Computer Chess, an offbeat comedy that was a sleeper hit at Sundance (released Aug. 9 in D.C.), takes place in the 1980s, just before the artificial intelligence explosion. Focusing on computer chess machines, and the eccentric human programmers behind them, the story takes place over a weekend-long computer chess tournament. To capture the mood of the period, the film was shot almost exclusively in black-and-white on low-grade analog video.
The film follows a cast of oddball characters. Most memorable is the eager, albeit socially dysfunctional graduate student, Peter Bishton, played by Patrick Riester. Bishton’s dedication to his work — often at the expense of human connection (a romantic interest quickly fizzles out) — concludes on a truly unexpected note. Overall, the film is a study of contrasts: man v. machine, preprogrammed behavior vs. creativity, and so on, with the lines becoming increasingly blurred.
While computer chess machines from the 1980s are no match for today’s supercomputers, the film argues that they laid some of the groundwork for artificial intelligence as we know it today. To get a better sense of how these machines work and whether or not they’ve been important in AI’s development, I interviewed Dr. Ben Goertzel, chief scientist at the financial prediction firm Aidyia Holdings in Hong Kong. Dr. Goertzel is an expert in what is formally known as “artificial general intelligence” (AGI) — the quest to create advanced autonomous thinking machines (with intelligence beyond the human level) – and splits his time between Hong Kong and Rockville.
On the surface, the machines represented in the film are not terribly complicated: a human operator manually enters each move that corresponds to a move on the chess board. Standard game rules apply. More interesting is what happens between the moves; that is, how these machines think — even if that thinking is rather primitive.
The principal force at work here is alpha-beta pruning — a search algorithm that explores branches of possible future moves for the game and trims or “prunes” from those that look so unpromising that they don’t need to be explored anymore. For example: if knight takes rook, but leaves queen vulnerable, then that branch is pruned. “IBM built a special chip [for Deep Blue during the 1980s] that would evaluate chessboard positions super-fast, enabling alpha-beta pruning to be used with a larger lookahead [seeing several moves ahead] than standard hardware would enable at that point in time,” Dr. Goertzel explains.
For alpha-beta pruning to work, however, there’s another piece involved: heuristics. These are experience-based approaches utilizing shortcuts to problem-solving. “While some basic mathematical heuristics can be used, chess programs supplant these with heuristics coded by human experts,” Dr. Goertzel says. Going back to the earlier example, “IBM consulted with several chess experts to design favorable heuristic evaluation functions.”
So have these computer chess machines, as the film argues, laid the groundwork for today’s AI, or AGI even? “They haven’t really done so, but there is a common concept among alpha beta pruning and some AGI approaches: the idea of search through abstract spaces,” Dr. Goertzel says.
What does that mean, exactly? Well, the basic idea is: you’re faced with an initial situation where you would like to achieve a certain goal. The search is the process of figuring out the different ways you can do that. In chess, you’re trying to win the game by checkmate; in route planning (when you’re using Google Maps, for instance), you’re trying to get from Point A to Point B. But let’s be honest: the machines in Computer Chess are not going to be very helpful if you’re stranded in the middle of nowhere.
Computer Chess has two things going for it: an eccentric bunch of characters, some of whom leave quite an impression, and some profound existential questions. What is “natural” intelligence? And how does “artificial” intelligence figure into that?
As is often the case, this leads to yet more questions. Still, some of the characters are unnecessary — one of the contestants, Michael Papageorge, is more of a buffoon than a fully-formed character — and the story sometimes meanders, lacking narrative focus.