How to Use AI to Improve Your Training | Dr. Andy Galpin
In a Nutshell
AI excels at organizing data, spotting patterns, and quickly editing training programs at scale, but cannot reliably judge quality or deliver context-specific solutions because it relies on synthetic data and general population averages rather than expert benchmarks. Users should avoid asking AI to assess whether results are “good” or “bad,” since it lacks access to high-performing reference populations and will default to broad, non-expert answers. Human coaches remain in rising demand precisely because real expertise is required once training or nutrition plans become complex or individualized.
These notes were generated by AI and may contain inaccuracies.
Artificial intelligence is constantly growing and will only get better. Up to the day this video was recorded, the personal stance is that AI is a tremendous tool. If it can help you organize, find patterns, or sift through massive datasets of your own information, this is good and productive.
AI has difficulty helping you evaluate information. If you took your blood test results, sleep tracking data, or training program and put them into artificial intelligence and asked it to provide you with a summary of your training volume in the past month, that's wonderful. You could ask it to draw up a plan, or write a new program for you, such as replacing every shoulder press exercise with an alternative on the machine.
It's really useful, especially if you're a trainer and you have 15 or 20 programs, and you suddenly decide you don't want to do trap bar deadlifts anymore. You can simply ask AI to replace every "Trap Bar Deadlift" exercise with dumbbell deadlift exercises. This would have taken hours and hours in the past, and now you can do it in the blink of an eye. All of those things are great uses for artificial intelligence.
The weakness is when you upload your data and ask AI to make judgments. Is this okay? Is this bad? Is this horrific? There are many reasons for this. Most AI engines or large language models operate in one of two ways. First, AI looks for the loudest sound in the room, and that doesn't necessarily mean expertise. The voices of those with the most experience and who speak the most truthfully are not always amplified.
Do not allow AI to judge whether something is good, bad, or terrible. AI simply does not have the data to do that. The second general way AI operates is through what is called synthetic data, meaning that it does not provide you with information based on any real data. This has both advantages and disadvantages, but in the field of health and performance, this is a disadvantage.
When using synthetic data, you are not evaluated and compared to the person you really want to be compared to. You are evaluated in comparison to the general public, who are usually people who are not in good health or are not in the same health as you or the level of health you aspire to. When asked to give the optimal range, not the range of the sick, AI does not do that. All AI does is confine values.
If you asked AI: "What is a good VO2 max level for my age?" and you are 35 years old, it will answer that a good VO2 max range is between 40 and 45. When asked for something wonderful, it would say the great range is between 45 and 50. This may be enough sometimes, but if it's more complicated, AI doesn't have a database of high-performing people it can really rely on.
Another area where AI faces significant difficulty is with context-dependent solutions. This is where experience begins to become truly important. If you simply wanted to use artificial intelligence and said to it, "Give me a basic training program" or "Give me a basic nutritional program," for many people, most of them, this is probably a good thing.
If you took the average person who doesn't listen to this podcast, and who you probably don't spend time with, and said "put this into something and just give me the basics of nutrition, the basics of exercise, give me a training program for 3 days a week," AI can do that well. The difficulty lies in saying: "Okay, give me that, but add this to it. Add this to it. Put that in it. Include this part in it."
Because AI will only rely on the loudest sound in the room, it will keep things very general and basic and will lack experience. If you don't need the expertise, it's perfectly acceptable to use AI. But if you need it, be warned against using it because AI lacks that real rating.
It has become very clear in the last two years that the demand for human expertise in health and performance is increasing. This sector was severely impacted during the COVID-19 pandemic. Everyone stayed away from hiring coaches, mentors and the like because they were doing everything at home. Then artificial intelligence agents began to appear, and now things are moving in the opposite direction.
People are using coaches more and more because they want to know someone with real experience, and because they desire that personal interaction. Use AI as a tool for organization, for getting things in order, or for getting paperwork done, but be very careful when using it for evaluation and providing solutions.
Keep Andy Galpin in your library
Save the videos and channels worth coming back to, and find them again in one place.
