Vision in humans is a highly complicated mechanism. We evolved to symbolically respond to the objects we see. When we see a photo of a loved one, our brain is triggered emotionally rather than just computing the photo as geometric shapes. We also evolved to see symbols in objects that barely have resemblance to reality. We see faces where none exist, like this smiley face =).
A funny example when this adaptation fails:
Would a robot be able to make this same mistake? It may be possible to have its or his AI strong enough to make the mistake, but it seems unlikely to be invented in the near future. This is partially why making a self-driving car is so convoluted. How could its AI identify an old woman driving in front of it? In which case it would have to tailgate until she pulls over. On a more serious note, there are many instances where symbolic human vision is needed.

1 comment:
Yah, it's amazing that they've been working on that visual recognition program since the 1950s and still haven't gotten it right.
Post a Comment