isn't this the same findings as the old "we trained to identify huskies, but instead we identified snow" problem?