Essay.02-Make.AI.Surprised.by.Magic.Shows.Build.a.More.Human-like.Machine-ZhuoyingLi-2100017764
Keywords: intuitive physics engine, reasoning
Abstract: This paper explores the potential of Artificial Intelligence (AI) to develop and utilize common sense for reasoning and shock response, drawing inspiration from human engagement with magic shows. Specifically, the study investigates the human cognitive capacity for intuitive physics and how violation of these intuitive principles during magic shows often evoke wonder. Drawing parallels, the paper explores how an AI can mimic this learning process to assimilate common sense and utilize it for reasoning. Despite the considerable achievement of data-driven methods such as GPT-4, these methods lack efficiency in out-of-distribution dataset generalization, creating a need for a new approach. The paper concludes with a proposal for a unique synthesis of cognitive science and deep learning, fortifying the push towards "small data for big tasks" and a more human-like AI system.
Submission Number: 43
Loading