These classifications are amazing but the fact that the first image in the article is classified as "a dog wearing a wide-brimmed hat" and not as "a chihuahua w

These classifications are amazing but the fact that the first image in the article is classified as "a dog wearing a wide-brimmed hat" and not as "a chihuahua wearing a sombrero" is telling of how far we are from true understanding of images. Only a human possessed with the relevant cultural stereotypes (chihuahua implies Mexican, ergo, the hat must be a sombrero) could make that conclusion. Even so, I firmly believe that at this rate of improvement, we're not far from that kind of deep understanding.