NMAT PSYCHOLOGY 7: Sensation and Perception - The Surprising Reason Machines Still Fail at Seeing

Have you ever wondered why you can recognize a "school bus" or "keyboard" from any angle, while the world's most advanced Deep Neural Networks (DNNs) often fail? In this video, we dive into the fascinating world of human visual perception and explore why current AI models are not yet adequate behavioural models of our core object recognition abilities. We uncover the Crocodile Conjecture, a theory explaining why humans are immune to the tiny adversarial perturbations that trick AI into misclassifying a banana as a crocodile.
We also explore the fundamental principles of Gestalt Psychology, including reification, multistability, and invariance, which allow our minds to perceive "the whole" as something else than just the sum of its parts. You will learn about the laws of grouping—such as proximity, similarity, and closure—that help us organize complex visual fields into meaningful shapes. Additionally, we discuss the mystery of Prosopagnosia, or face blindness, a condition where damage to the fusiform gyrus leaves individuals unable to recognize familiar faces, even while their other visual processes remain intact.
Finally, we break down the critical difference between sensation—the physical detection of stimuli—and perception—the psychological interpretation of that data. Whether it’s understanding Weber’s Law and the just noticeable difference in stimulus intensity, or exploring how top-down processing uses your past experiences to guess what you’re seeing, this video is a complete guide to the science of the human eye and mind.
#VisualPerception #AIvsHuman #GestaltPsychology #Neuroscience #FaceBlindness #SensationAndPerception #DeepNeuralNetworks #Psychophysics