The concept of intelligence is often viewed as being solely confined to the human brain, but what if intelligence is more than just individual cognitive abilities? What if it's also embedded in the environmental infrastructure that facilitates thinking? This idea challenges the traditional view of intelligence and highlights the importance of considering the role of external factors in shaping intelligent behavior.
The article explores the idea of System 0, which refers to the underlying infrastructure that makes complex cognition possible. This includes both biological and non-biological systems that contribute to intelligent behavior. Current AI systems, particularly large language models, are skilled at pattern recognition and prediction but lack an evaluative core, which is essential for true understanding and intelligence.
To achieve true artificial general intelligence, AI systems need to be designed with an evaluative core that enables them to assess their own knowledge and uncertainty, prioritize goals, and make decisions based on values and norms. A potential solution is to integrate symbolic reasoning and knowledge graphs into AI systems to provide a more robust and human-like understanding.
The development of evaluative AI requires a multidisciplinary approach, incorporating insights from cognitive science, neuroscience, and philosophy to create more sophisticated and human-centered AI systems. By rethinking AI development and focusing on creating systems that are not just predictive but also evaluative and goal-driven, we can unlock the potential for true superintelligence.
