among pros and athletes.

Alex Howes  08:04

Honestly, I think that some of this new AI stuff, I mean, it has a lot of potential. And personally, it was something we were we were pushing a couple of companies to do a number of years ago that I mean, it seems easy, right? To look at the weather and either you know, you put in x workout when it’s rainy, and you put in why workout when it’s not, or oh, it’s super hot. So down regulate the power of numbers by 10%. Like, it’s a lot of so seems simple. And things are just finally catching up. So I mean, I think there’s a there’s a place for it. Do I think it replaces like personalized coaching? For some people for most probably not. There’s a lot of subtleties to cycling. Like, if you’re just trying to get an extra five watts and your threshold, AI is probably gonna get you there. If you’re trying to get a line across the finish line first, like

Kiel Reijnen  08:58

good luck. I didn’t think about it like, you know, CTL CTL a useful, valuable number that we like to reference a lot. Sure, but it doesn’t account for sleep deprivation from kids international travel, home stress, it doesn’t account for you getting sick, it doesn’t account for you know, like there’s all these things that it doesn’t understand that a coach does. So I think Alex is right, like it can be a really effective tool for the the minutiae, you know, like the really small details like, Oh, I’m going up to 9000 feet tomorrow, you know, let’s change my numbers by X percent. They can just do that without someone having them go in and make the calculations that’s really nifty, but it doesn’t understand the stress on your body when you sleep up there. So you need both things.

Rob Pickels  09:44

That was something I was gonna bring up at the start of this. We’ve now had about six or seven companies that are developing AI training software. Reach out to us to come on the show. And Robin I actually avoided this for a while because we still don’t know where we stand on the software. We’re still on clear on where it is going. And wanted to have a better understanding of that before we did an episode on it. The reason we chose you is for exactly those reasons. And so our audience knows you have developed a software package. It’s athletica.ai. But you are a researcher, you have been involved in the science of exercise physiology for years, you bet at the forefront of a lot of the key debates. And we know that you are taking this from an athlete focus perspective, you’re taking this from a scientist perspective, as a matter of fact, I think we’re going to talk about this a little later. You’ve been using your your software with Dr. Siler to do some actual research.

Dr Paul Laursen  10:39

That’s right. Yeah. So I’m really fortunate that, again, your your colleague, and he’s on crosstalk all the time, Dr. Sylar, he is actually leading the charge with a conglomerate of universities and tech commercial partners, alongside a large grant application in Europe, to really build PhD scholarships that take the emerging tech, the emerging sensors, and make understanding of those, continue to build research around it and to continue to prepare tomorrow sports scientists and coach to you know, because this, this world is in front of us, and we don’t want students and coaches to be come redundant through this. We want them understanding, contributing to and building their own ability to serve as athletes because again, I believe I don’t I don’t think that things should change at the coalface. But our service can become better. So yes, Stephens, on to your point, Stevens on board with this, you know, again, just kind of like me, he sees he sees where it’s going, we’d rather be in it, then then outside of it, I think.

Rob Pickels  11:56

Good. So let’s dive into this. And really, what we’re going to try to do through this podcast is get a two questions. One is, how is this going to reshape training and important differentiation there? As I said, this isn’t will at reshape training, I think we’re going into this with the belief that it is going to reshape, so how’s it going to reshape? And then the second thing we’re going to talk about is what are the positives? And then what are the dangers? And what are the fears? But let’s start with I think a very important question, which is, let’s define what artificial intelligence training software is. How would you define that?

Dr Paul Laursen  12:33

Yeah, I had to research this question. I and I had a good chat as well with my, my head, AI guru, Andres ignobly. Athletic Athletica, and basically, at the I checked out with GPT. Actually, so first of all, let’s define intelligence first. So when we just look to intelligence, it’s the ability to defined as the ability to acquire, to understand to apply and adapt knowledge to solve problems, to reason to learn from experience, to exhibit creativity, and adaptability in various situations. So it’s pretty, pretty long winded, but this is, you know, think about human intelligence first, that’s its definition. And then artificial intelligence is then the development of computer systems that can perform tasks that typically require that human intelligence that I just described, that we all possess. So things like problem solving, things like learning things like understanding natural language, recognizing patterns, making decisions, and then adapting to new situations. So it’s leveraging those human capacities at the computer sort of level. And then there’s different branches of AI like machine learning and neural networks, but we won’t get into that. But that’s the definition so we can kind of maybe start there. So I

Rob Pickels  14:01

have to admit when you said you asked chat JpT to define AI intelligence offering like, this is where we find out if chat GPT as a person. Yeah. Well, it’s good looking. Everybody loves it, like okay, it’s got an ego.

Rob Pickels  14:16

I love the irony that you went to chat GPT to define artificial intelligence or intelligence, because also as we were prepping for this episode, I went to chat GPT and asked it this is this is the input, right a cyclocross workout with the influence from Paul Larsen and it spit back. Paul Larsen is renowned for his contributions to high intensity interval training and it goes on. So I want to I want to know from you, how do you feel about this workout? Start with 20 minutes of easy pedaling to increase your core temperature and prepare your muscles incorporate a few short bursts of speed for five to 10 seconds to activate your fast twitch muscle fibers. Main set micro intervals, three sets of eight by 20 second full gas efforts with 10 seconds rest, rest for four minutes between sets. And then there’s some running dismounts, you know, workout it has and then 15 to 20 minutes of easy spinning. Is that a Paul Larson did chat GPT crack the Paul Larsen code

Dr Paul Laursen  15:16

and said, That’s 100% that’s an athletic workout. There you go. So you got it. He just needs Chad GBT,

Rob Pickels  15:24

you know, but I do think that it’s interesting, right? Because ultimately, this question that Trevor asked is, What is a i, and, you know, we talked about something like chat GPT. And people in the know, love to say that, hey, this is a large language model, it knows absolutely nothing about doing workouts, right. And so that’s different from what athletic is doing, because there is an understanding of the athlete and, and the changes that are occurring, the adaptations, you know, chat. GPT is a large language model, which means all it’s doing is putting combinations of words together, that makes sense based on everything that is learned and been trained on in the past, right. But I think that when we’re talking about these training, softwares, we are doing something a heck of a lot more sophisticated than what chat GPT is able to do. Because GPT doesn’t really know if this is a workout or not it read it somewhere and associated it with Paul Larsen and was able to kind of pull it out of its memory bank and spit it back out to me, not necessarily reasoning, or making adaptations or changes

Dr Paul Laursen  16:25

will totally and it doesn’t know actually what you go and do at the end of the day, right? Like, even with that description that you described, there’s, there’s 100 different ways you could actually do that. So you know, the next step is to actually, you know, put a sensor of some sort on you to actually see what you what you actually go and do. But it’s a good it’s certainly can form a good starting place. And we can start to start to begin to see the advantages of this. And it’s why both of us to prepare for this. This interview. It’s very interesting that we both have are already leveraging this artificial intelligence.

