Blondie24

Blondie24 is an artificial intelligence checkers -playing computer program that earned its name from the screen‑name used by a team led by David B. Fogel . The whole stunt was billed as a “let’s see if a machine can beat humans at something that should be easy” experiment, and it actually worked—sort of.

The screen‑name was dropped on The Zone , a dusty corner of the early‑web where hobbyists still pretended board‑gaming mattered. From 1999 onward Blondie24 squared off against roughly 165 human opponents and managed to climb to a rating of 2048, which translates to “better than 99.61 % of the playing population” on that site. Not exactly a grand slam, but enough to make the developers grin and the critics snort.

Design‑wise, Blondie24 leans on a classic minimax search tree, but the evaluation function isn’t some hand‑crafted heuristic dreamed up by a bored graduate student. Instead it’s a deep‑learning convolutional artificial neural network that gulp‑downs a vector representation of the board and spits out a single scalar value. That value then gets fed back into the minimax routine, which proceeds to prune branches like a gardener with a vendetta.

The real kicker is how the neural‑network weights were obtained. Rather than hand‑tuning them, the team threw a population of Blondie24‑like agents at each other in a self‑generated tournament. Each win earned a point, each loss subtracted two, and draws were ignored. After a few hundred rounds the losers were culled, the winners were mutated, and the cycle repeated—a textbook case of evolutionary algorithm in action. This whole iterative cull‑and‑breed loop is what the community now calls neuroevolution .

What makes Blondie24 noteworthy isn’t the raw playing strength (though it’s respectable); it’s that the program never needed any human‑crafted expertise beyond “here’s a board, figure it out.” The entire skill set emerged from points earned in a blind evolutionary race. In other words, the AI learned to play checkers the way a cat learns to ignore you: by sheer, indifferent persistence.

David Fogel , together with his partner Kumar Chellapilla, chronicled the whole circus in a handful of papers. Fogel also penned a book—Blondie24: Playing at the Edge of AI—detailing the trials, the triumphs, and the occasional existential crisis that comes with watching a machine stumble into competence. The book was published by Morgan Kaufmann Publishers in 2002, complete with an ISBN that probably sits on a shelf somewhere, gathering dust.

The significance of Blondie24 rippled outward, influencing later work in Genetic Programming and spawning a whole sub‑field where neural nets evolve without a teacher looking over their shoulders. It proved that you could, in fact, let a computer figure out a game’s strategy from scratch—provided you’re willing to sit through a few thousand iterations of “whoops, that didn’t work.”