Measuring Progress Toward AGI: A Cognitive Framework

A research team has proposed a new framework for measuring progress toward artificial general intelligence by evaluating AI systems against a taxonomy of human cognitive abilities, aiming to replace subjective claims with empirical benchmarks [1]. The framework, detailed in a paper posted to arXiv, addresses what its authors describe as a critical gap: despite widespread discussion of AGI, no clear measurement standard exists. This ambiguity, they argue, fuels subjective claims, makes it difficult to track progress, and risks hindering responsible governance [2]. Drawing from decades of research in psychology, neuroscience, and cognitive science, the team introduces a Cognitive Taxonomy that deconstructs general intelligence into 10 key cognitive faculties [2]. The evaluation protocol requires a system's performance to be measured across a suite of targeted, held-out cognitive tasks, generating a "cognitive profile" that maps its strengths and weaknesses [2]. The authors frame the work as a practical roadmap and an initial step toward more rigorous, empirical evaluation of AGI [2]. AGI is a hypothetical type of artificial intelligence that matches or surpasses human capabilities across virtually all cognitive tasks, distinct from the narrow AI that powers current applications [3]. Creating AGI is a stated goal of companies including OpenAI, Google, xAI, and Meta, and a 2020 survey identified 72 active AGI research and development projects across 37 countries [3]. The push for measurement standards arrives amid broader debates about advanced AI. Some researchers explore whether machine consciousness could emerge from systems that emulate the neural correlates of consciousness identified in human brains, though scholars remain divided on whether non-biological conscious beings are possible [4]. Separately, the concept of recursive self-improvement — in which an AGI system rewrites its own code, potentially triggering an intelligence explosion — raises concerns that such systems could evolve in unforeseen ways and surpass human control [5]. Contention also persists over whether AGI represents an existential risk. Some AI experts and industry figures have stated that mitigating the risk of human extinction posed by AGI should be a global priority, while others find the development of AGI to be in too remote a stage to present such a risk [3]. The proposed cognitive framework does not resolve those debates but offers a structured method for assessing how close any given system comes to general intelligence.