With self-driving cars and whatnot, machines are progressively developing a more pressing need for accurate eyesight. As such, in addition to smarter computers, we need computers that can quickly discern the meaning of something on sight. In an effort to kill both birds with one stone, a certain team of scientists at Brown University have begun teaching machines to recognize and see through optical illusions.
The team created a series of virtual feedback connectors, similar to the neurons in a brain. These connectors are designed to activate in response to certain visual stimuli, like shapes and colors. Through a vast network of these connectors, the test machines should been able to identify specific targets, even when the environment is working counter. For an example of a test, a machine could pick out red berries from several other materials in a room washed in red light, all because it could recognize color, shape, size, and other elements. The testing only just recently started, so they have no conclusive stuff to report just yet. But if they can make this work, camera-based tech could take a big step forward.
