I would claim that this guy essentially goodharted himself into passing this online test. If you train for a month using always the same small set of clean single-note audio samples, of course you’ll end up distinguishing them. This is like building a neural net which solves just MNIST and writing “we achieve full handwritten characters recognition” in the headline.
Also, note that the online tool he uses doesn’t let you hear the next note until you’ve correctly guessed the previous one, and this sort of feedback makes the exercise significantly easier (after the first note you could basically achieve near 100% accuracy with relative pitch alone).
I would claim that this guy essentially goodharted himself into passing this online test. If you train for a month using always the same small set of clean single-note audio samples, of course you’ll end up distinguishing them. This is like building a neural net which solves just MNIST and writing “we achieve full handwritten characters recognition” in the headline.
Also, note that the online tool he uses doesn’t let you hear the next note until you’ve correctly guessed the previous one, and this sort of feedback makes the exercise significantly easier (after the first note you could basically achieve near 100% accuracy with relative pitch alone).