Rob Pickels  16:56

I did not go to jet GPD. And now I’m worried that if I blog and build it road cycling training plan based on Trevor Connors just gonna answer, don’t bother. Just don’t. I do want to throw in one important thing here. And we did an episode with Alan cousins, I believe was episode 256. We’ll put in the show notes, where we talked about this, but it’s really important to understand there’s a lot of people out there saying, Oh, we’ve got this AI software, that to me doesn’t really meet the true definition of AI software, where meaning they’ve created a software package where they just put a whole lot of information into it a whole lot of if and statements. And it looks like it’s coming up with very novel stuff. But really, it’s only capable of an output, using what has already been given what has already been told, I’m sorry, that was a horrible way to explain it. To me AI software has to be able to think it has to be able to do novel thinking where it will take a whole bunch of data, but it is able to draw its own conclusions is able to give you information that wasn’t something that somebody already put into it. Does that make sense?

Dr Paul Laursen  18:04

Yeah, well, so garbage in garbage out, right? Gago is the term that everyone uses. And it’s absolutely right. And again, back to Alan cousins, who I highly rate and we’ve we’ve chatted offline about this area, and we feel that I shouldn’t speak for him. But you know, it really needs to be led by domain experts. And isn’t necessarily this is why I think both Alan and I have kind of gone into it a lot more is because you can’t just throw this AI at anything without some, you know, fundamental domain expertise and backing. And again, that’s why we went through the process of building a book and, and course and getting coaching experience first from our end. And I daresay whatever Allen is building, I would probably rate that as being effective as well, because there’d be that foundation of coaching and science that would be behind it

Rob Pickels  18:56

and what he’s doing and maybe there’s a better way to explain what I didn’t explain. Well, there’s he’s using neural networks. So the difference here, the example that it’s given in a lot of the places I’ve read about this is trying to get a computer to recognize photos of cats. So what I was getting at with some of this AI software that isn’t true AI software is they just put in hundreds of 1000s of pictures of cats. And if you then you show this software, another picture of a cat. And that picture is already in the software. It’s gonna say yes, that’s a cat. But it’s not actually thinking it just has a lot of data in it. Where there’s neural networks, what you actually do is feed in a lot of pictures, you actually don’t tell it what is a cat. And then it has to learn for itself and start figuring out which pictures are a cat which pictures aren’t and it does that by making a lot of mistakes. It then gets corrected and it keeps learning and figures out for itself how to identify what is a cat and that’s novel thinking

Rob Pickels  19:52

an animal that wants to murder you in your sleep. The definition of a cat lover

Rob Pickels  19:59

so Let me get to the big question that I think a lot of people have, I’m going to start with, I was actually listening to a show this weekend where they were talking about AI software as this kind of Pandora’s box, and pointed out a lot of the CEOs of these big AI software companies signed a statement, saying that there is a real danger to AI software, we have to be really careful. And when they were asked, well, if you think this is a danger, if you’re worried about this, why are you developing it? And they all gave the same answer, which is, well, if we don’t do it, somebody else will. And their belief was maybe we’ll do it a little more ethically than those other people would. But there’s kind of this recognition that we’ve opened a Pandora’s box, and I get with things like chat GPT, that that’s a that’s a real question. But in AI training software, have we likewise opened a Pandora’s box here?

Dr Paul Laursen  20:57

It’s certainly possible, the big elephant in the room is really training peaks, because they are the, you know, the leaders in our field, that how’s the majority, I would think of the data or maybe even Garmin, you know, one of these two companies, because they could, you know, I’m just started thinking back to, you know, what you were sort of describing with the neural networks, right, you need large, big data sets. So that’s not the way we’ve gone about things that Athletica, and I don’t think Alan really could go that way, either. Because we don’t have the massive, you know, data sizes, that those, those are required. We’ve taken more the domain expert approach, where this is what, you know, this is what best practices, this is, what coaches would do, in this context, these sorts of things. And I think we know, yeah, I mean, we can talk about this later. But there there is kind of dangers as well, I think we’ve we’ve certainly heard about these sorts of things, you know, from vote rigging, to, you know, creepy, kind of, you know, sales tactics, you know, when your speaker or whatnot, is, is potentially listening to you and all of a sudden see an ad that you’ve been, you know, sort of seen so slowly. Yeah. I mean, I don’t know where it’s kind of all going with that. But that’s, that’s not what we’re doing. As a kind of mentioned to you before, we’re trying to answer our key why, right. And that’s always been like, try to make a tool to help athletes and coaches answer that key question, what’s the best training session that I do? Now? That’s what excites me. So but but yeah, it’s it’s, it’s interesting. It’s, it’s a crazy, crazy kind of area, isn’t it? Yeah. I

Rob Pickels  22:39

think that this, you know, Pandora’s Box question. It opens up a lot of questions about, ultimately, what could go wrong, but the one I want to focus on, I think, is the biggest question, and that is the appropriateness of the recommendations that’s given. And what’s interesting to me is we constantly on this show, talk about individual responses, individual variation, you can give two people the same workout, and they’re going to adapt, perform, recover everything differently. But Dr. Larson, I think that what you’re saying right now, and I totally agree, is when we discuss artificial intelligence, in the training, there have a very large data set of gazillions of individuals right now, it feels like we’re talking about a group mean and group changes to work out. For me, I wonder, how do we rectify? And is there a mismatch between working with an individual like a coach can do versus, you know, artificial intelligence having to be trained by this massive amount of data? And does that help the individual ultimately?

Dr Paul Laursen  23:46

Yeah, again, we get yet to sort of see, I don’t see it certainly isn’t working today. And I don’t think anyone would use that if you’re getting all of a sudden a recommendation, it’s based on big data. Maybe it’s, you know, maybe it’ll go through sort of a beta phase when some of these larger companies we spoke about. But that issue that you mentioned, with load and load response is so, so vital. That’s really what we’ve been working on for ages. And our our training logic we use like the banister, fitness and fatigue modeling profile that you’ll see on your own training peaks. If you go to the performance management, like that’s basically if you want to look at, you know, how it’s kind of operating, we have a version of that that’s working. And then we’re looking at the individual response to load thing as well. So maybe when the big data kind of context, you could, you know, imagine if you had 1000s of individuals, and you met sort of an individual profile that matched one of those profiles, maybe it would pull its recommendation from based on its own machine learning, neural network kind of process. But you know, we’re not we’re not there. We’re not doing that. It’s that’s all possible, I guess.

Rob Pickels  25:00

It’s funny, you know, I think that you take Garmin and whoop, and all of these other companies right now with with wearable technology, and they all are giving recommendations. And I’m not necessarily saying that it’s exactly what we’re talking about today, because I don’t think that they’re driven by AI. But I will say, inevitably, it always recommends that I do the opposite of whatever I’m planning on doing, which either means it’s terrible at its job, or I’m terrible at my job, I’m not quite sure which right now, but I’m banking on myself, I’m not going to lie.

