AI Coaching

With all the talk of AI doing everything these days, has anyone tried out coaching?

In theory, plug in a couple previous training cycles and your corresponding race results, and see what the AI spits out as your next training plan.

Just curious, as I haven’t seen anything out there on the subject specific to triathlon.

Try it?

You can always ask ChatGPT to write you a plan and see what it spits out

In the long-run, it probably be able to do as good a job as any generic plan that a coach churns out, only thing missing will be that personal 1-1 touch
.

With all the talk of AI doing everything these days, has anyone tried out coaching?

In theory, plug in a couple previous training cycles and your corresponding race results, and see what the AI spits out as your next training plan.

Just curious, as I haven’t seen anything out there on the subject specific to triathlon.

it would be an interesting SW, and AI, that you can program with your previous seasons data, plus specific seasons objectives, and general reqs… .and then print out what to do each week attending to specific constraints for the week, using your watch data and a kind of feedback

I believe TriDot is an AI powered coaching system. There’s a UK club / team pushing it at the moment, but i’ve heard from several other coaches who’ve looked at it and say it’s no better than ‘ok’ if the athlete hits every session perfectly. If you miss a session it throws it out completely.

I believe TriDot is an AI powered coaching system. There’s a UK club / team pushing it at the moment, but i’ve heard from several other coaches who’ve looked at it and say it’s no better than ‘ok’ if the athlete hits every session perfectly. If you miss a session it throws it out completely.

I think that with enough training it would be better, what does it need?

  • push to receive “subjective” feedback, so with this kind of dialogue + data collected from training being able to follow up the progression
  • being able to receive specific restrictions for weeks (like work travels) or injuries or illness
  • being able to use all that information to recalculate the training plan

i think that the ai is capable to do that, but as a training plan, or as any coach… it shall be “teached” before

I believe TriDot is an AI powered coaching system. There’s a UK club / team pushing it at the moment, but i’ve heard from several other coaches who’ve looked at it and say it’s no better than ‘ok’ if the athlete hits every session perfectly. If you miss a session it throws it out completely.

I think that with enough training it would be better, what does it need?

  • push to receive “subjective” feedback, so with this kind of dialogue + data collected from training being able to follow up the progression
  • being able to receive specific restrictions for weeks (like work travels) or injuries or illness
  • being able to use all that information to recalculate the training plan

i think that the ai is capable to do that, but as a training plan, or as any coach… it shall be “teached” before

One thing I think many misunderstand (or maybe I do as well) about Chat GPT is it is a language model not a true AI. It simply predicts based upon reviewing a large quantity of written text what words should come next. In and AI model the model learns from data. The model is fed a large quantity of input data and resultant data and from that it interprets how best to answer questions based upon the desired results. You then have a feedback loop, it says do this to achieve results x, the adjusts based upon true results… and so on. In fact Chat GPT is a good front end and back end application for a true AI model where the output is not likely a text or speech type this. There is an AI training model called AI Endurance and they have incorporated Chat GPT as a way of getting input and output data to the user. https://aiendurance.com/blog/chatgpt-in-ai-triathlon-running-and-cycling-training

In the long-run, it probably be able to do as good a job as any generic plan that a coach churns out, only thing missing will be that personal 1-1 touch

This^^^^

It strikes me that these days we call writing training plans and workouts “coaching” mostly because there are lots of people who call themselves “coaches” who really just spit out training plans.
The reality is, this is NOT what coaching really is.
Coaching is about guiding an athlete, having a relationship with that athlete and helping them to navigate training and racing. The “workouts” are the smallest part of what a good coach actually does.

Honestly this is less a vent about AI and more a rant about the water down B.S. “coaching” we see in 2023

A good coach isn’t even remotely worried about AI because their relationship with the athlete can not be driven by algorithms

In the long-run, it probably be able to do as good a job as any generic plan that a coach churns out, only thing missing will be that personal 1-1 touch

This^^^^

It strikes me that these days we call writing training plans and workouts “coaching” mostly because there are lots of people who call themselves “coaches” who really just spit out training plans.
The reality is, this is NOT what coaching really is.
Coaching is about guiding an athlete, having a relationship with that athlete and helping them to navigate training and racing. The “workouts” are the smallest part of what a good coach actually does.

Honestly this is less a vent about AI and more a rant about the water down B.S. “coaching” we see in 2023

A good coach isn’t even remotely worried about AI because their relationship with the athlete can not be driven by algorithms

You are 100% right but alot of coaches sell “generic” 12 or 16 week plans online these days. I think those days will come to an end. An AI platform will do a better job as it will modify it based on fresh data etc. There are also a segment of the athletes that go to coaches for basic plans but never really interact with them, think that line of revenue will also be endangered

Has anyone tried Humango (https://humango.ai/)? I haven’t, but it is purportedly a triathlon coaching app driven by AI. I did look at it briefly, and it seemed quite robust, but I didn’t thoroughly test it.

