This is Optimizera weekly newsletter sent every Friday from a senior Verge reviewer Victoria Song that analyzes and discusses the latest phones, smartwatches, apps and other gadgets that promise to change your life. Optimizer arrives in our subscribers’ inboxes at 10 a.m. ET. Report yourself Optimizer Here.
This time last year, I had cut my four-mile run time by 16 minutes, was lifting three to four times a week, and had lost 10 pounds after six months of training. I felt incredible. Then life happened.
A year later, I hadn’t run more than 5 miles in three months, had gained those 5 pounds back through stress, and was plagued by injuries, illnesses, and other health issues. This was largely due to factors beyond my control. Frustrated, a month ago I decided to test three AI trainers and fitness plans: Fitbit’s AI health trainer, Peloton IQ, and Runna. I would try them out while training for a 5K race to see if I could improve my time, which has dropped from 31 minutes to 38 to 40 minutes over the last year.
Long story brief, I ran this 5K race last week. I improved my time by a full five minutes. After I told all three AIs to go on a trip.
I’m not generally against AI coaching in health and fitness apps. The amount of data is often overwhelming. For a lifelong achiever, recognizing and accepting your limitations is a constant journey. Having an clever guide to check in with me when I’m not being realistic or slipping into a negative mindset is – in theory – a good idea. However, reality is never that elementary.
The problem with these trainers is that they can demystify the training AND adapt it to your individual circumstances. By configuring these shoes, you set the AI a specific goal – lose weight, get fit, run a distance in a specific time, or some variation of this. Using chatbots like Fitbit’s AI trainer, you can tell him other details, like “I’m starting a new medication” or “I’m prone to shin splints and have access to a Peloton bike.” Theoretically, this helps the AI better tailor its recommendations.
For example, Fitbit’s AI suggested that since I was coming back from a two-week illness, I should incorporate gentle steady-state Zone 2 cycling, walking, and running to take a break from my daily responsibilities. He generated a program consisting of three training sessions per week. Not bad.
However, the IQ peloton was a bit more relaxed. I had to complete three workouts to unlock the AI insights, and in the meantime the suggestions were based on my previous workout history from four years ago. Meanwhile, Runna’s AI-generated plans were more or less based on the survey. After some tempo-based workouts, he or she may adjust your goals or provide advice. If you are ill, traveling or injured, it is your responsibility to notify the app.
First, these AI features don’t hold you accountable. It’s very uncomplicated to figure out your way to a longer rest. If you want tough love, you have to to say artificial intelligence is what you want. Even then, you can always turn it off in the settings if the AI is getting on your nerves. Let’s say you feel a little tired but can still exercise. You tell the AI, “I’m tired today.” You may need to hear, “Just go out, see how you feel, and quit early if necessary.” Instead, you’ll probably hear, “Oh, no problem, go easy on yourself and take an extra day of rest!”
I consider myself a disciplined person, but I am human. When I had bad days, I manipulated Fitbit’s AI to get me out of trouble. He never called me out on my bullshit. With Runnie, it’s very uncomplicated to end the program and start over. I could ignore Peloton IQ’s weight lifting suggestions if I don’t want to suffer from muscle soreness the next day. There are no consequences other than remorse. This can also be rationalized. Sometimes the thought of explaining yourself to an AI trainer seemed so much work that it became more appealing to avoid it altogether.
This is much more complex with a human trainer, doctor, running group or accountability partner. Whenever I’m in a rut, I can count on my spouse to look at me and say, “You know you always feel better after a run. Just go for 15 minutes. If I want to miss a race in the cold, my friend will remind me why I signed up for it. I don’t want them scolding me the next time I go to the doctor. Yes, it does cause some anxiety. But I don’t want to let these people down. It motivates me to show myself. The AI may be able to read my metrics, but it’s not smart enough to recognize when, psychologically, I might need support or a break.
Another problem is the one I wrote about earlier: the obviousness of the advice.
It’s okay if you’re a beginner. At the beginning of your fitness adventure, any information is helpful. But if you’ve been at it for a while, it’s usually a repackaging of things you already know. Runna noticed that my pace was inconsistent. I start too fast and feel tired by the end. I’ve known this for the last 10 years. Fitbit’s AI often told me to aim for eight hours of sleep a night and try a bedtime routine. I’ve known this since childhood. Peloton IQ occasionally helped with some strength training, but that was it.
Additionally, AI often needs hand-holding. Fitbit told me that I should run on the treadmill due to the frigid weather. I don’t have a treadmill AND I hate running on them. I would prefer to run outside in the cold or replace runs with bike rides indoors if the temperature is below 30 degrees Fahrenheit. He confirmed my preferences and planned another run on the treadmill.
After two weeks of alternating between abusing and following the AI’s instructions, I ran my annual 5K on Thanksgiving Day. My legs felt heavy. My songs worked Thread for me and I was busy How I did. My race is about a week and a half away and this would be a benchmark of sorts. Every 30 seconds I felt like I was interrupted by an AI voice telling me I was behind or ahead of the pace.
I had to take three walking breaks, felt cramped halfway through, and hated every second of the run. My time was a paltry (for me) 41 minutes.
My posthumous AI analysis of the run left me reeling. In summary, the observations of the three trainers can be summarized as follows: Well, there’s nothing in your metrics to indicate that you weren’t well rested. Have you eaten before? Remember that you want to refuel before your run. Plus you were inconsistent in pace. Try to save energy at the beginning so that you can finish your workout well. Sleeping seven to eight hours a night can also help! Did you want to include medication side effects? Need to change your exercise schedule to effectively limit it to the race?
It was around then that I realized I was advising against using AI should trust my gut. My gut told me I was overwhelmed with all the data. I spent so much time training various AI tools on how to train me that I began to dread my workouts. I deleted my Runn plan. I took off my Fitbit and pressed pause while testing. And while I continued to use Peloton in class, I ignored the AI features. I changed my mindset and stopped trying to beat my 5K time and started just enjoying the energy on race day.
On race day, I didn’t look at my watch almost the entire time. I had no idea what my splits were, but I was proud that despite the more hilly route, I didn’t have to take any walking breaks. According to my Apple Watch, I finished in 36 minutes. That was five minutes faster than my turkey trot and a faster average pace than all the other runs during my AI fitness tests. It wasn’t a perfect run, but it felt like it Good start.
This is what improving your health is all about. A big part of it is the mental battle between who you were and who you want to be. The AI can’t really engage with you on this journey because it doesn’t really know you. Ultimately, it’s up to you to know what’s best for you. Sometimes this means the AI is about to shut down.