No, he is demonstrating that digitization (sampling) of any audio waveform must be reproduced as the same waveform, because there is only one possible waveform that can fit the sampled points. This is true regardless of the filetype/compression used, though at very low sampling rates (below 8 bits/second) information is noticeably lost because there aren't enough sample points in a given timeframe, and then you get distortion during playback.
I'm not sure specifically where the line is, but there are physical limits to how fast a speaker cone can vibrate to reproduce the waveform no matter how detailed it is (it would depend on the specific playback device), regardless of whether the source is analog or digital. There's a point where the level of detail of the sampled waveform doesn't matter, because there's no audio equipment in existence that can reproduce that waveform with any accuracy, and 320 kHz is definitely above that point.
It is more likely that the "muddy" quality of the MP3 files you're referring to is related to choices made while converting/compressing the file, such as the differences in dithering that he demonstrates later in the video.