Is There A Future For Artificial Intelligence In Coaching ? | Tri To Disagree

[@gtn]

Do you think there is space for AI coaches in Triathlon? 👀

[@trbeyond]

What AI will never replace is the human words of encouragement and support that a good coach provides. I know that I am probably on par or even inferior to AI when writing training plans for my athletes. But I am confident that my athletes success has been from the other things I provide as a coach

[@gtn]

do you feel it is too easy to ignore AI compared to a human too?

[@squngy0]

Good discussion, but I believe you missed one big point: most normal people already don't have a human coach. Most amateurs are currently just following a generic plan they found online. AI could make personal coaching far more accessible to more people, even if it will not be as good as a human yet.

[@gtn]

That's a really good point! Would we see more people trying out plans if they were coached by AI?

[@RealFrenchWill]

@@gtn That is exactly what I did. Humango is much cheaper than a plan with a coach on TrainingPeaks and feels like I am getting 80-90% of the value. It's not perfect but much better than no coach at all.

[@technodork101]

Yes, exactly! I would try an AI coach if it was significantly more affordable than a human coach. Currently, I use training plans that I find online, but I feel like AI could do a better job at adjusting things if I miss a workout or something like that.

[@JimCullen]

The "AI-assisted human coach" idea is a really good one. It's actually what happened in the world of chess decades ago. Computers have been able to beat the best in the world at chess for decades, but a human working _with_ a computer as a tool to assist them can beat even the computer on its own.

[@quangha1095]

Hi @gtn, is it possible for you to make a video that would go into details how to be a good support crew for a triathlete friend who is doing an adventurous race? Saw your videos on Norseman races and wonder if you can provide more insights on the support vans!

[@gtn]

Ooo that's a great idea 💡 thanks! Watch this space 👀

[@jazznroll5]

Polarized training or 80/20 makes creating your own training plan very easy. All you have to do is allocate time per intensity for each week, increasing time for 3 weeks and dicrease for 1 week. 80% of the time you will go easy up to a specific heart rate and 20% threshold with some vo2max intevals. Job done.

[@marie-louisekarlander2366]

After Kona I took a break from my coach 😭I had since 2018. Economy was behind that decision. So I tried HumanGo , I will write to them to. They were really good to me. But I choosed to not continue . I liked the running workouts, but the swimming was not in my level (?). I tried to put how long time, how many meters I prefer to swim. But the workouts were so short so they were like a warmup. The bicycle was not that bad. They were even hard sometimes, often one brutal warmup. I needed to warm up for make the warmup. In that case I think Hugo shall know my age (57) and know my capacity pretty well from data. Bicycle is my best part but I am not a professional. Actually I had expected more from AI. So now I’m coaching myself use Training Peaks and really think my coach did a great work. But it takes time to do the planning. //Marie-Louise

[@fstrey]

You can set the frequency and minimum time for pool sessions. I also had to adjust it as some for sessions would take longer to go to the pool than training. Since then it's working amazingly well

[@rikardengstrom7916]

I just have a problem with the broad term AI. An AI-driven coach would probably not consist of one AI-model. Another thing to consider is what data was used to train the different models.

[@uflux]

I use Tridot Ai coaching. It's great! Never seen more improvement than with that technology

[@gtn]

That's awesome! Have you ever had a human coach? 🏊🚴🏃

[@uflux]

@@gtnno a doubt I would so tridot works perfectly for me

[@DaveStecker]

I used to write my own training program as I viewed it as a hobby. I found myself analyzing data and programming far longer than what my sessions would take. This took away too much meaningful family time. I just got to the point where I just want to be an athlete. Invested in an AI training program and I’m excited to see what happens.

[@gtn]

That's super cool! Did you have strong sports science knowledge going into it? You must have learned a great deal 🙌

[@DaveStecker]

I did. I work in sports medicine and I also studied for the Strength and Conditioning Specialist credential (CSCS) and few years back. So I spend most of my days at a local high school designing training programs and evaluating sport injuries. I really enjoy it.

[@JimCullen]

Title says this is a coach's corner episode, rather than a tri to disagree one.

[@fstrey]

Tl;dr: Comparing apples to apples: AI > fixed plan. If I could I'd also have a coach on top of the AI. I've switched from a fixed plan to HumanGo this season, and the dynamic threshold calculations after each session made a massive difference on my progress. Instead of having to test my performance and adjust treshholds manually, the AI do that on the go for me. My pacing, power and load will adjust for each session. The right coach would potentially be better, but expensive, and as an average AG I'm not at rhe position to get a coach.

