Reframing superintelligence: Comprehensive AI services as general intelligence

Abstract

The concept of comprehensive AI services (CAIS) offers a new perspective on superintelligent-level AI, moving away from the traditional paradigm of self-improving, rational agents. CAIS envisions AI systems as services produced through research and development, providing a spectrum of functionalities, from narrow tasks to broad automation of human work. This approach emphasizes the role of human goals and preferences in guiding AI development, with services evolving through a recursive process of improvement. Unlike self-modifying agents, CAIS presents a world where general superintelligence emerges through distributed systems, already equipped with powerful capabilities. This shift in perspective reframes the potential for an intelligence explosion, the nature of advanced machine intelligence, and the relationship between intelligence and goals, prompting new considerations in AI safety and strategy.