Moravec's paradox is a fundamental observation in artificial intelligence (AI) and robotics, first articulated in the 1980s by Hans Moravec, Rodney Brooks, Marvin Minsky, and other pioneers. It posits that while it is relatively simple to program computers to perform high-level reasoning tasks typically associated with human intelligence—such as playing chess, solving complex mathematical equations, or performing on standardized tests—it is immensely difficult to replicate the seemingly basic sensorimotor skills of a young child, such as walking, recognizing a face, or navigating a physical space.

The paradox is rooted in the counterintuitive reality that tasks appearing effortless to humans often require the most significant amount of computational resources. This is explained primarily through an evolutionary lens. Human skills associated with perception and movement have been honed over millions of years of natural selection. These "older" skills are deeply embedded in our biological machinery and operate largely beneath our conscious awareness, making them appear easy. Conversely, abstract reasoning and formal logic are evolutionary recent developments, likely less than 100,000 years old. Because these skills have not been refined by the same vast stretches of biological optimization, they require conscious effort and feel "hard" to us, despite being computationally simpler to model using digital logic.

Historically, this paradox led to significant miscalculations in early AI research. Pioneers in the 1950s and 60s believed that once they conquered "hard" problems like symbolic integration or theorem proving, "easy" problems like vision and common sense would follow. This optimism contributed to the first "AI winter" when those predictions failed. In response, researchers like Rodney Brooks proposed "Nouvelle AI" in the 1980s, which focused on building machines that prioritized sensing and action without the traditional overhead of complex internal representation. By the 2020s, the massive increase in raw computing power predicted by Moore's Law finally allowed AI to begin making substantial inroads into the perceptual domains Moravec identified as the true challenge.