Thinking Machine 7: Go
Play Go against a transparent intelligence.
Place a stone on the board to begin. Or, if you prefer, simply watch the computer play itself.
What is happening?
The artwork is an artificial intelligence program, ready to play a game of Go with the viewer. Although the rules of the game are simple, the resulting contest is a subtle contest for territory.
If the viewer confronts the program, the computer's thought process is visualized on screen as it searches for the best move. A map emerges from the traces of potential future positions it considers. Those traces become a key to the invisible lines of territory in the game as well as a window into the spirit of a vintage thinking machine. Indeed, no LLMs here: this is good old-fashioned AI, based on simple and transparent search algorithms.
What do the images mean?
As the machine thinks, a network of curves appear on the board, outlining territorial influence in possible futures. The diagram below shows the process for a single potential position contemplated by the computer. It's an artistic, visual approximation. In the default color scheme, black and white curves suggest each player's territory; in the color view, orange curves show areas more likely to be influenced by black; green curves outline regions where white dominates. Overlaid, the curves for many possible positions create a visualization of the thinking process.
Why is the computer so [easy/hard] to beat?
The Go playing engine is not meant to be an expert. If you're a tournament player, you would clobber most casual players, and you'll clobber Thinking Machine 7 too. If you barely remember the rules of the game, the artwork may clobber you instead. The engine uses only basic algorithms; the goal is not to make an expert Go playing program but to lay bare a strategic search.
What were Thinking Machines 1 to 6?
The first six versions of Thinking Machine were based on chess. An early piece in the series is on display at the NY MoMA through June 13, 2027. And version 6 may be viewed online, where you can read more detail about the series. This page, number 7 in the series, represents a new direction for the theme.
Did you use AI to make this?
The concept, design, look, feel, and text are all human. But the code itself was written with AI help. In that sense this is a conceptually recursive work, visualizing AI with the help of AI.
Who created this series?
Martin Wattenberg and Marek Walczak, with contributions from Johanna Kindvall and Fernanda Viégas. Additional thanks: Eve Wattenberg for advice on interaction, Tzu-Sheng Kuo for many useful comments on Go play, Gus for pointing out people might want to play as white, and Helena Vasconcelos for a last-minute nudge for a black and white color scheme.