How Can AI Pass the Obama Scale Test
Keywords: Affordance, HOI, Functionality
Abstract: Understanding the complexities of computer vision, particularly in humor recognition, is a challenging task. This essay delves into the intricate nature of computer vision as outlined by Karpathy in his 2012 blog post. The focus is on the "Obama scale" image, which humorously captures a moment where Barack Obama leans on a scale while another person stands on it. Humans can easily grasp the humor in this image, but AI struggles to do the same. This essay explores the various tasks that an algorithm must comprehend to "get the joke" in the image and assesses how far current technology is from achieving this level of understanding. Drawing from recent research in computer vision and human-object interaction, it discusses the prospects of AI in achieving a comprehensive understanding of such complex visual scenarios.
Submission Number: 24
Loading