Human vision is hardly perfect - it can be fooled by certain situations very easily indeed. It can event be influenced by prior expectation. But it still basically works nearly all of the time, with the brain/eye system able to make sense out of even unfamiliar scenarios most of the time. If it wasn't able to do that, we'd all be dead. Optical illusions are fascinating, but I bet most of those particular triggers - at least not ones that are potentially dangerous - don't crop up all that much outside their very carefully contrived scenarios. Bias about what we're expecting to see is probably largely advantageous - we can go back and check, and re-evaluate our interpretations because our bias flags up unexpected things which are mostly actually wrong.
Then the researchers introduced something incongruous into the scene: an image of an elephant in semiprofile. The neural network started getting its pixels crossed. In some trials, the elephant led the neural network to misidentify the chair as a couch. In others, the system overlooked objects, like a row of books, that it had correctly detected in earlier trials. These errors occurred even when the elephant was far from the mistaken objects.
And as for the elephant itself, the neural network was all over the place: Sometimes the system identified it correctly, sometimes it called the elephant a sheep, and sometimes it overlooked the elephant completely.
Today’s best neural networks for object detection work in a “feed forward” manner. This means that information flows through them in only one direction. They start with an input of fine-grained pixels, then move to curves, shapes, and scenes, with the network making its best guess about what it’s seeing at each step along the way. As a consequence, errant observations early in the process end up contaminating the end of the process, when the neural network pools together everything it thinks it knows in order to make a guess about what it’s looking at.
“The human visual system says, ‘I don’t have right answer yet, so I have to go backwards to see where I might have made an error,’” explained Tsotsos, who has been developing a theory called selective tuning that explains this feature of visual cognition.
https://www.quantamagazine.org/machine-learning-confronts-the-elephant-in-the-room-20180920/
No comments:
Post a Comment
Due to a small but consistent influx of spam, comments will now be checked before publishing. Only egregious spam/illegal/racist crap will be disapproved, everything else will be published.