High-dimensional spaces are common in deep learning nowadays, and there's lots of room in high-dimensional spaces so any potential quantization errors are proba

High-dimensional spaces are common in deep learning nowadays, and there's lots of room in high-dimensional spaces so any potential quantization errors are probably insulated from incorrect recall by virtue of the fact that things are so far apart in the original space.