Rob Pickels  25:30

And as you know, you should never coach yourself, well, maybe not figure out how to do what you want to do, as opposed to what you should do. So I was going to ask two questions. And you’ve already answered one. And that one was, where do you think it’s going? And your answer was? I don’t know. But let me ask you a slightly different question. Think five years down the road, AI training software is now much more established? Where would you like to see it go? Ideally, what’s the role that would play? How would we be using it? Well, the

Dr Paul Laursen  26:00

data is really clear that the research is really clear is that we work best when we’re working alongside it. So that’s going to be the secret sauce is that is going to be coaches that, you know, know how to use the tools that are that are put in front of them, right. And that’s why we have a coaching platform as well as an athlete platform we have, you know, because we know that that’s, that’s critical, right? So it’s not hard to find the research that shows that athletes still want the human touch, of course, there are athletes that will be you know, will be self coached. But the majority still want that human touch. And they want the coach to know that they can leverage those tools. So that’s, I think that’s where I want to see things as I still want. I just want more coaches to adopt alongside these these tools that are being being put out there seeing the value of it saying, Oh, wow, I didn’t know I could, it could analyze that for me. And that really, that changes the game in terms of the prescription that I’m going to make today for sure. And yeah, that’s where I hope it kind of goes, I think, you know, you guys did an excellent post on your craft for coaching. Joel Friel just sort of said said is not as much, right. So, you know, training peaks founder, and you know, he sees it. So you just kind of got to move with the times really at the end at the end of the day. And I hope that, and I know that I think I think that’s probably eventually where we where we will go coaches will come to the party eventually there. It’s a little slow right now. And there’s going to be natural resistance. But that’s where they have every new innovation that happens and comes on the scene. Let’s hear

Rob Pickels  27:41

from Dr. Andy Pruitt and Larry Myers, who expressed some of the resistance but also show why it may be important for coaches to be involved.

Dr. Andy Pruitt  27:50

For me, it’s a bad thing that AI doesn’t know me, although they swear they can get to know me, I want to be able to talk to a guy or gal that I can relate to, and who feels my pain on a regular basis.

Larry Meyer  28:02

So I grew up in a small town on a farm and I listened to my dad, Bs with other farmers all day long. And I was so bored to tears. But I didn’t realize that I was learning a ton, which was I was learning how to communicate with people. And I always wanted to be an expert. And I think I’m an expert. But the reality is just like Andy, it’s your ability to communicate with your clients. And so forever, I will say that, you know, this technology is it’s okay. But you always, always, always need to have someone that you’re engaging with who knows you and can turn you around if they need to.

Dr. Andy Pruitt  28:41

Human Interaction is so important. There’s proof now that personal relationships are more important as we age than diet and exercise. Personal relationships. Friendships are more important than diet and exercise as we age. So, man, if I had to rely on Rob’s computer to make me feel warm and fuzzy, I’m in trouble.

Larry Meyer  29:04

You know, in our our community, the cycling community, which is huge, right, I’m not a racer, but I’m in the cycling community. I know a ton of people in the cycling committee and I love it. It’s so important for us we have tragedies happen and we have amazing wins and amazing triumphs and in our in our community, we need that and that’s not something that AI can do for us.

Rob Pickels  29:28

Listeners, this is a great time of year to expand your trading knowledge join fast talk laboratories now for the best knowledge base of training signs on topics like polarized training, intervals, data analysis, sports, nutrition, physiology and more. Join fast talk labs today and push your thinking and your training to all new heights. See more at fast talk labs.com/join.

Rob Pickels  29:55

So let me hit you with another question because this is one I’ve been asked a couple times now including on that At that module for craft a coaching, it was my opinion of how is this going to impact training? I’m going to give you the answer that I’ve been given. And then I really want to hear your response and see where we go with this conversation. But the the answer I give people is to say, look at Zwift Zwift has become very, very popular. Zwift has racing, and it’s now become its own legitimate form of racing, though, like there are people that just raced on Zwift there’s a lot of prize money for some of these whiffed races, it is a legitimate form of racing. But here’s the thing that I find really interesting about Zwift, as much as they have tried to mimic actual racing, and they’ve gotten really close, I’m actually pretty impressed at what they’ve done. It’s not the same, drafting is not quite the same. The big difference is you can get boxed in out in the road on Zwift. You just ride through the person Yeah, you do, you can’t box somebody in. And these little differences, you ask anybody who does a lot as wept rice, and they will tell you, it leads to a need for very different assets. And the strategy is very different. So you don’t train the same to be a good Zwift racer, as you would train the same to be a good road cyclists out on the actual road. And this isn’t a statement of which is better, which is worse, it’s just different. And the response I get people on, they asked me about what is the impact of AIS training software is it is going to come up with different training plans, it’s not going to be exactly the same as what a human would do. I’m not going to make a statement of whether that’s better or worse. But I think just like Zwift racing is different from out on the road racing. I do think training plans and how you train using AI software is going to shift and it might be very small differences that lead to big differences in your form. Or it might be big differences. But I think it’s going to fundamentally change how we train. What’s your thoughts on that, Dr. Larson?

Dr Paul Laursen  32:05

Well, I have to admit I’m not a I’m not a big swifter, I’m going to try and get into it a little bit more this winter. But my personal preference where I’m based, is just to be outdoors, because it’s beautiful here, but I mean, ultimately, the end of the day, it’s we’re trying to increase our capacity to output power. Right, and across the whole power profile. So that’s our aim. It’s one of the tools on Athletica is, here you are, here’s what you your, your fingerprint looks like. And now Can we can we build it and make it better. So I don’t know, I don’t see that being that much different in his with context as well, there must be still at the capacity of output power, and must be fundamental to success in that domain. But I again, I don’t have the experience. So

Rob Pickels  32:52

let me use an example using your software. Something that I think is pretty novel for most athletes is, you know, the I tried this myself, you your software will pump out a training plan you so you give it your your goals. So let’s say my race, target race is 12 weeks from now, it’ll build a plan that and I can look at the entire 12 weeks leading up to that race. But then you have this wizard, where on a particular day, I can go in and go you know what, I’m fatigued today, and I tell it, I’m fatigued. And then it pumps out a different workout. Or I can say I’m short on time, and say I only have 45 minutes and will pump out a different workout. It has this ability each day to change the workout that you’re going to do. And that’s not you know, I’m sure some people work with coaches where they can call the coach every day and say I don’t like the workout you gave me Give me a different workout and probably eventually the coach is going to fire them. But generally, the way people are trained is if they have a training plan. You have this on Tuesday of this on Wednesday, and you just do it. You’re now given this ability to say I actually want to pick my dinner a little more tonight.

