We're obviously not training it on the correct answers, like with those grid images where we identify the motorcycle, so are you suggesting this model is in sea

We're obviously not training it on the correct answers, like with those grid images where we identify the motorcycle, so are you suggesting this model is in search of data around how inaccurate people are when guessing between frog and toad?