A system can outperform people at a game without being able to handle a different game—or an uncertain real-world situation. That gap matters when deciding whether a useful AI capability deserves to be called intelligence. section 7.2
Intelligence involves more than fast calculation or success at one task. It includes responding to situations, making judgments, and applying what has been learned when circumstances change. Artificial intelligence refers to technologies that perform tasks traditionally associated with human intelligence; performing such a task does not, by itself, establish broad human-like ability. section 7.2 section 1.1
A game champion meets a new problem
A human chess expert can learn another game’s rules and apply strategic thinking. A specialized game-playing system may struggle even when the rules of its own game change. Calculators offer a simpler warning: they exceed human calculation speed, yet speed alone is not treated as intelligence. section 7.2
Human abilities also draw on experience that is not always consciously available. A child learns to walk through experimentation; an experienced driver may navigate with little conscious attention. Beliefs and learned habits can guide judgments, sometimes introducing hidden biases. These processes contrast with machine models built around explicitly stored data and algorithms. section 6.1.3
Claims of machine intelligence need careful scope. The Turing test proposed judging whether an observer could distinguish a machine’s answers from a person’s. Later predictions went further, imagining machines surpassing human intellectual activity and improving their own designs. Those possibilities should not be confused with demonstrated generalization. For practical AI decisions, ask how a system handles changed conditions, incomplete information, and uncertainty—not just how well it performs its showcase task. section 1.1 section 9.1 section 7.2