Dr Paul Laursen  33:59

Well, that’s right, then that’s because again, fundamental is that context over content, always. So we all have lives and which are crazy and busy and forever changing. You know, you’re a little bit under the weather right now, Trevor, so you wouldn’t be doing a you know, a standard, a standard workout right now. Right? You’d need to have something else in front of you to really make you feel better and and get you on the on the road to back to where you want to go. Right. Yeah, I

Rob Pickels  34:29

went out for a run last night and I’m using the term run in quotes because I was doing 14 minute miles. Exactly. I think kids were walking past me.

Dr Paul Laursen  34:38

Exactly. So we recognize that very early on. You know, this has been a slow evolution, again, been building this machine since 2015. And that was just really apparent that people wanted wanted alternates. And that’s really where the workout wizard kind of comes into play. And again, it’s a cool machine learning tool because it’s not 1000 different options. It’s just like, it’s one or two for any given context, right? Because I don’t think people, if you give people 1000 different options. It’s like, Well, which one do I choose? Right? So we give like a bit of a helper along the way, in terms of which, which one you probably want to pick. So yeah, it’s a, it’s a cool innovation that people are pretty, pretty stoked on. And again, it’s available for coaches and athletes.

Rob Pickels  35:24

But let me just throw a hypothetical idea. And this gets back to this is going to change how we train and there might be slight, but it could have big impacts. So I’m not gonna say I fully agree with what I’m about to state, I’m just going to make this statement, I want to hear your response. There is an argument that if you build out a training plan for somebody, and they get to Tuesday, they have a heart interval workout there, and they just go out and say, I have to do it, I don’t feel like doing it. That that builds a certain toughness in athletes, and there’s certainly coaches out there that are very big on God and do the workout, stop whining, and it’s going to make you more resilient, it’s going to make you tougher, because you can’t show up to a race and say, I don’t really want to do those hill climbs today, it’s changed a race. So I could make the argument that well, allowing that customization day to day, could actually I agree at better physiological effects. Is there a mental component that you lose, or we teach in athletes, they do it you feel everyday, if you don’t feel like going hard, don’t go hard, and you lose some of that resilience?

Dr Paul Laursen  36:25

Well, maybe it’s philosophical, but I mean, for me, I, I really believe in my own coaching, I work on trying to bring feel to the table for for the athlete, I want to teach them to develop that themselves. I want them to know that they shouldn’t do that 2010 workout that that Rob found on chat GPT, they shouldn’t do that workout. In Trevor’s case, yesterday, when he when he you know, there’s no way you were gonna get dive into that and do it, do it well. And if you did, you’d been yourself even further in terms of healing yourself and getting yourself back. So I’ve always been not about the no pain, no gain, I’m more about the teaching an athlete in terms of how they can develop that feel for Yeah, knowing knowing the right session at the right time.

Rob Pickels  37:18

Picking workout sessions isn’t the only place where the software may need a coach to help. Let’s hear from Sonya Looney talking about some of the other nuances where human is still needed.

Sonya Looney  37:29

I think it’s beneficial in that it can be a sidecar approach to bring a lot of information concisely to somebody. But relying solely on AI without having a human component to it might be a problem. Because there’s a lot of nuance involved whenever it comes to mental health. And a lot of it needs to be talked about more by the individual. And yeah, I think it’s just there’s a lot of nuance there that AI might miss. But I’m not an expert in AI. So I don’t really know, but it’s still very young. And I think that there’s a lot of gaps that would need to be filled by a human, at least right now.

Rob Pickels  38:06

Trevor, I think that this brings up a concept that we have talked about throughout the multiple episodes that we’ve done on AI, and that is this technology is best used in conjunction with a coach, right? And that coach is able to assess the athlete and say, hey, you know what, you really do need to back down and this AI software is going to help me figure out what the best workout is. Or the coach can say, No, you’re just being a little bit soft. Like there’s just allergy out who knows what it is, but they’re making that decision. And I think that it’s really important that we always consider this as this is a tool. It’s not the only tool, my toolbox is filled with hammers and wrenches and everything else, and we can’t just use a hammer on everything. And you know, again, I do know that some athletes will utilize this themselves. And maybe they fall into that trap. Because while they’re just you like you said, you can’t really coach yourself, you shouldn’t coach yourself because you’re going to make bad decisions. But a coach that’s using this is able to use it as the tool that ultimately helps him coach his athlete better him or her their athlete better. Look, I was

Rob Pickels  39:15

presenting that as a hypothetical, I will tell you I’m more on the Adjust every day. I mean, I still remember back when I was in school. In one of our textbooks we read about the East German teams or their Olympic sports back when they were absolutely dominating. You had the athletes basically living at a center with physiologists. And every day, you know, athletes would get up and they would come and get checked by the physiologist and the physiologist say, here’s what you’re capable of doing today. And every day they would adjust the plan based on that and that made them absolutely dominant. And there was a time when you basically had to be an Olympic athlete to have that sort of customization now I see this AI software making this available to everybody.

Rob Pickels  39:59

Yeah, I mean to put that into context. I think I can talk about this now because it’s been many years. I was at the Olympic training center for a long time doing biomechanics research. And we monitored rate of force development on all the Olympic weightlifters every week. And that led to the changes in their training for the following couple of days. Yeah,

Dr Paul Laursen  40:17

exactly. I mean, that is that is exactly where this is going. You’ve got a physiologist, you know, potentially, and a bio mechanist as well, right? Like, look at a technology like plan tiga, which is an emerging foot sensor, basically, you put your insole into, which is a force plate, and you put that into into your shoes, or your cycling pedals. Right. And this is, my colleague, Matt Jordan, who worked for the install does work for the Calgary base with the Olympic training center there. This is his one partner, his innovation. So that’s coming. And just just to Rob’s point, like you will have a biomechanics and a physiologist on board with you on any given day to make those assessments. So yeah, back to your five year question. That’s where I hope things are things are going in this is the

Rob Pickels  41:09

power rate of AI, is that as coaches, we can integrate more information than I think athletes can because of our expertise. And because it’s it’s the daily job. But I do believe that coaches still have to pick and choose what information they want to integrate into their decision making. There’s 100 different sensors or vitals that we could be monitoring. But we realistically can maybe utilize three or four of them to make decisions. But something like AI might be able to pull in all 100 of those different signals and make some sense out of what just looks like noise to the rest of us.

Dr Paul Laursen  41:49

That’s the biggest advantage right there. We just don’t have the intelligence, one of the one of the features back to the initial definition of intelligence is it’s like it’s like RAM, right? Like, it’s your ability to process info. And we just can’t keep up with the machines in terms of that processing capacity. And think about the overwhelm sometimes that you feel if you’re looking at like a data set, or whatever, like, like, it just takes too much time to go through that and pick out the important stuff. Well, we can leverage the machines to, you know, grab that, that data, make a very quick analysis, and, and pull out those red flags. So it’s right in front of you. And, you know, once it’s done enough, and you’ve developed the trust in the system to say, Oh, this is getting this is getting things 99 out of 100 times, then it’s now it’s like, okay, this is a pretty solid tool. This is really helpful kind of thing.

