Applications of Generative AI in the Fitness Industry
How Fitness companies can capitalize on emerging AI capabilities
Overview
We will remember 2023 as an inflection point in technological innovation and when Generative AI entered the general public's lexicon. For the past six months, a bright spotlight has highlighted this developing technology driven by media interest in attention to the sheer amount of investment capital flowing to this vertical. I can say from experience of looking at countless pitch decks that the term Generative AI may have supplanted "democratization," "platform," "crypto," and "web3" as the vital buzzword that every early-stage founder is looking to squeeze into their marketing materials. And although I do think the term is misappropriated at times, and some of the applications being created with it are gimmicky at best, I truly believe that this will be one of the defining technological advancements of my generation.
Generative AI will impact every industry in one way or another, the Fitness Industry included. I wanted to take this opportunity to explore some potential applications of Generative AI within the Fitness Industry and note how some companies are already integrating AI solutions into their products. First, let me start with a very non-technical high-level overview of what Generative AI is and how we got to where we are today.
How We Got Here
Artificial Intelligence (AI) is the simulation of human intelligence in machines programmed to think, learn, and perform tasks. Until recently, its capabilities consisted of performing analytical functions such as evaluating large data sets to produce insight at a speed and depth much more efficiently than a human would be capable of. Generative AI is a relatively new subset of AI and entails producing original content rather than just analyzing existing data, replicating human creativity. This breakthrough has significant implications for various knowledge workers, given it can create unique text, audio, video, images, and other forms of content that historically needed human oversight.
The idea of Generative AI has been around for a while but gained significant developmental momentum in 2015 when Google Research published a paper titled "Attention is All You Need." This paper started the arms race for all major tech companies to work behind the scenes on developing larger and more capable models that could analyze language inputs and produce accurate outputs through different modalities. Datasets and compute power grew at exponential rates that allowed the development of more sophisticated Large Language Models (LLMs) such as OpenAI's GPT-3 and GPT-4, Google's LaMDA, and Meta's LLaMA 2, among many others.
OpenAI, the company behind the GPT-3 and GPT4 LLM, helped bring the idea of Generative AI into the mainstream when it introduced text-based Chatbot ChatGPT and its image generation application DALL-E, free for public use. Mass adoption of its applications, along with a reported $10B investment from Microsoft in early 2023, lifted the floodgates on companies seeking to build capabilities on top of the work that OpenAI started and introduce Generative AI into existing or new products and services.
Existing Use of AI in Fitness
Although Generative AI is a relatively nascent technology, many companies already use more analytical AI within the Fitness Industry. This technology is helping power real-time insight into our bodies that was previously unimaginable. Below are a few examples:
January AI: January AI collects biometric data through a continuous glucose monitoring (CGM) device and utilizes AI to synthesize the data to provide real-time analysis of a user's biological response to various foods and activities.
Tempo: Maker of a Smart Home Gym that includes weights and barbells, Tempo uses embedded sensors and biometric inputs that allow the system to track speed, form, and other metrics. Integrated AI can then use that data to provide actionable insight to the user and help design workout routines.
Zone7: Zone7 uses motion capture technology powered by AI to analyze injury and sports performance data. Their systems can detect potential red flags in an athlete that can that lead to injury prevention.
Asensei: Asensei is an AI-powered motion capture tool that allows fitness professionals to analyze client's form and progression more effectively. The company is currently piloting its room-scale product that would notify a fitness class instructor when a participant needs particular attention and help the instructor oversee larger classes.
Whoop: Wearable device pioneer Whoop uses AI to synthesize the vast amount of data it collects through the wrist-worn strap and provide insight into one's strain, recovery, and readiness to respond to fitness activity.
Applications of GenAI in Fitness
Over the next 2-3 years, we will see many Fitness companies shifting to integrate Generative AI into their existing product or service offering, along with new entrants, using the technology as the cornerstone of their value proposition. Below are some of the areas ripe for disruption:
Personal Training
Text-based Generative AI can already replicate a large portion of a fitness trainer's job description, designing comprehensive workout routines. Suppose you add motion capture and biomechanical analysis technology that can provide accurate feedback. In that case, technology has essentially replaced the role of a personal trainer. Many companies are already building in this space, including Aaptiv, Fitbod, Fitness AI, and Freeletics.
There has been pushback from the personal training community on replacing their service with AI saying that the technology could never accurately assess form and provide the human touch needed for practical gains in strength and conditioning. This thought is probably true for the most educated and highest-cost fitness trainers. On the other hand, anecdotally, I've walked into several big box gyms and watched fitness trainers scroll their phones while counting reps, not providing much value beyond that. Using Generative AI will ultimately be a net positive when measuring gym engagement and a generally healthier society. Applications can soon democratize personal training, making it cheaper and more accessible for anyone to have access to fitness knowledge.
One final point on this topic is that GenAI in personal training doesn't need to be entirely disruptive and replace trainers. The best trainers will use tools to streamline their workout design process and free up more time to scale their services to more clients.
