Liberties

What is a game? Ludwig Wittgenstein famously chose this nebulous concept to illustrate what he meant by “family resemblance,” where the individual members of a class can be determined to fulfill no necessary and sufficient conditions for admission, and instead only share some traits with some others in the class, others with others. Yet we can at least identify two types of game, which seem not just distinct from one another but very nearly opposite. One class of games, which includes peek-a-boo, charades, and musical improvisation as representative instances, is characterized by free expressivity. It is the manifestation of what Friedrich Schiller called the Spieltrieb, the “play-drive,” which is innate in all human beings insofar as they are free. The other class includes chess, fencing, and wargames as its representative instances. If there is still some dose of freedom operating in this sort of game, it is freedom under severe constraints. The purpose here is to win, and one does so by means of strategy aforethought. In such games, serendipity and spontaneity are disadvantages. While some such games may, like Schillerian free play, be “fun” (especially when you win and the other guy loses), at their outer edge they shade over into a domain of human endeavor that has little to do with leisure at all. At their most serious they can determine the fate of the world. It is this latter sort of game alone that machines are capable of “understanding.” Strategy games, in other words, are essentially algorithmic. A good portion of the history of computing has been dedicated in fact to training machines up algorithmically in such narrow domains as chess, then Go, and more recently a full array of natural-language processing tasks. This training has facilitated the gradual progress of the machines from domain-specific “weak artificial