AI Can See Optical Illusions
And it might teach us about our own brains The post AI Can See Optical Illusions appeared first on Nautilus .
Two centuries ago, a toymaker named Charles Benham discovered an optical illusion that produced colors from a spinning white disk with black lines. This phenomenon, now called "subjective color," has puzzled scientists for the past 200 years. The prevailing theory suggests that it arises from feedback within our visual cortex, which predicts and anticipates visual information to fill in gaps.
Recent research has unveiled more peculiarities of the Benham's Top illusion. For instance, arcs at varying distances can evoke different colors, which can even reverse when the direction of rotation changes. Interestingly, artificial intelligence networks can also perceive this illusion, as demonstrated in a new study conducted by researchers from the National Institute for Basic Biology in Japan.
They created an artificial neural network (ANN) to predict the next image in a sequence, trained on first-person videos of natural landscapes. When the ANN was presented with Benham's Top, it produced images showing faint colors, just like a human observer would. Even when the spinning disk was altered to change the arcs' movement or direction, the ANN displayed the same color responses.
The researchers concluded that the ANN had learned to process visual information similarly to our biological neural networks, supporting the feedback model of subjective color. This finding provides further evidence for the theory that our visual cortex generates colors in moving black-and-white images through feedback.
Written by urgent.news from Nautilus's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
