Will A.I. Replace Coaches? - Starting Strength Radio Previews

[David A. Rowe]

The crux of A.I. functioning correctly is getting it to correctly perceive reality. That has always been the most difficult problem to solve, and I'm not sure if we're at the point where it could be done more economically than just hiring a coach.

[Brodie Butland]

There’s another fundamental reason why AI can’t replace coaches: identifying the flaw is only part of the battle of coaching—the other, more important aspect is providing cues to correct the lifter. I can have five lifters with nearly identical form faults but have to use different cues for each of them because of their individual psychology, training history, sensory perception preferences, etc. I’ve had to use some cues for lifters that would never be in an AI cue database. In short, AI can’t get into someone’s head in a manner needed for proper coaching.

And that’s assuming only one form flaw that needs corrected. But in reality, lifters may have several form flaws, and coaching becomes an exercise in prioritization. If someone consistently isn’t using hip drive in a squat and caves their knees slightly in the hole on a rep, in most cases IT IS WRONG to cue knees out because the far bigger problem needs to be focused on and addressed first. (There’s a free pro-tip for those who want to be evaluated at seminars.). But how realistic is it that AI will be able to make those determinations, quickly and in real time, using cues that will work for a particular person?

[Chiaki Nitya]

Every once in a while I get into a bit of a solipsistic panic at the thought of ubiquitous AI in human society, and then I remember that humans, in general, are pretty fucking stupid, and therefore any human-created AI is going to be inherently limited by our species own intellectual shortcomings.