Rob Pickels  42:43

So I in this question of where’s it going? How’s it going to change training, I quickly wrote down what I thought were the positives and dangers. And I think you both are touching on this, let me quickly give you my list. And then please respond to this. But the positives are exactly that. We’re going to talk more about this in a minute. But it can process a ton of data. And it can find things find trends that humans potentially couldn’t. That was what Alan cousins was talking a lot about of it might look in places and find trends that we would even think to look for, like it might very well look at what you’re having for breakfast every morning going, boy, whenever you have this for breakfast, you don’t train as well. And that’s something a coach would even think to look at so that that’s the positive. To me, the dangers are athletes becoming overly reliant on it, and over trusting it. Because as you said, there is that issue of garbage in garbage out. And sometimes that AI software might come up with something really out there that you shouldn’t be doing. The other danger for me is there are things that the AI software just won’t ever understand. Let’s take a minute to hear Robin carpenter talk about what happens when we asked Chet GPT to build a plan. Just keep in mind AI software designed to build training plans is different.

Robin Carpenter  44:02

All the new AI training software I actually obviously I haven’t used it because I’m a cheapskate. And I’m self coached and just hanging out here guessing in the dark, but it’s actually something I really enjoy doing. But there was a little trend going around I think a little while ago when chat GPT first got released to the public of people asking it to write training plans for a cyclist. And some of the stuff that was coming up with was just hilarious. Just bananas workouts where you’re doing a threshold workout of 40 plus minutes at FTP five or six times a week. And yeah, you we kind of expect these things to get more powerful and learn over time. And they are fascinating from a technological point of view. But the idea of letting a computer put your body through the wringer in something that can really your head can really be balanced on a knife’s edge. Definitely seems somewhat dicey to me. And frankly, we just don’t have the ability to quantify everything yet. We we’d love to write, we’d love to have an we’re trying, right? There’s all sorts of stuff where you input input, your use counting your sleep hours, it’s counting your HRV. It’s all these, these proxies for, for recovery. But in the end, there’s nothing quite like the subjective qualities of a human mind.

Rob Pickels  45:33

There was both of you, what’s your responses to those? And did I leave any positives or dangers out?

Rob Pickels  45:38

Well, I think on the positive side, Trevor, I think that we’re going to see an objectivity that comes from AI software like this, in order to take your breakfast example. I think that as athletes, we oftentimes do things to explain some data points away, oh, my heart rate was a little high on the workout today. But that’s because I had a burger for dinner last night, and oh, gosh, my heart rates always high when I eat burgers. But maybe that’s true, it’s probably not true. It sounds like an excuse to me, and it’s able to pull with that larger data set. It might know every time you’ve eaten a burger, all the training that’s happened, it has 100 data points, and oh, my God, look at that your heart rate really is higher. So I think that the objectivity is really important. What I will say, though, on a danger side, my hope is that as we and yes, I do think that there is going to be some false positives, or just some some poor decision making based on low quality data. But my hope is that this is a pretty common danger that’s cited, and the robustness of the software ought to be able to handle that such that the software should be able to recognize outlying data, that is not correct, and does not make decisions thereafter. And I know in the beginning, that probably isn’t able to happen. But one would hope, as we talked about five years down the line that this is baked into, and those mistakes aren’t made, at least not very often.

Dr Paul Laursen  47:07

So for me, the big one, was a couple of things in terms of we’re talking impact on training, right, and the positives and negatives. And what are the fundamental things first of all of training that we want to that we know lead to success? Well, one of them is training consistency, right? So this really gives us a potential tool to impact training consistency. And we just saw this recently with, we had one athlete using athletic, she’s used it, she’s an ambassador, but she’s used this thing for two years. And she, you would look at this athlete, and you would think, no way she’s going to go to Kona, or even do well there. Well, she did an 1109 kona Ironman was 25th, in our age category in the 44, to 49 category, and she followed our principles, all the various different tools. And when we look back at her data, while she trained every day, but like one, you know, kind of keeping healthy and following kind of the guidance. So that is that is one thing that it can do. Number two, it can educate us as we go along. And this is, I’ve spoken with a colleague, you know, even send back who’s like, you know, he’s the editor of IGA SPP. And one of you know, one of his visions is that we’re seeing this in the classroom actually, where we, you know, we learn best by doing, we can actually educate ourselves as we kind of go along in terms of the things that are important. So again, a hope for Athletica, and maybe other platforms, but is to actually use this as a training tool and get these insights into our training as we kind of go along. And I’m seeing this as we’re working towards implementing GPT into Athletica. And we’re really in the phase right now where we’re refining the prompts and those sorts of things. Because the prompts that you give it make a difference in terms of the output, but it is indeed pulling out all of the data that we need and seeing the things that we can’t, and giving us that really quick insight into what’s important, which frees up that mental space to do other things. So those are some of the positives that I see. And I guess Yeah, danger potentially, you know, again, to your point becoming over reliant on it, because you gotta you definitely gotta keep tabs on things because it’s you know, even as we’re as we are right now, with what I’m seeing in beta version with GPT is it’s not always perfect right now. It’s gonna get better and better. So yeah, we really have to be diligent at this point in the game earlier on, but yeah, it will get better and better I believe.

Rob Pickels  49:53

So the big question I have is who in the future is going to be consoling the the AI software when it starts? complaining, every day I bust my butt building the perfect workout for this person and they never do the damn workout.

Rob Pickels  50:07

Well, it can develop feelings. And when the athlete is upset because their performance wasn’t very good, they can yell at the coach. And then the coach can yell at the software and the software has to feel bad about it.

Dr Paul Laursen  50:18

There we go. Yeah, no, it’s emotionless Yeah, for sure.

Rob Pickels  50:22

So we got a couple other things that we really wanted to cover here. And one of them we started to touch on, and really interested in hearing your perspective on this. But it does feel like one of the the biggest positives of this AI software is addressing a problem that we have created for ourselves, which is data overload, we have more and more sensors, we have more and more data. And we have gotten to that point where for a coach or an athlete, to go through all the data, even from a single workout and try to analyze all of it, they could spend hours, they could probably spend more time than they spent working out. And it seems this is something that the software can do really well.

