I definitely could be wrong on this but it’s always been my impression that they utilize everybody’s data for the app’s model training and toggling those items off would just mean essentially not making those adaptations in your own training.
I’m hoping the update yesterday from Strava was a “nothing to see here”“carry on as normal” message but it’s hard to deduce that following the pretty detailed original comms.
Like a number of people here, I cancelled my Strava subscription and even if nothing does actually change, Strava can forget about me returning as a paying subscriber. As DCR says, it’s almost as though they look for ways to try and screw over their user base . Honestly, I’m done with it. My feed these days is only full of commutes, yoga sessions and Zwift rides and I’ve no interest in that.
The Strava Athlete Intelligence sounds like a CV play. It is a real joke. E.g. why do they comment on my pace after I do a TrainerRoad workout? But of course all tech companies must use AI to be relevant.
DC quickly updated his page with the article on it:
Strava’s Big Changes Aim To Kill Off Apps
However, ultimately, it doesn’t actually change things – in fact, kinda doubles down on it. In it, Strava basically says three things:
A) 3rd party apps can’t show data beyond the user itself. Just as before, this still breaks coaching apps. Strava tries to confuse people by saying “the overwhelming majority of existing use cases are still allowed, including coaching platforms focused on providing feedback to users and tools that help users understand their data and performance.” – however, what they’re actually saying is 3rd party coaching apps can provide feedback, as long as that’s computational feedback. Coaching apps still cannot let human coaches see the data. Had that been the case, Strava would have simply said that, and notified coaching apps. They’ve done neither, and multiple coaching apps I’ve talked to have confirmed Strava hasn’t changed their position, but has scheduled meetings to talk about the impacts.
B) Artificial Intelligence and Machine Learning is still prohibited. Again, Strava tries to keep distracting people by saying “training models” is not permitted. Yet the terms are very clear, both training and usage isn’t permitted. Still, as any company can tell you, even if it were just training (which again, it isn’t), these models can’t get better without training on the same data for whatever application purpose they are using. After all, last I checked, Strava themselves doesn’t give us permission to allow/deny their access to our data for AI training purposes.
C) “No impact to most developers” – In the update, they say “We anticipate that these changes will affect only a small fraction (less than .1%) of the applications on the Strava platform”. Except, that’s the most disingenuous way of writing it. Strava has tens of thousands of connected applications, the vast majority of which are tiny hobbyist applications that may have 1 user or a handful of users. Instead, it’s the big ticket applications that matter here with the most users. Strava isn’t say 0.1% of users, just 0.1% of applications.
–
Ultimately, it seems like Strava’s API itself needs an update if Strava believes users need more control. And to be clear – I’m 100% behind giving users more control. Strava could easily add check-boxes for “AI data usage” and “Show data publically” to the 3rd party platform API authentication dialog (just like they do today for other features). They could easily require 3rd party developers to ensure consent (just like they do today for other features). And they could easily require developers to adhere to various rules (just like they do today).
VeloViewer posted an update on Strava: https://www.strava.com/clubs/45060/posts/33582286
Thank you for all the comments and concerns around the recent changes to Strava API use. We have always had a really strong relationship with Strava, and are currently working our way through these changes with them. We will let you know more when our position is confirmed - this is likely to be in a couple more weeks.
It’s such a minor thing all things considered, but I’d be bummed out if they can’t do their end of year infographic anymore. I look forward to it every year.
That is my impression as well.
To iterate I am of the mindset of if I, as a user, authorized the integration between TrainerRoad and Strava then they should be okay with the fact that I am okay with TR using my data as they see fit. Maybe add more verbosity when enabling integrations on how the third party will use the data: kind of like how the Apple App Store shows what permissions an app requests when you download it (e.g location, microphone, contacts, etc.).
With that being said I imagine TR not having a toggle switch for users to opt out their data completely from AI or ML processing might be an issue in Strava’s eyes.
Edit: I just remembered that Strava’s new policy forbids ALL AI processing for third party apps so even if TR had a hypothetical toggle switch to opt into any AI processing it would still break this new policy as turn said switch on would break that policy.
Can Trainerroad remodel their app to record activities such as : running , walking , swimming … ? The app could then be free to use and ,if TR wanted , act as a social network,data exchange platform ; like Strava used to. Obviously adaptive AI and training plans are behind a paywall. I would suggest that it would be a great way for TR to grow their business as adaptive AI plans could be rolled out for other sports .
-1 vote for trying to turn TR into what Strava was.
We are still waiting on key cycling features (WLV2) which are keeping them plenty occupied along the many recent updates to the calendar, plans non-cycling imports & consideration and the like. I’d rather not have them diverge even further from their basic skills and focus until those are well handled at the very least.
I found the project manager/scrum master in this thread!
But I agree. My dream is a TR <> intervals.icu integration opens up and I can make intervals my hub.
Agreed, I don’t think TR have the bandwidth for a distraction like this. As it is, just working around Strava and having to do new data integrations is going to be a lot of work. It’s a real blow.
AI is introducing all sorts of data use issues. If you have a toggle to say I don’t want you including my data in AI or ML processing even if that information never leaves the confines of this app. App of a sudden if you get a ton of users opting out of AI that makes your AI features less effective.
Strava subscription just canceled, runs out in May, but hopefully they will see that they are loosing with there actions.
100%. The model gets better with more data. The toggle switch was more of a thought experiment than anything else. If you subscribe to a TR account you are subscribing to their core features: one of them being AI and ML.
Yep, lines up perfectly with my experience. If there’s insufficient growth for the quarter, pick a spot and squeeze hard, even if you don’t actually understand what you’re squeezing, why you’re squeezing it, or what it is you hope to squeeze out of it. I suspect (with no basis in fact other than experience) a round of layoffs may be on the horizon next.
Strava have issued a clarification
According to them this change affects less than 1% of users. In other words they aren’t bothered about us
Already covered by DCR and linked above as misleading, wrong, garbage… (my adjectives).
Less than 1% of apps, not users. We have no idea of how many users are in that supposedly 0.1% of affected apps but I imagine they do… which is why they’re framing it in terms of apps and not users.
Anyone use tapiriik? Looks like a good alternative
I’ve started changing the name of my Strava workouts to “I cancelled my Strava subscription and you should too” and description to “Reprehensible data sharing changes in the recent terms, see dcrainmaker dot com for details.”
I figure that might help get the word out to folks who aren’t forum sickos like we are…
Hah! What does your Athlete Intelligence summary tell you? Please share!