[@SBoots29]

I believe AI will benefit the sport. For me living far away from the ability to access a coach, group exercise ECT., I rely on AI and of course the You tube coaching. It gives me a sense of motivation and discipline. This of course cannot give me feedback on form and technique. With so much data though there could be a possibility on how to interpret so staying in tune with my body is something AI and not even a coach is going to know. Cheers

[@davethedogdude]

You need to use an AI model specific to training method desired and also programmed to not extrapolate what it does not know (as in hallucinate and make up random crap). Chat GPT for example right now will gladly design you a 4-hour cycling workout that includes 2 full on 1 hour FTP efforts as part of that 4-hour block. Or invent a 55 minute 135 TSS workout that is also impossible to complete. I can attest to this from experience goofing around with that as an idea.

[@nataliamartinkova]

Two different AIs failed to coach me to my goals in two successive years. In 2022, my performance lagged by 10 minutes over a half-marathon distance, and in 2023, by 15 and 22 minutes, respectively. Moreover, as James was saying, the AI-mandated watch harassment turned training into a chore that was difficult to enjoy. My VO2max decreased by 5, and the overall experience was negative.

[@SubtleForces]

I agree with James but would go even further: family, work, life, love, despair, health, sickness are just other data points which can increasingly be captured. The subjective can be eliminated. What can't be is the aspect of purpose and the selection of meaningful goals. That might require a conversation with a human...for now.

[@gtn]

That's a good point, the human aspect is needed for that end goal. Do you AI struggles to give out reasonable goals?

[@SubtleForces]

@@gtn I think it can give reasonable goals easily but a meaningful goal is much harder. AI can reasonably tell someone who has done a a few 5k and now his second 10k that he can try a semi. But is that a meaningful person for that person at that particular time in his life? That requires knowing about his career goals, his family and social life and how to balance his aspirations in these domains with the training requirements? Also, what if he actually prefers speed and would rather add some cycling which he only does occasionally and yet can't swim? What about even more subjective questions such as related to ageing or health or even aesthetics or self-image? I am not saying AI will never be good enough for such conversations, but that would require very advanced general purpose AI and that's probably still at least a decade away.

[@MladenSaracevic]

Cost factor is another argument for AI coaches. AI coaching is for amateur triathletes often more affordable than a human coach, who is available like an AI one. For lot of us is AI one the best starting points, if you need kind of a personalized training plan.

[@utuber221]

AI needs a lot of datasets to make a good judgemental decision. In essence, you as the athlete are training the AI and not the other way around. A company that offers such services would need to explain how these datasets are being trained to form your benchmark. Unfortunately, everyone is unique and has many variables at play. It is early games for AIML with LLM and training them. But will be exciting in 1 year or 2

[@klewis2048]

These "AI training platforms" seem to be little more than statistical models built from small-to-medium sized data sets of athlete historic training and performance. I'd love to have one of these companies come here and explain exactly what they have, how it works, and what the "artificial intelligence" component actually is, because I'm not seeing it bringing real advances in training tech, or anything that allows for the more nuanced approach a human coach would be able to provide.

[@MarkBridges1]

That will only happen if that company does a paid sponsorship with GTN

[@atlantaswelder]

Ai will never be as good a good coach because a good coach understands and almost vicariously views and lives the training arc of his player / pupil

[@winklertribe5268]

Because humans are sentient, and AI can not discern psychological/emotional nuance, AI can not replace a human coaching

[@Antonseidl_]

AI Coaching would be helpful if there would be a platform that actually uses AI and not just backwards calculate a generic plan 😅

[@svdb2981]

AI will be the tool in the furiture.

[@CarmineEspositoIT]

AI is 100 times better that just a program you downloaded/bought online and that will work just fine for the majority of people

[@pilifos129]

What are you using?

[@miguelsaez340]

Ollie’s power numbers have been leaked. 830W in 20’ is not to be trifled with

[@basildaoust2821]

I'm not going to waist my time with this shit. No matter what you say it will be wrong unless the answer is actually no matter what the state is now it will get better faster then anyone expects and coaches will all be AI at some point. Imagine if everyone had their own fucking coach that could run, swim and bike along with them and have all the intelligence of every coach that ever lived and check it all validate or trash the junk, and get better even while the athlete is doing the current workout. Now the AI would know not to change the workout underway while doing it that would just be bad I would think, unless the AI could tell that status of the runner and get them to say just do a light run for 15 minutes I would like to change our workout, I know you expected to do ABC today but I think you would benefit more from CDE today and that will improve your X goal by 22% or however you think an AI should talk. If it is still learning then right now might be a bad time but whatever you think will eventually be wrong. First chess was to hard then it wasn't, the Go was to hard then it wasn't, this will continue for everything you think should be too hard. I know some AI drawing suck but soon they won't because soon they won't be children drawing. If you think you understand you probably have no clue what is happening right now and I'm glad I no longer work and do not have to try and defeat AI to keep my job. Good luck everyone.

[@raginald7mars408]

there will be artifical Athletes - out performing anything - even themselves

[@ironman140.6]

Mark your "argument" was not apples to apples. AI vs not AI has other properties, not the angle you brought. FAKE argument.