Fooling Computers With Optical Illusions Is A Step In The Right Direction
from the maybe-my-laptop-will-appreciate-escher-soon dept
Just a few weeks ago, computers were learning to understand knock knock jokes. Now, computers take another step towards being human; they are fooled by optical illusions, just like us. By creating a computer program that learns how to comprehend different shades of gray similar to the way that an infant learns, the computer also falls prey to White's Illusion, an illusion in which bands of gray appear different even though they are the same shade. This is an important step in computer vision since this illusion is not a physiological one, where optical sensors send the brain the wrong image, but rather, a cognitive one, where perception of an image is misinterpreted. As we get closer to true artificial intelligence, what other kinds of human faults will computers be able to emulate? And, considering that computers are already better than us at certain tasks, will those faults be amplified? After all, "to err is human, to really screw up you need a computer."