Lionhead Studios’ 2001 god-game Black & White, led by Peter Molyneux with AI by Richard Evans, centered on a giant animal creature that the player raises and trains. Its defining feature was learning: rather than following a fixed script, the creature watched the player, tried things, and adjusted its behavior based on reward and punishment, developing a distinct personality over the course of a game.
In the official Game Developer postmortem, dated June 2001, Molyneux described the ambition plainly: “The Creature mirrors you and your actions, so in Black & White we’ve got a game in which part of the game itself learns from everything you do.” He stressed that the apparent intelligence came from genuine learning mechanics, not randomness, so “no two players experience identical creatures.”
Under the hood, Evans combined several classic AI techniques - decision trees for opinions about how to satisfy a desire, simple perceptrons for the desires themselves, and player-driven reinforcement to shape which behaviors the creature adopted. The creature also tried to infer the player’s intent from their actions, a form of empathic learning. Black & White became one of the most-discussed examples of adaptive, learning-based character AI in a commercial game, and a touchstone for designers exploring believable, trainable agents.