Can an AI-Powered Fitness App Outperform a Human Trainer? – CNET

Tech News

Products You May Like

An AI trainer can tell me to jump, but it won’t care how high. 

It was a realization I had when trying out FitnessAI, an iPhone-only fitness app that uses artificial intelligence. FitnessAI produces personalized workout routines based on information you enter, like your height, weight, gender, and available equipment at home or at the gym. But relying on technology alone has its limitations in honing the human body. 

I compared Fitness AI to the app Future, which connects you to a personal trainer virtually. With Future, I wasn’t just answering to myself or inputting numbers into an app. I was dealing with a real person who had real expectations. And with my bro-ish lizard brain, I wanted to show him that I could push myself harder.  

It shows the fundamental differences between training with machines versus humans. FitnessAI says the app aims to slowly ease people into new routines, meaning it won’t throw you into a new workout methodology. That’s because it can’t understand and synthesize who you are as a person, nor register your specific needs or nuances. It can take a few weeks for the app to start varying workouts in more unique ways as it crunches numbers, collects data and recognizes patterns. 

The AI-ification of everything is upon us. Since the launch of ChatGPT, the world has become entranced by large language models. From generating poems and resumes to computer code, ChatGPT rummages through troves of online data and forms human-like sentences, making old-fashioned Google searches seem pedestrian by comparison. 

As ChatGPT became the fastest-growing consumer product in history, there’s been a surge of companies integrating more AI into their operations. A Google search of “Fitness AI apps” will net you at least 20, including Fitbod, Freeletics and AI Gym. By one estimate, the global digital fitness industry will reach $26.5 billion by 2026.

There’s a reason I didn’t turn to ChatGPT for a workout program. Sure, I can add specific metrics to a ChatGPT prompt, but it’s essentially “autocomplete on steroids” and will churn out a response by predicting what the next best word should be. And when I told ChatGPT to create a routine around my theoretically twisted ankle, it recommended exercises while standing and bearing weight, which could make an injury worse. FitnessAI doesn’t have a chatbot built in, although that will soon change. 

Testing out AI in fitness 

Launched in 2019, FitnessAI relies on machine-learning algorithms to optimize sets, reps and weight for each workout. The app has a slick interface. Every workout features a 3D-anatomical model of the human body, showing the undulations of each muscle group, how to do each workout and which muscles are targeted. After completing the exercise, you can tell FitnessAI if you used higher or lower weights, and the app will use that data to adjust future workouts. 

Screen of fitnessAI app with upper arm day

FitnessAI app

Screenshot/CNET

My first session focused on shoulders and abs. I started off with five sets of barbell shoulder presses (five reps each), then switched to dumbbell shoulder presses. Because of fatigue, having two shoulder press exercises in a row didn’t feel completely efficient to me. After a dumbbell lateral raise, I moved on to a weighted decline crunch followed by some standard crunches. It was far from a grueling workout. Granted, I could have gone through and added additional workouts, but for the sake of testing, I kept it as is. 

The following day focused solely on biceps. Just like my previous session, there wasn’t a lot of variation. (Granted, there’s only so much that can be done with bicep exercises.) Going from a dumbbell curl to a barbell curl seemed repetitive. If it were immediately followed by a hammer curl, which targets the long head of the bicep, that would allow different muscles to be hit. After five different bicep exercises, my arms were shot. 

On chest day, I ran into the same issue, with FitnessAI recommending two workouts that largely targeted the same muscle groups: an incline bench press followed by an incline dumbbell press. I also didn’t see the point of starting off with a bench press and ending with a Smith machine bench press. I was surprised FitnessAI didn’t consider something like a cable chest fly, which allows for a wider range of motion.

At $15 per month, FitnessAI might be a bit steep if you compare it to numerous other fitness apps on the market. Still, I see it as a powerful workout tool that can log your activity and show you exactly how to do exercises. The app can vary workouts, improve efficiency and help you gradually achieve progressive overload, which could lower the chance of injury. It takes time to get the most out of it, so don’t expect a massive shift to your routine right off the bat. It also won’t recommend a different workout style unless you specifically indicate as such in the settings. But doing so requires some background research on your part. Overall, FitnessAI seems better suited for more experienced gym-goers. 

FitnessAI has confirmed that it’ll be integrating a ChatGPT-like chatbot into the app later this year that uses deep learning AI. It’ll be able to “shape your preferences, give feedback on your workouts, and get questions answered on everything from nutrition to form,” said Justin Bingham, chief technical officer of FitnessAI. 

And while you might be able to converse with a future version of FitnessAI, it still might not motivate you as much when answering to a real person. 

Read more: Could Your Future Workout Buddy Be a Robot? 

