That’s actually a pretty good analogy; both rely on throwing out information that the algorithms perceive as un-seeable or un-hearable. If you’re doing it on a high quality large source file, it’s usually pretty good, but if you have a janky and/or small source, you end up with artifacts and bad quality.
As far as the film analogy goes, what gave me the idea is that I have a photo manipulation software package, and one of the options that you can buy for it is the ability to impose stuff like film grain, film exposure curves and film color responses onto digital photographs. So for example, you might go take your photo with your spiffy Canon EOS DSLR in full color, but on the back end, you can make it look like it was taken with Tri-X 400 or Kodachrome 25. Interesting, but only really necessary if you LIKE the look of film.