A Tinkertoy computer
that plays tic-tac-toe



by A. K. Dewdney

indirectly kicks an "output duck," a bird-shaped construction. The output duck swings down from its perch so that its beak points at a number- which identifies the computer's next move in a game of tic~tac-toe.

   What precisely does the read head scan as it feels its way down the monolith? Nothing less than 48 rows of Tinkertoy "memory spindles" encoding all the critical combinations of X's and O's that might arise during a game [see illustration on opposite page]. Each spindle is a sequence of smooth spools connected axially by sticks and arranged in nine groups of three each, one group for each square of the tic-tac-toe board. The presence or absence of spools from a group indicates that a corresponding square of the tic-tac-toe board is vacant or is occupied by an X or O.

   The Tinkertoy computer is not fully autornatic: a human operator must crank the read head up and down and must manage its input. After the computer's opponent makes a move, the operator walks to the front of the machine to adjust the core piece inside the read head, registering the contestant's move. The operator then pulls on a string to cock the core piece for its impending whirl of recognition. When it discovers a memory that matches the current state of the game, the core piece spins, and the computer indicates its move.

   The best way to understand how the machine works in detail is to recount the story of its creation at the hands of the M.I.T. students: Erlyne Gee, Edward Hardebeck, Daniel Hillis, Margaret Minsky and brothers Barry and Brian Silverman. Most of the group has long since graduated and entered various computer professions. Perhaps the best-known team member is Hillis. He was the moving force behind Thinking Machines, Inc., which produces the well-known parallel supercomputer called the Connection Machine. (Perhaps Tinkertoys have something to teach us.)

   In 1975, when Hillis and Brian Silverman were in their sophomore year, they participated in a class project to build something digital from Tinkertoys. The students sat down to play. One made an inverter­a logic device that converts a binary 1 signal to a 0 signal and conversely. Another made an OR gate; if either of the device's two input signals happened to be a 1, then its output would also be a 1. It quickly became clear to the students that Tinkertoys were "computation universal," the theoretical term for a set of components from which a fully program

"I first had that experience [universality of computation] before I went to school. There weren't any [computersl yet, but we had toy construction sets. One was called TinkerToy.... What's strange is that those spools and sticks are enough to make anything."

in preface to LogoWorks

How many of us remember Tinkertoys, those down-home kits of colored wooden sticks and spools with holes in them? Amid our childhood constructions of towers or cranes, how many of us pondered the outer limits of the Tinkertoy world? Did we conceive of contraptions that reached the ceiling? Perhaps, but we lacked the kits or the time to make it

happen. Such a Tinkertoy fantasy took place several years ago when a student group from the Massachusetts Institute of Technology constructed a computer entirely (well, almost entirely) out of Tinkertoys!

   From a distance the Tinkertoy computer resembles a childhood fantasy gone wild or, as one of the group members remarked, a spool-and-stick version of the "space slab" from the movie 2001: A Space Odyssey. Unlike the alien monolith, the computer plays a mean game of tic-tac-toe. A Tinkertoy framework called the read head clicks and clacks its way down the front of the monolith At some point the clicking mysteriously stops; a "core piece" within the framework spins and then with a satisfying "kathunk"

The first three levels of the tic-tac-toe game tree

 120  SCIENTIFIC AMERICAN October 1989

Page 121


 Introduction - Table of Contents