computational-complexity
Replicating the evolutionary process that produced human intelligence requires computational resources far beyond current capabilities, even with optimistic projections like Moore's law.
1 chapter across 1 book
Superintelligence: Paths, Dangers, Strategies (2014)Nick Bostrom
Chapter 2 of Bostrom's Superintelligence explores various plausible technological paths to achieving superintelligence, including artificial intelligence, whole brain emulation, biological cognition, and human-machine interfaces. It emphasizes the definition of superintelligence as an intellect vastly outperforming humans across virtually all domains, and discusses the challenges and theoretical considerations involved in creating AI, particularly through learning systems and evolutionary algorithms. The chapter critically examines evolutionary arguments for AI feasibility, highlighting the computational complexity and unknown efficiency gains in replicating or surpassing natural intelligence.