and missing all the important stuff…
1 Like
Which is… ?
look in the mirror before pulling out
1 Like
the coach and the athlete…
1 Like
A bit of a sweeping statement… but I kind of agree
1 Like
it’s a relationship business…
1 Like
Sean would have passed his test if only
in fact, if AI does anything it will highlight just how much of a relationship business it is…
3 Likes
Just remember, l am a bot apparently, I train myself !bots… there is a truth there.
Enjoying the conversation, I don’t really have much to add, other than when ai has a real impact on training I will have retired years earlier. It will happen but Im not sure most of us will be around to see it. Might be wrong, we’ll see. (My family arent living on Mars, which was a version of the furture semi recently)
1 Like
Totally agree.
He passed it on the third atempt… while we were not looking.
I’m not going argue as clearly you have far more knowledge of how neural networks are developed than me but I have been programming as a career since 1994. I am far from an expert in AI, not many worldwide are, but I know how computers work. Big Data is only as powerful as the algorithms used to interpret it. I know that there are some seriously clever algorithms behind Neuro Networks written by a handful of mathematicians far more intelligent than anyone on here that none of us would ever understand. I also know they create an image of the machine learning by analysing user defined training data, testing that data using user created test sets and then analysing the results of those tests using developer created algorithms. Using those results and patterns in that data that it finds the machine can then predict results based on new variables inputted. That prediction can then becomes test data once the results are confirmed and over many years those predictions can potentially become more accurate with more training data to use, this is the “self learning” part
You’d be training an AI model for an outcome of improved performance based on the historic data gathered, that allows it to suggest courses of actions that lead to the outcome, run simulations then do the real world actions to compare. Some of those will be wild and have to be trained out, and you’d need some guard rails.
There’s no reason it can’t be individualised, once you have a generally effective model, have it further trained on an individuals data.
The problem I think is not how good it is, it’s the believability. At least for M40+. They just don’t believe less can be more, same problem real coaches have except AI can’t come up with some bullshit personality the sheep will follow. And it can’t explain why, so it will need a bullshit generator to go with it.
Do you have a robot to do cotswold, or just got some poor actor in?
There’s every reason it can’t be individualised
There’s missing the point and, well, just missing the point…
You are assuming that somehow AI will produce programmes with less volume.
If consistency is already an issue, surely the data will suggest that athletes under train
They’ve been at this for nearly 60 years already
That’s Eliza from 1966
Hopefully the current iterations are a little more helpful
3 Likes
It doesn’t appear to be so.
Consistency appears to be an issue because when people do train, they over do it, because they over do it they’re more likely to do the next session at poor quality, etc. typically it’s the weekend group ride and or long run where people blow the doors off. So there’s gained from the mon/tue workouts.
As opposed to consistent series of shorter, easier structured workouts through the weekend. So volume/load is overall lower but more consistent quality dosing.
My Garmin has said rest day for the last 4 mornings despite the training readiness being 70 today.
I have of course ignored it 
5 Likes