Nutrition Coaching
Going hand-in-hand with personal training, Generative AI can provide many of the services a traditional nutrition coach offers. Right now, you can punch in a macro profile and list the foods in your fridge, and ChatGPT can produce a meal plan that adheres to your guidelines.
Nutrition advice will also be a sensitive area for Generative AI disruption, given the health implications and the emotional connection people feel towards food and diets. Because of this, I foresee the most likely scenario as existing coaches using Generative AI applications as a tool to reduce busy work and scale to a broader audience. This application once again has the positive possibility of bringing nutritional guidance to people who otherwise would not have been searching for this information or not be able to afford it.
Connected Fitness
Connected Fitness content providers and device manufacturers will utilize Generative AI to expand the size of their content library, which is a crucial point of differentiation within the digital fitness industry. Connected Fitness companies could create custom and personalized workout content in real-time using the large datasets these firms have already collected and continue to build, along with user inputs such as preferences and biometric data. One will no longer need to scroll through an endless Peloton library to find a 30 min, Alex Troussant-led, hill-focused, Hip-Hop workout. Instead, they can input their search criteria, and if it doesn't already exist in the library, GenAI could create this workout in real-time.
Fitness Media and Content Creation
Perhaps the most apparent application of Generative AI within fitness is the creation of Fitness Media and Content, particularly in written form. Many major media companies have already openly stated they are using Generative AI to augment the amount of content they create (Buzzfeed). For companies that rely on creating a high volume of content consistently, such as newsletter writers, Generative AI can be a game-changer for efficiency.
I predict that much of the short-form content we know today will be automated soon. Human work will be more focused on long-form content such as investigative journalism and providing editorial oversight over machines' content.
Prescriptive Wearable Devices
Mass adoption of wearable devices and the volume of biometric data they can collect has increased precipitously over the last decade. Most wearable companies have also integrated analytical AI into their product to quickly synthesize the collected data and provide insight into our bodies. Integration of Generative AI to take this data and create more prescriptive actions is the next stage of wearable development. My personal opinion, which I discussed in a previous post (What’s Next for Wearable Devices), is that the proliferation of wearables has not necessarily led to a corresponding improvement in health outcomes in part due to the lack of prescriptive capabilities of most devices. People collect their data but don't know what to do with it.
Kristen Holmes, VP of Performance at Whoop, echoes this sentiment in an interview with Wareable, in which she gave an example and discussed how one could proactively tell their device that they'll be drinking alcohol that night, and it responding with a plan of action as far as what to eat and what to do to minimize the effect on your body. Wearable companies that can use technology to drive more actionable insight will create stickier products that lead to better health outcomes.
Meditation and Breathwork
The recent COVID-19 pandemic helped accelerate the growth of meditation and breathwork companies such as Headspace, Calm, and Brthwrk. Like connected fitness devices, these companies compete in part based on the size of their content library and ability to provide personalized and compelling sessions to users that will lead to stickiness. Using Generative AI to personalize the content on the fly can result in people adhering to the practice for longer, increasing the likelihood of renewal. One can enter a prompt like, "I'd like a 12-minute meditation focusing on Wim Hof style breathwork. I want there to be light rain sounds in the background, and I want the narrator to sound like David Attenborough." If an application can deliver real-time content creation with this level of specificity, I believe it will lead to more engaged and satisfied users. Additionally, the cost profile of these companies will change dramatically if they no longer have the same expenditures for content creation.
Gaming and Virtual Worlds
Although our transition into the metaverse is going slower than Mark Zuckerburg may have initially expected, there are existing use cases for virtual worlds that can be enhanced and created with Generative AI. Gaming is the most straightforward scenario. Whether through VR applications like FitXR or more traditional connected fitness devices like Zwift, people are becoming more comfortable working to achieve fitness goals in virtual worlds. User-inputted prompts that tell the application what they would like their virtual world to look like, who'd be joining them, and customization of the workout within the virtual world can be transformational for fitness games and VR applications.
Closing Thoughts
One final question I've been spending some time with not specific to the Fitness Industry is who else will capture the value created from the Generative AI innovation wave? A stakeholder that I don't think is often discussed is the owners of the underlying data. Generative AI models are only as good as the data used to train them. The data quality becomes even more critical as developers build applications for specific verticals, like health and fitness. The wave of innovation has moved faster than regulators can keep up with. Still, I think it's inevitable that restrictions will be put in place that prevent models from having complete reign to use whatever data they can find. Owners of data and IP will open up new revenue streams to sell or license their data to applications that need it to build effective models. Whether or not using the data themselves, opening the door to selling it will be a tailwind to the numerous data harvesting machines we interact with daily (wearable devices, social media, healthcare providers, etc.)
As with any disruptive technology, there will inevitably be some negative externalities. With AI more broadly, I think many jobs and responsibilities will become obsolete, undoubtedly leading to economic pain for some in the short run. But ultimately, I believe Generative AI applications will be a net positive for the fitness industry. The most enterprising companies and entrepreneurs will evaluate ways to integrate this technology to enhance their existing products or service. More importantly, I think it will help bring fitness and health solutions to more people at a cheaper cost, which is what we ultimately need to drive behavioral change in our society.