Bringing AI to Autonomous Systems: From Cognition to Collective Intelligence

Summary

This paper examines autonomous systems as a possible advanced stage of artificial intelligence and argues that their development requires both connectionist and symbolic AI, together with systems engineering. It proposes a general design and evaluation framework based on an agent architecture whose behavior is composed of cognitive functions organized around long-term memory containing the agent’s evolving knowledge. The framework addresses how sensory inputs can be connected to structured information in memory, how agents can make decisions and plan toward goals, and how multiple agents can coordinate to combine individual and collective intelligence. The paper treats trustworthiness as broader than the behavioral reliability expected of traditional systems: it also depends on cognitive properties and on whether an agent uses its knowledge validly when making decisions. It outlines possible methods for evaluating this cognitive dimension of trustworthiness. The authors conclude that a substantial gap remains between the aspirational vision of autonomous multi-agent systems and the current state of the art.