Comparing my AI trainer to a real one 

When I started out with the Future personal training app, I was connected to Brett Carroll, a trainer based out of Florida. He asked if I wanted to continue with my five-by-five workout routine (five exercises at five reps near my max weight limit). I deferred to him, saying I was doing this routine for a while and not seeing the results I wanted. He instead opted for a block routine, breaking up my workouts between lower and upper body days. And the workouts wouldn’t solely focus on weight training. He integrated lots of stretching and movement exercises, which I’d been neglecting. 

The workouts created by Carroll were some of the most difficult gym sessions I’ve ever had.

I think of Future as asynchronous personal training. It’s not that Carroll is there on a video call the entire time I’m at the gym. He creates workouts for me and the app autoplays videos of how to do specific exercises and stretches. It also has chimes for start and stop times and gives recommendations on how to properly do exercises. I can also go into the app and input variations I made to weights. Carroll can then adjust accordingly. 

My first upper-body day was tough, forcing me to circuit through multiple muscle groups with short periods of rest. At the end of the hour, my shirt was noticeably drenched. 

Man doing bench press

An upper-body day focused on progressive overload.

Imad Khan

After a one-day break, I went back to the gym to do a lower-body workout, focusing on funny-looking lateral duck walks and cossack squats, where you lean your body in a lunge from side to side. After some bodyweight squats, I grabbed two 30-pound dumbbells and did Bulgarian split squats, a balancing-squat variation that focuses on one leg at a time. From there, I had to immediately jump to “bound with stick,” a movement where I jumped laterally from one leg to the other. The combination of lifting plus explosive movements left me ragged. And that was only the first complex. 

The second complex of barbell single-leg deadlifts (while holding a 45-pound barbell) and 30-pound dumbbell step-ups was making me question my years at the gym. Had I really made any progress at all? These relatively simple exercises at lower weights were draining me. Typically, I can do a normal barbell squat with 275-pounds of weight on my shoulders. While I was only supposed to take 30-second rests, I found myself needing more than a minute to recompose myself. 

Exercises with Future exercise app showing man doing barbell bench press

Future app 

Screenshot/CNET

That first leg day left me feeling so depleted that I skipped the following day’s workout. Thankfully, Carroll was able to adjust my routine accordingly to help me recover.

Speaking of recovery, Carroll recommended supplements. I could also talk to him about my diet and macro goals. This is something companies might restrict with AI fitness apps due to liability concerns. When asking ChatGPT what supplements I should take, it recommended whey protein, creatine, fish oil pills and multivitamins. Its macro recommendations seemed incorrect based on my body type and weight-loss goals — it proposed 360 to 540 grams of carbs, as opposed to the recommended 170 to 200 grams of carbs from online macro calculators.

At $199 per month, Future is far more expensive than FitnessAI, and even pricier than a lot of gym memberships. (For the purposes of this article, Future granted me a trial membership.) But given that personal training sessions at a gym can vary between $50 to $150 or more each, a month of fully curated workouts via Future isn’t a bad deal. 

Human over machine 

It’s early days for AI in the health and wellness space. Eventually, FitnessAI, and likely other fitness apps, will integrate AI chatbots so you can ask it questions and have it respond. Until then, you’ll be relying on apps that, at best, can crunch some data and put together a decent routine. 

I had no particular qualms when using FitnessAI. If you’re on a tighter budget, I’d recommend it for those who already have gym experience and would like to log their workouts on their phone. (Granted, plenty of apps do this for free.) There are numerous workouts in the app so even the most polished gym goers will find something new to try. With FitnessAI, it’s on you to maximize your workouts, and you’ll need to reach out to outside sources about diet, supplements, stretches and cardio. 

Bottom line: AI-powered software and algorithms can’t generate a workout routine for you based on your thoughts and desires — and it might not be precise in accommodating injuries. 

But everyone is different. Some may prefer logging numbers into an app and not having to deal with the pressure of reporting back to a person. In my case, getting real feedback from a real trainer is what kept me motivated, a sobering realization of my own social conditioning. Either way, when it comes to gains at the gym, I’m opting to trust a human over a machine.


Editors’ note: CNET is using an AI engine to help create some stories. For more, see this post.

Products You May Like

Articles You May Like

A Scrapped Overwatch Netflix Show, A Proposed Call Of Duty RTS, And Four More Details From An Upcoming Book On Blizzard
5 ‘dumbphones’ that can still run WhatsApp
Amazon October Prime Day: 80+ Early Deals Across Tech, Home Goods, TVs, Appliances and More
Our 8-Part Plan For Alex Pereira To Become A Combat Sports GOAT Begins Tonight
Best Prime Day Apple Deals: Score AirPods, iPads, MacBooks and More at Major Discounts

Leave a Reply

Your email address will not be published. Required fields are marked *