New AI Framework Defines Intelligence by Process, Not Output

A new paper proposes a framework called Process-Constituted Intelligence, arguing that true intelligence lies in the iterative process of creation rather than just the final output. The authors, including Michael Richardson, suggest that current generative AI models are only weakly equivalent to human cognition because they learn from traces of human processes without instantiating those processes themselves. The paper outlines seven features to define strong equivalence and proposes design principles for AI that would better embody these processes, along with methods for auditing AI to test for strong equivalence. AI

IMPACT Proposes a new theoretical lens for evaluating AI, potentially influencing future AI design and evaluation metrics.

RANK_REASON Academic paper defining a new theoretical framework for intelligence. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• generative artificial intelligence
• Hugging Face
• Michael Richardson
• Process-Constituted Intelligence

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →