Wow, Artificial Intelligence. It’s everywhere. It seems like 50% of the people I know on LinkedIn are now working on an AI startup. Everyone is writing about it … so please, hold my protein shake.
Never in my living memory has so much excitement / hype / hysteria been attached to a technology in such a short amount of time. If you believe Battlestar Galactica and Terminator, we are witnessing the beginning of the end of mankind, and there is genuine fear among many AI experts. It’s actually hard to think of AI movies that end on a positive note - OK, maybe Stealth, where the AI sacrifices itself to save the pilot.
But will AI revolutionize fitness? I don’t think so. At least not in ways you might think, not for a long time. And I say that as someone who’s run software companies for over 20 years.
What is AI and what’s it good at?
Today, the buzz is about generative AI, with snappily named poster children like ChatGPT. The GPT part stands for “Generative Pre-trained Transformer“ - and now you see why they shortened that to the acronym. The other less famous acronym is Large Learning Models (LLMs), which allow an AI to learn by itself (unsupervised) based on a large set of input data, often human dialog (like conversations). This ‘unsupervised learning’ has been a game changer in AI.
People have predicted robots and AI that can do everything since those cheesy 1960s videos - see the optimism in this video. Every prediction has turned out to be wildly optimistic. But as per one of the suddenly famous voices as the Godfather of AI, Geoffrey Hinton, the big change has been big computers getting much lower cost. ChatGPT itself runs on Microsoft’s servers at Azure, and it uses hundreds of thousands of processor cores - imagine the football fields of servers in The Matrix.
When I was studying Chemistry at Durham University, back in the last century, one of my friends and the guy who was almost always top of the class in every exam made an interesting off the cuff comment to me. He said that you could get a monkey to get 100% in all the tests if it could just regurgitate the right part of the professor’s notes for each question and then use it appropriately. Now he was being modest here, he was a very smart guy, but he also had an incredible (near photographic) memory. But today, generative AI excels in questions, tests and exams, where the goal is to take a lot of information as input and present it coherently in response, potentially also applying it in ways that have also been ‘taught’. It can ‘discover’ new techniques, but it’s still quite narrow and based on the training data.
As far as I’m aware, ChatGPT and other generative AI tools struggle to have a predictable and intelligent editorial line. Now if you create your own giant learning model based on a ton of data in your particular paradigm or thesis, then sure it can spit out more of that, but if I wanted ChatGPT to write Tribe’s blog, it would struggle. Could it spit out a ton of generic essays that would help drive Search Engine Optimization (SEO) juice? Sure it could do that. But would it bring insight based on the specific ways we at Tribe interpret the world? No, not today.
Many people are saying that AI is like a calculator that can help you do your job faster and better. A lot of people under the age of 30 can barely do math in their heads today. But if you are the computer or calculator, like in the movie Hidden Figures, then it’s time to start reinventing yourself or you’ll be out of a job.
According a recent article in The Guardian, “AI could replace the equivalent of 85m jobs worldwide by 2025 and more than 300m in the long term.” Goldman Sachs made a similar prediction - that AI could automate 25% of the entire labor market.
What jobs could be revolutionized by AI? If you’re a taxi or Uber driver, or a truck driver, your days in that job are probably numbered. AI will revolutionize the transport industry - young kids today may never learn to drive cars. Software developers, attorneys, accountants, graphic designers, journalists, media analysts, stock traders, ... the list of endangered species goes on, though in reality I think many of these will end up using AI for greater efficiency rather than being replaced.
In a recent study, 80% of people who tried ChatGPT for therapy advice thought it was an effective alternative to human therapy sessions. The upside was a lack of inhibitions to be able to ask anything without judgement, and for sure much of the advice was good, but AI doesn’t have true empathy - it fakes it by mimicking empathy from humans expressing empathy in its training data. As a fun aside, John Oliver recently took the rip out of using an AI powered robot for training human therapists - which does also seem to have some irony.
So how is AI helping in fitness today?
- Movement Analysis: The high profile use of AI in fitness has been form analysis and rep counting. Lots of companies have worked on this, including Peloton, Tempo, Fiture, Altis, Asensei, Kemtai, Exer, … it’s a long list.
- Workout Customization: The CAROL bike varies resistance based on AI, Tonal does something similar with digital weight, FitnessAI uses AI to design personalized workouts, Orangetheory is using AI to determine each member’s max heart rate.
- Recovery Monitoring: Whoop is using AI to analyze wearable data collected all day long to measure recovery, Athlytic is doing a similar thing with data from Apple Watches.
- Managing Ads: Orangetheory has seen a significant reduction in its cost per lead by using AI to run its advertising, this is a mainstream marketing area in which AI will excel.
- Coach Co-pilots: Verb is using AI to assist and scale coaches and personal trainers, allowing automated data gathering and human-like responses mixed in with real coach responses, over text messages. FitGrid is doing something similar in using AI to help write personalized email responses and campaigns to a studio’s members. Mindbody as the incumbent is also trying to use AI to automate member communications. Every mainstream CRM is figuring out integrating AI.
- Tribe: We’ve taken a few small steps in AI so far. For example, we us an AI service from Amazon to automatically find cover and thumbnail pics from video recordings (versus coaches having to do that manually), and we’re working to integrate a recommendation engine based on AI - similar to the Netflix “you might also like” options. We have a few more ideas and tricks up our sleeves for the future.
I do think that for Pro sports players, where a 5% change in performance is a game changer, AI will likely have a big impact. There’s no shortage of trainers and motivation in this community, it’s more about finding ways to performance tune the engine to get every last ounce of horse power out of it.
But will AI revolutionize the broader fitness experience? I don’t think so. Not in the same sense that it will revolutionize transportation.
AI will bring more personalization, and it will also continue to amp up the gamification aspects of fitness with different types of progress measurements and achievements. I think that the combination of gamification and AI form analysis in fitness has been relatively unimaginative so far, perhaps in part because it’s hard and expensive.
For gym and studio back offices, just as with many other businesses, AI will allow automation of a lot of processes, allowing staff to focus on tasks that AI can’t do.
In fitness, a lot of people want (or think they want) six pack abs. I have the perfect formula for that - eat less, eat more protein, do more cardio, do more ab exercises, hydrate, reduce stress in your life and get more sleep. See - easy isn’t it? I just typed “how to get six pack abs” into Bing chat, and it said pretty much exactly the same thing. Magic. If only AI could do all that for me, I would hand over great quantities of US dollars.
What I lack isn’t knowledge, it’s motivation. Where can I find motivation? Well for most people, it’s a mix of gamification and community. Fitness is hard, gamification as a solo activity is much less gratifying than when you do it with other people - colleagues, friends at Crossfit, a personal trainer - just regular people.
As Jared Cluff, the founder of Caliber app said recently, “For most people, and certainly the vast majority of clients we see at Caliber, the value of a trainer extends well beyond the tactical role they play, and is much more about the motivational role they play, which won't be commoditized by AI anytime soon.”
So in a future in which we all sit around and AI does everything for us, will we still need fitness trainers? Well, WALL-E, would certainly say yes, unless we all want to die of heart disease in our 40s.
Now if we get AI we care about and want to impress, like in the movie Her with Joaquin Phoenix or in the TV show Humans (where the son of the creator falls in love with the created), well then that might provide genuine motivation similar to a personal trainer. But we’re still a long, long way away from that.