Having just completed an Ironman in Barcelona – a swimming, cycling, and running endurance challenge that took over 13 hours to complete – Jeffrey Williams was baffled by Strava’s newly-introduced “Athlete Intelligence” feature on his favorite workout tracking app.
At the end of the final segment, a 26.2 mile marathon that took him over 5.5 hours to finish, Athlete Intelligence – which uses AI to take those stats and summarize them into a few sentences – told him it was an “impressive long run,” and noted that his “training is clearly paying off.”
“They’re giving you this chipper ChatGPT – you just roll your eyes,” Williams told Forbes. He finds the feature slightly insulting.
“It has no context. It couldn’t differentiate a marathon where I had a profound emotional experience. I really did go to a [mental] place that I did not expect to. It can’t tell the difference between that and a Tuesday evening run.”
Strava users are notorious for obsessively pouring over their own stats – how many minutes per mile on a run, or how many miles per hour during a bike ride. Many follow their friends’ efforts and offer encouragement through the app’s “kudos” thumbs-up feature. But these AI-generated paragraphs, which were rolled out on October 3 to paying subscribers, are often generic or repetitive at best and unintentionally ludicrous at worst, many users say. And they rarely provide the insights users crave to help them improve their performance.
Instead, it tends to tell everyone that they’re “crushing it,” regardless if they’ve run a marathon or done a chill cool-down jog, per screenshots provided to Forbes — and the bland encouragement isn’t particularly welcome. Put bluntly, many amateur athletes don’t want saccharine cheerleading from a robot.
“It’s just a bullshit word salad of stats that are already conveyed pretty easily with the existing charts,” Arindam Bose, a longtime Strava user and cyclist in San Francisco, told Forbes.
“It was some nonsensical recommendation of fluff, as if I was trying to bullshit my way through a high school paper,” said David Maver of Philadelphia.
While many tech companies have been quick to adapt AI-fueled summaries and chatbots into their products, the results haven’t always been useful. Forbes reported in June how Target rolled out an internal chatbot to its employees, many of whom described it as a waste of company resources. In March, the Washington Post found that Amazon’s customer-facing chatbot did not “deliver on the promise of finding the best product for your needs or getting you started on a new hobby.” And that same month, a chatbot put out by New York City gave citizens bad and in some cases, outright illegal, advice.
Some users have revolted: After delivery company DPD launched a customer service chatbot, one user was so frustrated with the experience that she convinced it to write poetry lambasting its parent company.
Like many AI tools, Strava’s Athlete Intelligence sometimes hallucinates or makes the wrong assumptions.
On a recent four-mile run that had an elevation gain of just 72 feet, Athlete Intelligence told Maver of Philadelphia that his “focused workout” had “elevation changes throughout the splits suggest a challenging route,” despite Maver’s own characterization of the run as being “pancake flat.”
Iowa-based runner and race director Joshua Sun told Forbes that Athlete Intelligence was confused by his use of the phrase “Great Wall” in the Strava title of a recent run. Sun was referencing a local Chinese restaurant in Davenport, Iowa, but the bot congratulated Sun on his “run in China,” even dubbing it the “Great Wall Ultra…in a stunning part of the world.”
Strava’s CEO Michael Martin acknowledged that Athlete Intelligence is in an “early beta stage,” and isn’t really meant for experienced athletes, adding that anyone can easily turn the feature off. “What we have heard from our newer users but they have trouble understanding the data,” he told Forbes. “They are candidly looking for more encouragement.”
“The goal of Athlete Intelligence is for people to gain context and understanding of their performance and historical progress,” Martin said. “Some people are able to do that with the charts and data alone. We are providing a capsule summary and are doing so in an encouraging way.”
Some users Forbes spoke to liked the encouragement and the positivity of the AI.
Tzeitel Gutierrez of Black Diamond, Washington, appreciated it, even if “it’s a little bit simplistic.”
“Even on rides where I was struggling, it finds a positive twist, which is a little bit annoying – I need some tough love,” she told Forbes. “I kind of like it, I was skeptical at first, but it does get it right.”
Similarly, Sana Ajani, a runner in Chicago, was pleased that the AI was able to highlight a recent personal record for speed. “It looked over all of the runs that I’ve done and said, this is the fastest one you’ve done,” she told Forbes. “That’s not a connection I would have made otherwise.”
Strava was founded in 2009 by Mark Gainey and Michael Horvath, and built a devoted user base of over 120 million across nearly every country in the world. Within the last four years the company raised $185 million and was valued at $1.39 billion during a November 2020 funding round.
CEO Martin said the app’s new AI feature is built with “one of the leading AI companies” but declined to share which one. Late last month, Strava updated its privacy policy to specifically say “we use the information we collect to enhance the quality, reliability, and/or accuracy of our AI Features by creating, developing, training, testing, improving, and maintaining AI and [machine learning] models run by Strava or our service providers.”
