>Let's say we train the NN using equal number of human faces of all races, and animals faces. You're describing something that happened, and the data sets very

>Let's say we train the NN using equal number of human faces of all races, and animals faces. You're describing something that happened, and the data sets very much did NOT represent a smooth spectrum of skin tones. It was heavily weighted towards light skin. The premise is incongruent with reality. Your analysis is fundamentally flawed.