Title

From Artificial Failure to Artificial Perception

Abstract
Complex systems fail. I argue that the nature of these failures can be a blueprint characterizing living organisms and biological intelligence.Perception offers a convenient testbed for this idea: Artificial Intelligence has been focused on reproducing hman’s perceptual successes in object recognition, classification, and video prediction. However, in reality, human perception has numerous flaws, some of which are shared with other animals; perception has not evolved to reach perfect accuracy, but to be useful. Through the example of perceptual illusions and memory failures, I will try to convince you that imitating biological successes can be misleading; and that imitating failures offers a path towards understanding and emulating intelligence in artificial systems.