> try to politicise this Okay! "human learnability of abstract network representations" Which humans? Did they test it works equally well for all races and g

> try to politicise this Okay! "human learnability of abstract network representations" Which humans? Did they test it works equally well for all races and genders and body shapes, or is their model of human learning biased to work best for buff white men and is inferior for everyone else? What's in the abstract network representations, are they truly neutral, or are we just helping humans to learn from biased data that will make them subconsciously more racist/sexist/heightist?