tbh for tools like Shazam there's no fundamental difference between a database + hashing algorithm and a self-supervised model; both are great indexing & compression solutions, just for different scales of data.