In brief
A new A.I. system named Centaur has been developed by an international team of scientists to simulate human behavior in psychological experiments. This advance may help decode complex aspects of human cognition as traditional A.I. remains distinct from human thinking.
Artificial intelligence (A.I.) continues its rapid evolution, primarily by tech giants such as OpenAI and Meta, who are pursuing the ambitious goal of developing artificial general intelligence (A.G.I.). Despite substantial investment, A.G.I. lacks a concrete definition and remains more of an aspiration rather than a reality. Current A.I. capabilities include playing chess at championship levels and analyzing protein structures, showcasing its growing potential. Although chatbots like ChatGPT can generate human-like text, they still differ significantly from actual human cognition. Most A.I. systems excel in specific tasks but falter in areas outside their specialized training. For instance, although a chess-playing A.I. can strategize in chess, it is incapable of performing tasks like driving a car. Additionally, these systems make unusual errors, exemplified by allowing illegal chess moves.In an innovative approach, researchers have developed Centaur, an A.I. system that can participate in psychological experiments as if it possessed a human mind. This groundbreaking project, detailed in the journal Nature, seeks to utilize A.I. as a tool in cognitive science research. Understanding how humans learn, remember, and make decisions is at the heart of cognitive science, and A.I. can serve as a surrogate participant in experiments designed to test various theories about human behavior. For example, studies illustrate that individuals prefer certainty over risking potential losses—even when offered substantial gains. The insights gained from employing A.I. like Centaur could provide valuable contributions to the ongoing quest to unravel the complexities of human cognition and behavior.