Dr Paul Laursen  51:05

Yeah, no doubt. That is our key goal is to you know, again, as these 1000s of sensors are sort of out there. It’s like, well, what’s what actually matters? In the old The old adage is you don’t to measure measure what matters? Well, what does actually matter, probably follow a follow experts to find out what matters, I think we’ve got the key ones right now that you probably you know, if you’re a coach listening to this, you’re probably already using them in terms of the the movement, the power, the movement speed, and then the heart rate is an internal market monitor. But then now there’s, you know, all the whoop devices and aura rings. And yeah, it’s totally overwhelming, right? When you think well, what, what do I What do I look at as a coach, when what’s the athlete expecting and you know, you start to feel these panics. And then these sorts of things. Well, imagine if that panic could kind of you know, be alleviated a little bit and you had a system that was kind of, you know, your assistant that was that was making a little bit of a snapshot on on all these sorts of things. And then, you know, if a red flags raised, or then you kind of, you know, maybe we should pick up the phone with this athlete or whatever. So, yeah, data overload is, is definitely one of the biggest ones. And then also, with data overload is going to come data insight, you know, it’s like, I do believe we’re going to start to see things emerging, that really weren’t totally apparent before to us as coaches. And the the key example, that that’s forefront of mine is the concept of durability. Um, you guys have probably spoken about durability before, but not everyone is, it’s sort of the fourth dimension is not everyone’s kind of aware of the importance of durability. And this is really like, you know, with endurance, the maintenance of your power, your speed, relative to physiological stress, a lot of the science is really sort of showing now the, you know, you need this, it’s really all about fat burning, and you know, the longevity of your your fat oxidation into, you know, and probably those other structural components, but getting more insight into that. And again, having that tool to say, you know, here you go your, your aerobic D couplings off the chart, you probably need to work on, you know, acts in your training to enhance that. So these are some of the exciting things, I think, and again, back to that education tool, I think we’re gonna get educated by this as we go along. And coaching may change accordingly. So it’s kind of it’s exciting for me.

Rob Pickels  53:37

So let’s though look at a couple of the dangers. I’m interested in how you plan to respond to these, you brought up the garbage in garbage out issue. What happens if it’s bad data going in, and particularly because it’ll kind of be a black box, you know, Coach can look at the recommendations from the software, but it doesn’t know how those recommendations were created, what it was looking at, and whether it was looking at all good data to come up with those recommendations. So it seems to me that’s the danger of letting the software do all the data analysis. And how do you address that?

Dr Paul Laursen  54:10

Yeah, you’re it’s just it, this is the reality of the situation we’re in right now. And maybe it’ll always be that way, but it’s getting better and better. And the classic example is, so as we go into the weather that we started out, talking about, I’m swimming a lot more. I’m using my my Garmin to to capture my swimming. I can’t believe how far they’ve come across on the heart rate data from the wrist sensor. Like I’m actually like monitoring as I’m kind of going like oh, that’s that actually seems pretty spot on. So they’ve done something to improve this in the last couple years. Because two years ago, I would just I wouldn’t even wouldn’t even look at it. So sensor companies are getting better and better the data will like the integrity of the data will improve as we kind of go for record. And it’s because they’re actually using machine learning alongside that in the data processing portion, before they it actually comes back to the giving you kind of that output. This is, again, why we need, we need coaches on board to, you know, really understand common sense and what these, you know, these numbers should read. And, you know, coaches are responsible at the end of the day for their athletes. So they need to still sort of be over this, to see when these errors occur. But, you know, this is one of the things it’s on our dethleffs, as well as to be able to Red Flag data. That doesn’t make sense, right, like heart rate shouldn’t be through the roof at zone four, when a person is just sort of starting their, their warmup when they’re, you know, walking or jogging at Zone Two, you know, I mean, so that’s, again, another challenge for the companies that are that are involved. But it’s also an important responsibility for the coach to kind of look over and not become too over reliant on the on the system, I believe.

Rob Pickels  56:02

Yeah, as you were both talking, I was sort of off in my own little world daydreaming as oftentimes happens. And, you know, Trevor, I think that you had brought up the concept that we might be discovering things that we didn’t necessarily know because of this. And my mind went to Oh, man, yeah, it would be really interesting if we could pull like daily activity data off the bootstrap and understand how that plays into it beyond just the workout stuff that we’re talking about how does the daily data play into somebody’s adaptation and recovery. But then also, I began thinking, I don’t wear my whoop, strap all the time, sometimes the battery dies, sometimes I just forget to put it on. And I wouldn’t want the system to be confused by a day that had really low activity when ultimately it was really high activity. And then that led me You know, Paul, exactly to what you said, at the end there that on your dev list, you have to develop this ability to flag data as being inconsistent or incorrect. And yeah, I was just thinking kind of by channel like, Hey, don’t trust my whoop from this day, because the battery died, I forgot to where it whatever it was, in the beginning until the AI is able to discern that for itself. It’s a way that us as humans and athletes, and coaches can help the system flag data that really should not be taken into the equation 100%.

Rob Pickels  57:21

Someday, going back to it might be able to flag things, add a code to it and see, I’m gonna give you an example of myself, I’d be interesting how the software would respond to this, you’re talking about the the power relative to heart rate, and having to be able to say there’s something wrong with this data, because that heart rate isn’t right for that power. But as you know, I’ve got atrial fibrillation. When I go into afib, my power and heart rate completely don’t match up with one another. And what I’d be interested in knowing is whether software to have the intelligence to say that’s just bad data versus a there’s an issue going on here. Because I can tell you, I tell everybody that I first saw it on a ride on my 50th birthday. But I did go back through older data and went oh, there it is, it had happened before that I just didn’t know how to recognize it with the software be able to recognize something like that.

Rob Pickels  58:11

I’m gonna hop in here real quick, because something that I’m always impressed with Trevor is that we use just so everybody knows we use an AI software, when we do transcripts of our episodes, and it’s called otter.ai. The first time a guest comes on the show, we have to like click on Name, it just has like Speaker One, we click on the name, we say, hey, that’s Trevor Connor. What’s really amazing to me is that you could have had a guest from like the first 10 episodes of the show, the software is able to recognize that person 300 shows later, it knows boom, and it automatically says, Oh, that was such and such. And my hope is yeah, maybe in the beginning, it’s not able to say Oh, Trevor is in afib. But a coach or an athlete was able to identify that that was a fib and then the software knows, and it’s like, oh, I recognize this, and it finds it in all of the other pieces of training. And then maybe we even get insight into this specific workout or warm up or meal oftentimes leads to AFib in this athlete, it would just be amazing to see information like that. That’s what

Rob Pickels  59:19

I was getting. I think that’d be great. Because when I saw an AFib specialist, they told me, usually ever for a couple years before you see it. And that’s question, could the software see it way before you’re aware of it?

Dr Paul Laursen  59:31

Oh, I just depends, like, what does the data look like when you have a fib? Is it is it going, you know, up towards the two hundreds or is it? What are you seeing in terms of a number?

