Have you ever wondered where your coach has just magic-ed that session from? Is it part of a higher plan, or was it conceived on the spot? Maybe they asked an online chatbot to design it?
ChatGPT has exploded into our lives and our social media streams. Apparently, it will change everything, render student essay writing pointless, replace most of our jobs and much more. Undoubtedly, it will have a profound effect on many facets of our existence. Which facets and what impact will become clearer over the next few years.
Over the last few decades, we’ve seen a proliferation of data collection, analysis and use in Sports. My interests in this intersection between sport and technology are twofold; on a personal level can it benefit performance, and, more generally can it democratise access to the best training advice?
Chat GPT ingests massive amounts of content from all corners of the internet and synthesises the information into natural language conversational responses to queries. This information includes training designed by world experts, academic literature and blogs of professional athletes. Vastly more information than a human could ingest in multiple lifetimes. Now that a chatbot can integrate most of the information on the internet and pass exams to become a Barrister, Doctor and even Sommelier – surely it can come up with an effective Triathlon training program?
I thought I’d see for myself. I entered – what I thought are – a few key parameters and it generated a program. On the surface, it’s pretty impressive. In a few seconds, a training program appears on the screen. It’s periodised, progressive and structured with a combination of easy and more intense training. Although, you could be forgiven for finding the program rather generic, lacking in detail, and not suited to your specific goals or training style. It’s difficult to argue it offers any value over finding a published training program on Google.
ChatGPT is only one artificial intelligence-based training program option. It misses, what I think, are the real benefits of using AI to prescribe training. Most of us log a huge amount of data; power, heart rate and velocity; data from wearables; HRV, resting heart rate and breathing rate. Analysing this data, comparing it to trends and using the results to suggest adaptions to future training is something AI can, and already is, doing successfully.
I believe Athletica is already driving the optimisation of training prescription successfully.
Athletica takes this objective data-driven approach, to both create and dynamically update, a personalised plan based on a user’s ability, fitness and race goals. The AI runs all these data points through algorithms based on widely accepted sport science principles to create an effective and tailored plan to help the user make progress towards improving their race performance potential.
Another important aspect of training that ChatGPT cannot yet do relates to the all-important issue of solving for context. We all know how life gets in the way of the perfect plan. Unfortunately, I’m a bit too familiar with this issue. What do I do if I’m short on time? What do I do when the legs just don’t work? And what do I do if I feel a niggle coming on? Context.
This is where Athletica’s Workout Wizard AI steps in. It helps the user solve for their context using appropriate solutions based again on their wearable-derived fitness level, ability and goals, to make it possible for an athlete to change workouts appropriately on the fly. Ultimately, we are human beings. We’re fallible. What if that workout just isn’t doable today? Workout Wizard AI suggests other possibilities that elicit the same physiological gains but may be more psychologically palatable.
Athletica also incorporates natural language analysis of a user’s subjective training session feedback that gets fed back into the training plan for future sessions. This is an area where chatbots like Chat GPT could help Athletica and many athlete coaching platforms. Combining Athletica’s objective data alongside the athlete’s contextual history. Using a conversational chatbot approach to user interaction means the next generation of AI training programs could better address the needs of the un-coached (or hybrid-coached) athlete.
Data scientists talk about Garbage in Garbage out. Dodgy initial parameters will result in a dodgy output. It normally refers to the quality of the input data; is it measuring what we think it’s measuring? But, I think, in the case of this chatbot experiment the issue is the interface between the subject – prone to biases – the human being (me) and the AI. Why did I input those variables? Would other variables elicit a better output? How do I know what these optimal variables are?
This is why Athletica seamlessly integrates human coaches. The Athletica coach platform does the grunt work of designing, scheduling and updating training plans, allowing the coach to focus on the things that AI cannot deliver: emotional, strategic and psychological support, as well as an important sense-check when the AI fails, as it sometimes does. This, for me, is where the future of coaching lies. The ultimate combination of machine processing, human insight and personal relationships. Combining this approach with technical experts, scientific support and a sense of community is our aim at Brownlee Fitness.
Chat GPT isn’t going to disrupt the deep science AI training apps just yet. It’s also not going to be a worthy competitor to great personal coaches. It is incumbent on all services in the athlete training space to understand what is happening with AI and use the technology to enhance their services.
