1. 26
  1.  

  2. 6

    The neural network really likes brown, beige, and grey.

    I would guess that this is a factor of some central tendency. When neural nets don’t know what to do, they’re biased in favor of the mean (or median, if L^1 loss was used).

    Dondarf 145 151 226

    Horble Gray 178 181 196

    Gray Pubic 6 193 214

    Snowbonk 201 199 165

    Bank Butt 221 196 199

    Dorkwood 61 63 66

    Light of Blast 175 150 147

    Dope 219 209 179

    Stoner Blue 152 169 159

    Turdly 190 164 116

    I would invite this neural net over for dinner. Or let it name variables if I had an employer or client that I especially disliked.

    1. 6

      How could you leave out “Stanky Bean”? These names are fantastic

    2. 3

      Those paint names. “Stanky Bean.” “Grass Bat.” “Rose Hork.” ?

      1. 1

        The best is “Turdly”, which is, in fact, pretty turdly.