With all the talk of AI doing everything these days, has anyone tried out coaching?

In theory, plug in a couple previous training cycles and your corresponding race results, and see what the AI spits out as your next training plan.

Just curious, as I haven’t seen anything out there on the subject specific to triathlon.

ChatGPT:

Write me a training plan for a 50K race
Training for a 50K race requires a comprehensive training plan that focuses on developing the endurance, strength, and speed necessary to cover the long distance. Here is a training plan that spans over 12 weeks and covers all aspects of preparation for the 50K race.
Week 1-4: Base building phase
Focus on building aerobic base by running at an easy pace for 3-4 days a week. Gradually increase the distance of your runs each week, aiming to cover at least 25 miles per week by the end of the 4th week.Incorporate strength training exercises to improve muscle endurance and prevent injuries. Include exercises such as squats, lunges, planks, and push-ups. Do strength training twice a week.
Week 5-8: Endurance phase
Increase the mileage gradually over the next four weeks, with a long run on the weekends. Your long runs should start at around 10 miles and increase by 2-3 miles each week until you are running at least 20 miles.Add in some hill workouts and tempo runs to build endurance and strength.Continue with strength training twice a week, and make sure you stretch well after each workout.
Week 9-10: Speed phase
Focus on improving your speed and stamina by adding interval training and fartlek workouts to your routine.Do two interval sessions each week, where you run at a high intensity for a set amount of time followed by a recovery period.Incorporate one fartlek workout per week, where you alternate between running at a moderate pace and sprinting.Reduce your long runs to 16-18 miles to allow for recovery.
Week 11-12: Tapering phase
Reduce your mileage and intensity to allow your body to recover before the race.Cut your long runs down to 10-12 miles and focus on shorter runs at a moderate pace.Rest well during this phase and make sure you’re eating a healthy, balanced diet.
Race Day:
Start at a comfortable pace, and don’t push too hard too early.Take in fluids and nutrition regularly to keep your energy levels up.Listen to your body and adjust your pace accordingly.Enjoy the experience and celebrate your achievement!
Remember, every runner is unique, and it’s essential to adjust the training plan based on your fitness level, experience, and injury history. Also, make sure you’re getting adequate rest and nutrition throughout the training period to maximize your performance on race day. Good luck with your training and have a great race!

I asked ChatGPT to give me a plan for a “sub 18 minute 5km”

the first line was “to run a sub 18 minute 5km requires holding a fast pace of at least 3 minutes and 24 seconds per kilometer”

well I thought that was weird because an 18 minute 5km is actually 3m36 per kilometer.

then I figured, it probably thought that a “sub 18 minute” time meant 17 minutes, rather than 17m59s which is what I intended

I’ve been using athletica.ai for a while. Can’t speak to its AI credentials but it regularly adapts the training program to activity and performance. It works for me. I don’t really need or want a live coach for everything.

I gotta figure out a way to change my default typeface when viewing, because I keep reading “AI” as “Al”

https://youtu.be/uq-gYOrU8bA
.

I asked ChatGPT to give me a plan for a “sub 18 minute 5km”

the first line was “to run a sub 18 minute 5km requires holding a fast pace of at least 3 minutes and 24 seconds per kilometer”

well I thought that was weird because an 18 minute 5km is actually 3m36 per kilometer.

then I figured, it probably thought that a “sub 18 minute” time meant 17 minutes, rather than 17m59s which is what I intended

Chatgpt can’t actually do maths. It’s not trained for that. It might guess simple arithmetics but anything involving numbers has a high probability of bringing up allucinations.

Full transparency: I work for https://athletica.ai and we have been working on this for many years - building the latest in endurance sport science into an AI training tool. Alistair Brownlee wrote a guest blog post for us on this subject a few weeks ago: https://athletica.ai/chatgpt-vs-ai-the-future-of-endurance-training/

The key difference from ChatGPT that Athletica provides is that we dynamically update your entire training plan based on you previous sessions (pulled in from your Garmin, Strava or other data), so if you overperform, underperform or miss a session, your plan gets adjusted so that you still meet your training or race goal. We also provide alternatives with our Workout Wizard that deliver the same load and training value in a different session or modality. Finally, we have amazing human coaches and sport scientists that can answer any questions.

You can try it out for free without a credit card for 2 weeks.

While not labeled as AI, the newer Garmin watches do provide very specific training plans for short to medium distance run and bike events. I can definitely see this coming for Sprint to Olympic triathlons as well. Even now certain Garmin watches provide recommended bike or run workouts based on past workload and heart rate.

While I understand some people might need that one on one with a human coach, I love the idea of an AI generated set of workouts that change based on my current workload and my goals. Also people should not just think of AI as a fully automated entity, but a way of assisting people. For example, spell check is technically a form of AI, it’s not all Skynet :slight_smile: