Google DeepMind has introduced a cognitive science-based framework to measure progress toward Artificial General Intelligence (AGI), accompanied by a Kaggle hackathon offering a $200,000 prize pool to develop evaluations for critical cognitive abilities.

The framework, detailed in *Measuring Progress Toward AGI: A Cognitive Taxonomy*, identifies 10 key cognitive abilities essential for general intelligence: perception, generation, attention, learning, memory, reasoning, metacognition, executive functions, problem-solving, and social cognition. It proposes a three-stage evaluation protocol—benchmarking AI systems against human performance using held-out test sets and demographically representative human baselines—to assess capabilities across these domains.

To operationalize this framework, Google is partnering with Kaggle to host a hackathon focused on five high-gap cognitive abilities: learning, metacognition, attention, executive functions, and social cognition. Participants will use Kaggle’s Community Benchmarks platform to design and test evaluations against frontier AI models. Prizes include $10,000 awards for top submissions in each track and $25,000 grand prizes for the overall best entries.

Submissions are open March 17 through April 16, with results announced June 1. This initiative aims to bridge the gap between theoretical AGI frameworks and practical evaluation tools, leveraging community expertise to advance measurable progress toward general intelligence.