2 points
Neural networks are proof that we suck at defining things, but our labels are excellent.
Dr. Dan McClellan has a great segment on Data Over Dogma about Prototype Theory ( On YouTube, time counter at 35:17), in which he points out dictionaries aren’t authoritative in telling us what words mean, rather they tell us what words have been used to mean so far.
He brings up the word furniture as an example talking about prototype theory, and talks about how we have a general sense of what furniture is ( we know it when we see it ) but we cannot define a set of features that includes all furniture and excludes all things not furniture.
2 points
