Games provide a happy vehicle for studying methods of simulating certain aspects of intellectual behavior; happy because they are fun, and happy because they reduce the problem to one of manageable proportions.
Arthur L. Samuel, “Programming Computers to Play Games”
Games have fixed rules; gaming involves deception; gamers have opponents. The general game fabric, therefore, is not necessarily consonant with design. Architecture is not Monopoly, Parcheesi, or checkers. Such games assume perfect information, winning is explicit, and the process is composed purely of sequential acts—moves—governed by immutable, fathomable, and predefined rules. Design does not have a clear-cut format; so why is “design gaming” considered avant garde and fashionable? What good are games?
Games are a learning device for both people and machines. “Play and learning are intimately intertwined, and it is not too difficult to demonstrate a relationship between intelligence and play” (McLuhan, 1965). Games are models by which or with which learning takes place. They eliminate worrisome complications and perplexities by using artificially contrived situations. They involve the amalgamation of strategies, tactics, and goal-seeking, processes that are useful outside of the abstraction of gaming, certainly in design.
Historically, chess has been the machine’s baccalaureate. In 1769, Baron Kempelen constructed a fraudulent chess-playing machine, The Maelzel Automaton. The hoax was achieved by the labors of a concealed dwarf who observed the moves from beneath and manipulated a mechanical dummy. The need for such fraudulence has since been overcome with computing machinery. The pioneering works of Claude Shannon (1956) and the later efforts of Herbert Simon (and Baylor, 1966) and his colleagues have led to the development of chess-playing machines that demonstrate sophisticated techniques for intelligent decision making by strategically looking ahead. The approximately 1,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000 possible chess positions render it improbable that a calculating device can exhaustively search all possible courses of action. As a result, a chess-playing machine looks at local situations, looks ahead some small number of moves, and makes a speculation. Such techniques are indeed relevant to the construction of an architecture machine.
However, rather than map intelligent chess techniques into design tactics, let us concentrate on one key issue in gaming that is particularly relevant to design machines, that is, the relation of local actions to global intents. In architecture the local moves are embodied in physical construction and destruction procedures (whether explicitly executed by a designer or implicitly by zoning laws or the like), and the global goal is quite simply “the good life.” In chess, the consensus is that the global goal to win, by taking the opponent’s king, has little bearing on the local actions and the skillfulness of making these moves, particularly in the opening and middle game. The loser can indeed have played the better game.
In architecture, the losers are rarely the players. This is historically true, but let us assume that it changes and each resident can play the game with the global goal being the good life. The rules for achieving this goal are certainly unclear; they vary for each person, and, as in our Alice in Wonderland croquet game, they are ever-changing. Furthermore, in this game there is no coup de grace or checkmate; the global goal has no “utility function,” no cost-effectiveness, no parameters to optimize.
But the chess analogy suggests that a machine could learn to play architecture from local design pursuits and that these actions would be draftable without an absolute definition of the good life. A machine’s adroitness in design could evolve from local strategies that would self-improve by the machine testing for local successes and failures. In other words, we are suggesting that a machine, as well as any student of architecture, can learn about design by sampling the environment for cheers and boos. For example, in a tennis match a human spectator who is ignorant of the rules, scoring procedures, or criteria to win can begin to distinguish good from bad play merely by observing the applause of the other spectators.
Such learning by inference can apply to the breeding of intelligent design partners able to discriminate between plausible patterns and dubious forms. With a history of local punishments and rewards, an adaptable machine can evolve without a global set of values and adaptable rules to achieve them. Maybe nobody knows how to play, maybe everybody applauds at the wrong time, and maybe the good life is the wrong goal. But the thrust of the game analogy is that we do not have to answer these questions in order to proceed.