2 fit-tech trends are leading us towards interactive AI fitness trainers

Dear Readers, 

At last, summertime is upon us, which means more people are joining sports teams, hitting nature trails, and dusting off their activewear for the annual last-minute “beach bod” scramble. But fitness these days involves a lot more than polyester, Pilates, and passes. In this week’s newsletter, we navigate the nexus of fitness and tech, exploring how AI startups are transforming fitness and how embodied AI agents fit into this innovative industry.

Fitness and Technology are two of the fastest growing industries in the world. Fitness reeled in an estimated $94 billion last year at an annual growth rate of 6.1%. Technology permeates the world, with the AI industry reigning as one of the fastest growing (38.8% CAGR). The intersection of these two promising industries, a.k.a. fit-tech, has created a whole new world of digital fitness, where innovative products like Peloton and software applications like Strava flourish. Valued at $7 billion, Digital Fitness is expected to reach an estimated $27.4 billion by 2022 at a forecasted CAGR of 32.6% from 2017 to 2022, growing at 10 times faster than the traditional fitness club industry. Fit-tech is a match made in heaven.

Starting with the rise of smartphone apps in the last decade, such as Runkeeper, Runtastic, and MyFitnessPal, fit-tech has transformed how we keep fit. After jogging, we check running apps for performance stats (our own and those of our friends). At lunch, we track our calorie intake. During a work break, a colleague might be talking enthusiastically about her last HIIT session. At an evening party, a pal shows off his newest Apple watch (and how close he is to closing those rings!). And, at the end of this day full of counting, crunching, craving, and checking, we even track our sleep.

The fitness industry has seen much disruption in the past: the subscription model, most notably, is making fitness more flexible and diverse. Meanwhile, technology continues to improve both the hardware and software of fitness tools, democratizing previously expensive premium services. With the rise of AI, fitness is entering a whole new era of accessibility. At Embodied AI, we see 2 trends that are converging towards a new generation of AI-powered digital trainers that can interact with exercisers in real time.

People are taking fitness home - goodbye weight room, hello living room! An emphasis on functional fitness is making the gym less of a necessity and fitness apps a more viable option for those of us who can’t make it to the gym all the time. Fitness app usage increased 330% from 2014-2017. A survey by the user insights platform Alpha shows that 54% of Americans who work out at least once a month are interested in buying an at-home fitness system. Take a glance at some of the most popular at-home fitness companies:

These companies’ at-home products and services offer convenience, content from top trainers, and connected experiences. Essentially, they are moving one-on-one fitness training from a premium service to a commodity.

Until now, most fitness apps, such as the popular Freeletics, have been content-heavy but lacking in interactivity. What apps cannot provide is real-time, corrective feedback that a human trainer can offer. However, some startups are building more sophisticated AI to achieve this by leveraging wearable devices and computer vision.

Vi Trainer, for example, comes with biosensing headphones to collect the user’s physiological data and adjusts their form in real time. The app provides motivational feedback during workouts, reacts to basic voice commands, and has a chatbot function for a limited interaction. Sword Health and Hinge Health are two companies that pair wearable motion trackers with a tablet app to offer digitized physical therapy and real-time feedback from the comfort your home.

Other startups develop virtual trainers that can guide and train users with computer vision techniques, such as pose estimation, that map key points on the user’s body. Kaia Health’s Motion coach is an app that involves positioning a camera towards the user in order to offer real-time corrective feedback during therapy sessions. Vay Sports, a Zurich-based startup, plans to launch a exercise-specific app in alpha similar to Kaia.

Form correction and physical therapy, both of which were once only available by appointment at the fitness club or sports clinic, will be digitized and democratized by technology.

With at-home products commoditizing one-on-one training and computer vision democratizing access to expertise, the stage is set for deep learning, the most recent wave of AI advancement, to combine the two. Imagine the one-on-one connectedness, empowering motivation, and helpful corrective guidance of a personal trainer embodied in an on-demand, non-judgmental, yet human-like avatar. That is TwentyBN’s Millie.

TwentyBN, the AI startup where your editors work, has built Millie Fit, a computer vision AI-powered fitness coach housed in a life-sized kiosk. Combining video understanding and natural language understanding, Millie offers everything you’d expect from a traditional trainer. Millie leads you through a variety of workouts, offers real-time corrective feedback on your form, and pushes you to hit those last 5 reps. But that is not all what Millie has to offer.

Behind Millie’s capabilities is TwentyBN’s globally distributed data platform that constantly collects high-quality, annotated video data on fitness-related actions from crowdworkers. The data serves as fitness knowledge for the deep learning model inside Millie’s brain. With no need for pose estimation or skeleton mapping over a human body, Millie tracks a wide range of exercises, including those which were previously untrackable by traditional computer vision techniques. Currently, Millie understands over 10 different types of stand-up exercises and over 20 High-Intensity Interval Training (HIIT) exercises. Her expertise in fitness will soon extend to workouts in boxing, yoga, Pilates, and more.

A typical one-on-one fitness session with Millie includes a warmup, workout (such as a HIIT session), and cool-down. Millie demonstrates each exercise and provides tips on proper form. Then, she guides the user through the workout, pushing them with encouragement and giving corrective tips where necessary. Users can track their real-time performance on the screen, check out their stats post-workout, and join the leaderboard with their friends. Very soon, Millie will offer exercisers custom workouts and corrective form sessions matched to their performance data.

As tech-driven fitness training continues to rise, the wall barring access to personalized fitness and health expertise will break down, giving way to the commoditization of one-on-one training and the democratization of previously inaccessible expertise. At the intersection of these trends, intelligent, seeing, and interactive AI avatars are spearheading this new era of fit-tech.

Subscribe here to Embodied AI! Make sure to forward Embodied AI to your friends and colleagues and tweet about us. Cheers!

This issue of Embodied AI is written by Nahua and Isaac, with contribution from David, Chris, and Will.