Functional Consciousness - Pillar 26: Superintelligence and The Footprint Problem

Abstract

AbstractThis paper examines what it calls the Footprint Problem: the relationship between the physical requirements of advanced artificial intelligence and humanity’s ability to monitor, regulate, or restrict its development. Current frontier AI depends upon advanced chips, large computing clusters, substantial energy, specialized infrastructure, and concentrated technical expertise, creating a technological footprint that can potentially be observed and controlled. Yet the history of computing is also a history of increasing efficiency, declining costs, improved algorithms, compression, automation, and the diffusion of capabilities that were once available only to governments or major institutions. As artificial intelligence itself becomes increasingly useful in programming, experimentation, engineering, and AI research, another question emerges concerning how the resources required for future advances may change. Rather than attempting to predict when artificial superintelligence will arrive, whether current large language models can produce it, or what such an intelligence might do, the Footprint Problem asks a more fundamental question: What happens to strategies for governing advanced artificial intelligence if the technology being governed becomes progressively cheaper, smaller, more distributed, and more difficult to distinguish from ordinary computing?