A brief history of intelligence: evolution, AI, and the five breakthroughs that made our brains

Abstract

Intelligence evolved through five major breakthroughs spanning four billion years: steering in early bilaterians, reinforcement learning in early vertebrates, simulation in early mammals, mentalizing in early primates, and language in early humans. Each breakthrough built upon previous ones, creating increasingly sophisticated forms of learning and cognition. The ability to steer required categorizing stimuli into good and bad. Reinforcement learning enabled trial-and-error adaptation through dopamine signaling. Simulation allowed animals to mentally test actions before performing them. Mentalizing - modeling one’s own and others’ minds - facilitated social intelligence. Finally, language enabled the accumulation of knowledge across generations. This framework helps explain both the capabilities and limitations of modern artificial intelligence systems, which have replicated some but not all of these breakthroughs. While current AI can engage in reinforcement learning and pattern recognition, it lacks the ability to build rich world models or truly understand language. Understanding how biological intelligence evolved provides crucial insights for developing more human-like artificial intelligence. The evolutionary story also reveals that human intelligence emerged not from a single innovation but from the gradual accumulation of cognitive abilities over hundreds of millions of years. - AI-generated abstract