Rob Pickels  59:40

It’s an interesting response because I have both AFib and atrial flutter and they go hand in hand. So what I will see when I go into AFib Yeah, you’ll see my my heart rate skyrocket. And you know, I don’t see too often go up to 200 but it’ll be up in the 170s up in the 180s and use go shouldn’t be that high. But what will happen pretty quickly is then it will go into atrial flutter where it just flatlines round 141 45. So if I’m going really hard, and 145 If I stop pedaling, I’m 143.

Dr Paul Laursen  1:00:17

So there you go. I mean, if you develop a system, you know, machine learning system to recognize those patterns that you’ve just described, then your system red flags that, there you go, I was going to suggest using the fourth frontiers, which is a new heart rate monitor that actually does like live ECG. It’s quite, quite amazing. But you wouldn’t even need it in that context. Because if you develop the, you know, if you recognize if those patterns were kind of consistent, and they were consistent across others, even your Garmin heart heart rate monitor could potentially pick it up for you with the right the right system.

Rob Pickels  1:00:53

One seven days, that’s child’s play. Sit at 170 for six hour mountain bike race with my little bird heart. At the right moment, I could do that.

Rob Pickels  1:01:06

For both beginners and veterans polarized training is the best way to get fast and stay fast year after a year. And this is the perfect time of year to be thinking about how polarized training can help you in our new guide featuring Dr. Steven Siler explore fascinating and helpful topics, like how polarized training is different from sweetspot, how to bust out of performance plateaus, how to polarize all season, how to build durability, and how to time your high intensity work. With this complete guide from fast talk labs, you’ll have everything you need to polarize your training like a pro and unlock your Leap. Learn more at Bass talk labs.com.

Rob Pickels  1:01:49

What I’d like to finish out here, Dr. Larson is talking a little bit about specifically what you’re doing and what’s going on with athletica.ai. So the first question I’ll throw at you, you are using it to do some research with Dr. Seiler. Are you willing to tell us a little bit about what you’re doing and what you’re hoping to accomplish?

Dr Paul Laursen  1:02:09

Yes, for sure, for sure. So lots of different projects are underway. But the main one that and we’ve even started a pilot version of this. And the main one actually relates to the training load alongside the menstrual cycle in females within this is in collaboration with mera care, which is a fertility tracker, and we’re getting our so this is one of the research projects that we’re doing, but we’re getting our, our gals to, they pee on a stick every morning. And this mirror care device basically gives them a series of metabolites and HPA Ganatra hormones that are telling us insight into the menstrual cycle. And yeah, just really, you know, that’s That in itself is quite fascinating. But the real underlying question is, should a female athlete train to her cycle, and it’s really just a fishing experiment, it’s very similar to what HRV was, maybe back in the early days where we’re, yeah, we’re just it’s quite remarkable just how variable the menstrual cycle is. And then how training load which is calculated by Athletica is corresponding to these different phases of the cycle. Kudos to Phillip Cochise, who’s running the study and and all of the athletes that are that are part of that. And then we’re going to continue that with with Stephen and Monica truant is a professor at the same university with the Steven slathers at and works with the IOC on the red sort of area relative energy deficiency syndrome. And yeah, that’s just one of the research projects. But yeah, it’s we’re hoping to have you know, many of these types of projects, Steven might have talked about time wear, which is a reading device, basically measuring your, your respiration with with a shirt, the time our shirt. So they’d be the be some studies going on with that. So respiration, in fact, is probably telling us something completely different than we thought it thought it was, then then just your heart rate in terms of stressors. So that’s another project and on and on and on, I’ll kind of go like there’ll be some with plenty EGA and all these different sensors, super Sapiens, so blood glucose, blood lactate in the future, but you kind of get it right, like there’s all these sensors are kind of emerging, and that we want to use, use athletic as also an r&d kind of platform to be able to grab meaning from all of these and bring back insight to the coach and athlete.

Rob Pickels  1:04:44

So let’s kind of round out tell us some of the exciting things that you’re dealing with athletica.ai and I think I want to start with because I read about this as I was reading about your software, bringing in these GPT features. So you’re talking about actually Having the software read the descriptions of the workout that the athlete put in there and look for the trends there, which to me was really exciting to hear, because I always say, I can get more out of the athletes description of the workout than any of the charts.

Dr Paul Laursen  1:05:13

Yep, exactly, exactly. And I couldn’t agree more. That is the biggest insight is that I get to Trevor is like, what is an athlete actually saying in terms of their comments? So yeah, we and we’ve got that section. So we’ve got both rating of perceived exertion marker on after you do a session, and we’ve got feel as well, we can rate a session as hard and it feels really good. But we can have a session that’s hard and feels really bad. Well, those two sessions mean completely opposite things. And then corresponding comments as well, they really put the picture together, and they bring the context in play for the coach. So yeah, that’s we’ve got GPT assessing those variables in alignment with, you know, what was actually done from the device in terms of the power that was done for that workout and segments of that workout, too, which is pretty cool. And then also the heart rate response as well, if available. And yeah, how it how it sort of corresponded, it’s, it’s quite, quite remarkable. And then, you know, we can be as verbose or concise in terms of that, that output back to the coach and athlete too. So I’m seeing these in beta. But yeah, by the probably by the time this, this goes out, I’m seeing I’m seeing this in pre pre beta I should say, but that by the time this goes out, it should be in beta fully, and then won’t be too long till we do some final fine tuning and testing. And then that’s available for for coaches and athletes too, which is pretty cool and useful. And, to us pretty mind blowing. So I haven’t I haven’t seen anything sort of like that out there

Rob Pickels  1:06:49

yet. No, I haven’t seen that. I’m really excited to see how that plays out.

Dr Paul Laursen  1:06:53

There’ll be some growing pains, no doubt, we’ll get some, we’ll get some flack that it will, you know, won’t be 100%, I’m sure. But again, with with all these, this is the fascinating thing is that every time you build this new little bit of software, this little part of it, once it’s good, it stays. And then you find the next thing and you build out on top again. And it’s just it just, it’s like a Lego sort of structure. And it just keeps you just keep building on top of it. And it’s only average in the beginning. But slowly more work done. And with passion, as we all have on our team, it just actually become something.

Rob Pickels  1:07:32

That’s great. So going with your metaphor, which I like, what are some of the other exciting Lego pieces you’ve just added?

