Or: An Inspiring Story About The Human Condition; In Which Our Protagonists Fail To Pick The Intended Solution.
When I was a very young child – maybe five or six years old – I got my first gaming console: The original GameBoy. I was limited to half an hour of screen time per day, and there were frequent arguments about batteries, but it was so damn cool. I was extremely into that big ol’ brick.
Buying games was a lot more complicated in the Analog Days. You had to use your legs to walk to a store. I also didn’t have a lot of money, being a child, so I got a lot of second- and third-hand games. One of these was The Chessmaster (1991), passed on by a friend of my dad. It’s pretty common for smart people to think that they might be into Chess, only to find out that… yeah, no. He was glad to be rid of it, I was glad to have a new game. Everyone’s happy!
“Machine-Chess” has a long history. The ’90s were a particularly interesting time – the “golden days,” even, when humans and machines were increasingly evenly-matched. The infamous Kasparov – Deep Blue match represents a turning point in the narrative, and that happened in 1997. Sure, the GameBoy didn’t have nearly as much processing power, but GameBoy Chessmaster is an impressive little software package. According to this little test, it packs an Elo rating of “about 1558”, making it a decently mid-level player.
So how did I, a complete novice, perform against the perfect immortal machine?
I completely demolished it, of course.
It wasn’t even close. I’d win every game I played against it, on the maximum difficulty level. After a few weeks, I had beaten every game mode on every difficulty setting. I actually made my father ask his friend if there were higher difficulty levels in there somewhere, which probably did wonders for the guy’s self-esteem.
Soooo what happened here? I clearly wasn’t a Chess prodigy, since I couldn’t reproduce that success in games against human opponents. My games were – well, they were what you’d expect from a random five-year-old. I knew the rules and but I had no business winning except maybe against another child. So what happened?
Turns out there’s a “custom game” mode in there.
You know, one where you can set up your own scenarios? In case you want to try playing with a handicap, or something. I, being a person who did not speak English, had somehow found a way to fumble my way into that setting from other game modes. Having a vague understanding that the point of a game was to win, I would then proceed to fill the entire board with Queens. I wasn’t any good at Chess, but I had found a way to reliably win every game. (My parents would later explain that this was “not the point” of Chess, which I thought rather suspicious.)
So what’s the take-away here? Let me offer you some ways to interpret this story:
- Children are good models for how machine learning systems work. Five-year-olds are ruthlessly results-driven and don’t much care for the poetry of the checkered board.
- Humans can’t win against chess engines with sufficient processing power, and the same thing will eventually be true for every deterministic game, sure. But life, uh, finds a way.
- Your five-year-old probably isn’t a Chess prodigy. Sorry. But they might have a career in IT if they can reliably fumble their way into debug mode.