Measuring progress toward AGI: A cognitive framework

2026-03-17 · Source: Google DeepMind News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Advanced, medium

Summary

Google DeepMind has introduced a new framework and paper, "Measuring Progress Toward AGI: A Cognitive Taxonomy," to empirically evaluate Artificial General Intelligence (AGI) using cognitive science principles. The framework identifies 10 key cognitive abilities crucial for general intelligence in AI systems, including Perception, Generation, Attention, Learning, Memory, Reasoning, Metacognition, Executive functions, Problem solving, and Social cognition. To facilitate practical application, DeepMind is partnering with Kaggle to launch a hackathon, "Measuring progress toward AGI: Cognitive abilities," inviting researchers to design evaluations for five specific cognitive abilities: learning, metacognition, attention, executive functions, and social cognition. The hackathon offers a total prize pool of $200,000, with submissions open from March 17 through April 16, 2026.

Key takeaway

For AI Scientists focused on AGI development and evaluation, this cognitive framework offers a structured approach to benchmark system capabilities. You should consider participating in the Kaggle hackathon to contribute to standardized evaluation methodologies, particularly for less-explored cognitive areas like metacognition and social cognition, which could significantly advance empirical AGI measurement and potentially secure a share of the $200,000 prize pool.

Key insights

A cognitive taxonomy and evaluation protocol are proposed to measure AI progress toward AGI against human benchmarks.

Principles

• AGI evaluation requires diverse methods.
• Cognitive science informs general intelligence.
• Benchmark AI against human performance.

Method

A three-stage evaluation protocol involves assessing AI systems across cognitive tasks, collecting human baselines, and mapping AI performance relative to human distribution using held-out test sets.

In practice

• Design evaluations for learning ability.
• Develop metacognition assessment tasks.
• Utilize Kaggle Community Benchmarks.

Topics

• Artificial General Intelligence
• Cognitive Taxonomy
• AI Evaluation
• Kaggle Hackathon
• Cognitive Abilities

Best for: AI Scientist, AI Researcher, AI Engineer, Research Scientist

Related on AIssential

• Measuring progress toward AGI: A cognitive framework — A cognitive taxonomy and evaluation protocol are proposed to empirically measure AGI progress against human capabilities.
• Live Q&A With the Hosts: Measuring Progress Toward AGI - Cognitive Abilities Hackathon — Robust AGI evaluation requires targeted benchmarks isolating specific cognitive faculties, moving beyond general problem-solving.
• Live Q&A With the Hosts: Measuring Progress Toward AGI - Cognitive Abilities Hackathon — Robust AGI evaluation requires targeted benchmarks for specific cognitive faculties, moving beyond general problem-solving.
• The Secret is Out: How we're Building AGI — True AGI requires conceptual knowledge validation and integrated neuro-symbolic architecture, unlike statistical LLMs.
• Uneven Evolution of Cognition Across Generations of Generative AI Models — Generative AI models exhibit uneven cognitive development, excelling in language but struggling with visual-perceptual reasoning.
• ARC-AGI-3 — General intelligence is defined by the ability to learn new skills, not just perform existing ones.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Google DeepMind News.