Hi, xert is awesome at giving so much information about training performance, such as XSS, Peak, High, MPA, and so many more included in the latest release. Unless you are an expert, these are very difficult to understand in a more natural way. Athletica.AI is one of the best I found to give feedback on completed workouts explaining in simple terms how the workout contributes to your performance goals. Are there any plans to use LLMs to digest all this data to give users feedback in a natural language way on completed workouts and even how our fitness is improving, what to focus on based on recovery, etc.?
2 Likes
I’ll be curious to see if the trend towards LLM-based coaching and feedback sticks. I think it’s definitely interesting and new and as you point out, can provide an interpretation of the data in terms that are easily to understand and follow. I suspect it may end up being repetitive and too generic and the initial spike in interest might fizzle. We’ll see.
4 Likes
Agree, think we’re approaching the peak of the Hype curve:
So I tried a bunch of Xert related queries with ChatGPT and it generally was quite good. (I had tried previously and it was woefully lacking). I asked about Fitness Signatures, MPA, Smart Workouts and it generally was very good in the responses.
I then asked about Race AI, and it provided an answer that was not really aligned with what Race AI truly is. We just announced it so I can expect it to know. Then I asked about Forecast AI and it gave a pretty good overall answer. I was quite impressed.
Overall, I think the technology is pretty amazing and will help a lot of people learn about stuff more readily. It can also help those that write content be more prolific and improve readability of their documentation. I had put some of our recent blogs through ChatGPT and the rewording was quite good but I found it to have a style that I didn’t feel represented how it should read so it wasn’t used.
The answer to the Race AI question though is a bit concerning because it answered in a very authoritative way but in fact the answer wasn’t right. I find this aspect of these new Gen AI models a bit troublesome since you may not have sufficient knowledge or wherewithall to understand that it had done this on the topic you asked about. The more specialized the topic, the more likely it’ll get it wrong in fact and it’s these specialized ones are often where someone might genuinely be interested in getting more information.
In the end, we’re on that cutting edge of new models and concepts and I don’t expect ChatGPT to keep up with everything immediately. We’ll still need to write new content and explain new concepts that are not yet mainstream. GenAI will then use these to learn from. (Someone has to create new content!!). We’ll probably move towards allowing users to access a bot that gets trained on our content and endurance training content at some point. I don’t know if it’ll improve anyone’s ability to reach their fitness goals but it might make the system a bit more fun to use.
1 Like
It can indeed be good to get a quick overview over a domain you are no expert in.
But always check the references and if there are none or they don’t match the bot’s output, step back.
1 Like