Measuring Progress Towards AGI: A New Cognitive Approach

Artificial General Intelligence (AGI) has the potential to accelerate scientific discoveries and help solve humanity's most pressing problems. However, it is challenging to assess how close we are to this pivotal milestone, as there is a lack of empirical tools to evaluate the general intelligence of systems.

Tracking progress towards AGI will require a wide range of methods and approaches, and cognitive science provides an important element in this endeavor. Today, we present a new paper titled "Measuring Progress Towards AGI: A Cognitive Taxonomy," which lays the scientific foundation for understanding the cognitive capabilities of AI systems.

Along with the publication, we are collaborating with Kaggle to launch a hackathon, inviting the research community to assist in developing the assessments necessary for implementing this framework.

Our taxonomy is based on decades of research in psychology, neurobiology, and cognitive science. It highlights 10 key cognitive abilities that we believe will be essential for general intelligence in AI systems:

• Perception: extracting and processing sensory information from the environment
• Generation: creating outputs such as text, speech, and actions
• Attention: focusing cognitive resources on what is important
• Learning: acquiring new knowledge through experience and instruction
• Memory: storing and retrieving information over time
• Reasoning: drawing valid conclusions through logical reasoning
• Metacognition: awareness and monitoring of one's own cognitive processes
• Executive Functions: planning, inhibition, and cognitive flexibility
• Problem Solving: finding effective solutions to specific tasks
• Social Cognition: processing and interpreting social information and responding appropriately in social situations