Paper
A Definition of AGI
The lack of a concrete definition for Artificial General Intelligence (AGI) obscures the gap between today's specialized AI and human-level cognition. This paper introduces a quantifiable framework to address this, defining AGI as matching the cognitive versatility and proficiency of a well-educated adult. To operationalize this, we ground our methodology in Cattell-Horn-Carroll theory, the most empirically validated model of human cognition. The framework dissects general intelligence into ten core cognitive domains-including reasoning, memory, and perception-and adapts established human psychometric batteries to evaluate AI systems. Application of this framework reveals a highly "jagged" cognitive profile in contemporary models. While proficient in knowledge-intensive domains, current AI systems have critical deficits in foundational cognitive machinery, particularly long-term memory storage. The resulting AGI scores (e.g., GPT-4 at 27%, GPT-5 at 57%) concretely quantify both rapid progress and the substantial gap remaining before AGI.
Authors: Hendrycks, Dan · Song, Dawn · Szegedy, Christian · Lee, Honglak · Gal, Yarin · Brynjolfsson, Erik · Li, Sharon · Zou, Andy · Levine, Lionel · Han, Bo · Fu, Jie · Liu, Ziwei · Shin, Jinwoo · Lee, Kimin · Mazeika, Mantas · Phan, Long · Ingebretsen, George · Khoja, Adam · Xie, Cihang · Salaudeen, Olawale · Hein, Matthias · Zhao, Kevin · Pan, Alexander · Duvenaud, David · Li, Bo · Omohundro, Steve · Alfour, Gabriel · Tegmark, Max · McGrew, Kevin · Marcus, Gary · Tallinn, Jaan · Schmidt, Eric · Bengio, Yoshua