>>5
But you work out the rest from context, which makes this a nice illustration of bot failure. Cameras are terrible at preserving definition of skin tones, and there's very little depth information. I reckon photos from a known angle of emotionless faces with a preset light source would give very accurate results.
Male / Female is probably trivial but gamifying that would be quite problematic.
Meanwhile "cat or dog" has also turned out to be a very easy problem even without using external cues, consistent angles, knowledge of different breeds etc.