Dr Paul Laursen  1:07:37

Well, this one. And again, I’m the inventor of this is Andres ignobly. And he reminds me that isn’t AI per se, it’s very mathematical. But it’s we I just I’m finding it so useful. And it’s, it’s called our workout reserve. And basically, he’s he’s taking the maximal mean powers or maximal mean paces, right of any given workout. And he is comparing those outputs to the last six weeks. And he does that sort of in a time course it kind of runs the The analogy I probably try to use at first is the battery analogy or in terms of like, you know, how close are you to what you’ve done in the past, and we see this line kind of falling. And the closer the line falls from 100% down to zero. If it hits zero, you’ve hit like a personal best kind of thing, right? And this could be across any maximum mean power duration that’s on the spectrum, right? Whether that’s a five hour Ironman, or 10 seconds sprint, right? We know what’s been done in the last six weeks. And of course, we show you that. And why is this important? Well, again, in the context of training consistency, you probably don’t ever want to burn yourself on any one of those maximum lean powers or maximum lean mean speeds, any of those could cause issues in terms of repeating again, so we’re really finding this one of interest and value in the context of training consistency, it’s like, Okay, I’ve just, I’ve gotten down to zero on this workout, that’s great. But if you’ve ever gone like, you know, you get into this really, really dark place of like minus 50s. And stuff, you know, you might get injured or sick, sort of shortly after. So we’re finding it as a very useful tool in terms of just getting that regular signal, right across anything that you might be working on. And that’s shown in every single graph, you know that you’re in terms of your session analysis, we’re seeing that, that workout reserve, so that’s pretty cool. That’s not available anywhere else. We’re creating a Garmin app for that as well. That’s going to come out from a developer actually met on your forum. So thank you guys for for that. So you got some great, great individuals on that on that forum. Lots of real, real intelligent bunch. So that’s another one. The final one I’ll mention is the is the power profile again, we’re really, really proud of our power profile. This is Peter Leo’s work. And Peter is just this pro cycling tour coach, I’m working for the Australian program right now. But he’s he’s a, he’s a German. And it’s really from his work, we consider sort of the gold standard in terms of his power profile. And that does have some machine learning in it. But it basically whatever you do on your garmin or your any of your devices, we see that that power speed profile, which is on there with best practice, so and we mentioned the workout wizard already, but yeah, and I guess the underlying key thing that this is from Andrea, that he’s most proud of is it’s our under underlying logic in terms of the recommendation for next session for like, what’s the next training load that I should have? It’s going to get me to, to the events in the best best shape again, go back to that Cindy example, it’s all about getting the most out of any individual, when we did that with Cindy, using this logic for the last two years. Yeah, that’s what we’re sort of most proud of.

Rob Pickels  1:11:06

All right, well, I hate to say it, guys, but I think it’s time that we wrap this up. So let’s dive into our take homes here and Dr. Larson, we will start with you.

Dr Paul Laursen  1:11:16

Alright, so my take homes, I think we said it, it’s here to stay. And it really is all about getting on board with it and learning how to leverage the tools that are in front of us. And those that do I believe, are going to be probably and again, I’m echoing Joel Freels points on craft for coaching. So um, you know, it’s nothing new. But that is the key message. So coaches, yeah, I think they should probably, you know, invest a little bit of brainpower and I guess, finding out what’s out there, understanding the tools that are that are there and available for them. And running alongside of that you’re going to be better off in the end. The second one probably is just the recognizing the Gago the garbage in garbage out, is really important. And especially it’s going to be important in the earlier stages, where you know, the sensor data integrity probably isn’t up to speed yet. So the underlying principle is that athletes want coaches that know how to use these tools, coaches, that you know, your job is not going away. That’s, you know, athletes are going to stick with you. But they want coaches that can leverage the power that’s within AI. And that’s going to make you a better coach. Rob?


Yeah, Trevor AI in training is inevitable. The integration of AI in athletic training, it’s not hypothetical, it’s a present and lasting change. So athletes, trainers, coaches, they need to adapt to this technology, or risk becoming obsolete. The AI training software for my second one, it holds immense potential. It can offer insights beyond human cognition, it can process a lot of data. However, there are some dangers, right, where athletes can have over reliance on the technology and the software’s limitation and understanding nuances. And then my third one is, you know, handling data overload is something that AI is really going to be able to do and it can be instrumental in managing that data overload from all the different sensors, helping coaches ultimately make better decisions. But the most important thing that I think you should know is that that answer was written by chat GPT. And I just read it off my computer. It’s frightening. Oh, dear,

Rob Pickels  1:13:31

what was your question? What did you say to chat APD? Give me my one minute taiko

Rob Pickels  1:13:34

as Dr. Larson was talking, I wrote what are three take home messages from this podcast outline and I put the podcast outline and I didn’t read everything that it said it wrote a lot more than the one minute take home. So I was trying to like scan and read, you know, the salient points there.

Rob Pickels  1:13:55

I really don’t think I can follow that up. But I still have to get my take home.

Rob Pickels  1:13:59

So given the human

Dr Paul Laursen  1:14:01

it did a pretty good job. All right.

Rob Pickels  1:14:04

Yeah, that was good. Okay. Um, frightening. I really can’t follow. Up to Oh, no. So here’s my take on being totally frightened by that. What I found interesting this episode was there was a big question that we really wanted to answer, which is, where is it going? And we spent 30 seconds on that question, which was just all three of us gone. And I think that is what is important because there are real positives that can come out of this. There are real dangers. It can go in either direction. And that’s the issue. We don’t really know which direction is going to go and so my take home and what I got from this because I’ve been avoided to the software myself, is we particularly as coaches need to get involved. At the end of the day people like you Dr. Larson, who are developing the software, you are trying to sell a product you want it to be something that your customers are interested in and they can use. So we have a choice as coaches, we can avoid it. And then who knows what direction it’s gonna go in, or we can get involved and give the feedback and help shape it and make sure it goes in the direction that’s going to help us. And I think that’s my take home is I don’t think we should be putting our heads in the sand. I think we should be getting involved. And I didn’t read that judge if it didn’t come up,

Rob Pickels  1:15:27

or did it.

Dr Paul Laursen  1:15:32

I really appreciate that, that take home there as well. Trevor, and then I will just say, you know, if there coaches out there, we do have a coaching platform. And we’d love it. You know, we also have a coaching course, too, as a free course that we give on how to use Athletica. So you know, if this podcast did speak to you, please reach out to me directly. You can do that through the athletic website and take our free course, the coaching platform is free as well. There’s no charge for using it and reach out to me directly and if you do want to get involved.

Rob Pickels  1:16:05

Fantastic. Well, Dr. Larson, always a pleasure having you on the show. Thanks, Trevor. Thanks, Rob.

Rob Pickels  1:16:10

Great scene. That was another episode of Fast Talk. Subscribe to Fast Talk wherever you prefer to find your favorite podcast. Be sure to leave us a rating and a review. The thoughts and opinions expressed on Fast Talk are those of the individual. As always we love your feedback tweeted us @fasttalklabs or join the conversation at forums.fasttalklabs.com. Learn from our experts at fasttalklabs.com for Dr. Paul Laursen, Keil Reijnen, and Alex Howes, Dr. Andy Pruitt, Dr. Larry Meyer, Robin Carpenter, Sonya Looney and Trevor Connor. I’m Rob Pickles. Thanks for listening!