AI vs Human Coach

Training and Performance Fundamentals
1

As a programmer for 45 years, I have had a lot of time to think about the term AI… lately it seems all our collective interest is focused on this. It starts to open some interesting windows into both the current state and future state of AI.

What is AI? This is an interesting question. There is no clear definition. Usually it means the latest and greatest technology that we don’t fully understand or something on the horizon. For me personally machine learning is very interesting but no longer falls into the AI category. But this is just a personal definition and not the interesting part of this discussion.

LLMs like ChatGPT or Claude are now what most consider AI, but once we have AGI, or the next super intelligence LLMs won’t seem so smart either.

So how does this apply to training? Let’s start with machine learning. ML takes tons of data and tries to find patterns. For example. if day = Sunday then long rides = true… because lots of data points say this is true. People that do long rides on Sunday get faster. Not many people would think an IF statement is AI but on the spectrum put enough together and it starts to feel like it and how it has been marketed for a long time and generally considered AI in the past when it was new. I am not talking about any other product, just the idea. Machine learning is amazing at what it does.

Pattern matching is always wrong at some point. Your rests are too easy is probably a bad rule created by pattern matching but you never know. For sure wrong for some people. LLM’s hallucinate, machine learning creates bad patterns. Human coaches give bad advice too. What is the difference?

  1. You can’t have a discussion with machine learning. “Can you prove this is a good idea? I think XYZ.” Since you can’t have a conversation the only answer is of course it is right, the data says so…

  2. LLM’s - “Coach, are you sure about this idea? I think XYZ” AI answer - oh you are right, you are a genius your idea is much better than I was suggesting, do that.

  3. Human Coach - “Coach, are you sure about this?” Human coach answer - Here is why I believe this is true. You might be right for you but I would keep trying it. What do you want to do? Or yes, I am sure about this. I have seen it over and over again. Many people feel this workout is too hard at first but after pulling through they thank me afterwards… Their performance is rewarded by sticking to the plan.

For now, all three approaches — machine learning, LLMs, and human coaches — have real strengths and real blind spots. The question isn’t which one is AI. It’s that a sparring partner might provide more value than you would assume. Not saying everyone needs a coach, just don’t assume any of these are a drop in replacement.

3 Likes
2

I agree. Good coaches will already be embracing the new technology.

For example, there is nothing to stop a human coach using ML on the data she has gathered from her athletes over the years (or using something like Xert / TR to take advantage of a larger dataset).

That could help her come up with the basis of a plan for her current athletes.

She could use an LLM to challenge and refine that plan, along with the discussion points for each workout to address with the athletes.

The human element of the input remains. More the emotional / psychological aspects of the coaching relationship.

As with now, you don’t that get with a $50 a month coach, but if you have enough of the right coach’s time, you can build a truly effective relationship that gets a bit more out of you.

1 Like
3

Yes, exactly. With machine learning especially, there’s always a bit of 220 minus age built in. But no question there’s value in looking at the data, and ML for sure can provide insights that a human wouldn’t see without serious depth and time. The bigger problem I see is that there’s so much messy data and so much missing data into what goes into ML that it gets ugly fast. Ideas like your training stress balance is too low, maybe you should think about a rest can be insightful and well timed.

1 Like
4

What a great discussion! I have been messing around a lot with AI (thank you for the Claude suggestion, Alex!) and I use it to wright my articles these days. It always comes up with a good structure, which improves readability. I write the prompt, let AI write it, and then I rewrite the whole article.

The most important part of coaching is the interaction with the client. The back and forth and asking questions is where we can dail in. The thing is that every body is different and works different. It takes time to get to know this. Using AI can be very helpfull. I think my biggest concern is that an average user can tell when AI is hallucinating.

It is all about what is out there. For schedules, most schedules are 8 to 12 weeks. Some maybe up to 6 months. As a coach I look at the whole year and when the athlete is up for it, even more years.

I think it makes more sense for nutrition advise, since you can ask for research to back up the claims. But again, how can you check if it is good advise when you don’t know what you are talking about!?

Today, I was planning a roadtrip with our campervan. I had AI plan several trips. He worked them out very nicely, but when I was checking the km’s, he was regularly of by a few 100 kilometres.

My intake is AI is a very helpfull companion, but you need to check everything it says very carefully. In my company he is the intern. Great for doing bulk work, but you have to keep your eye on him. :wink:

1 Like
5

I like the spelling mistakes here. “wright” instead of write :slight_smile: Very human of you. I agree with your experience, must be validated. Or I say “are you sure” he says… no I was just pattern matching based on your experience and others mentioning having similar experiences…

1 Like