A neural net system built by Google has beaten the European champion in the Chinese game of Go, winning five out of five games and crossing a new threshhold for machine intelligence. In a Google post and a paper published today in Nature, DeepMind researchers revealed how the system was constructed and how it was able to succeed where decades of previous Go systems have failed.

Go has long been considered one of the hardest games to automate, making the new DeepMind system particularly interesting for artificial intelligence researchers. Facebook is working on a similar Go-playing system, touted by Mark Zuckerberg in a Facebook post just days before the Google announcement.

In broad strokes, Google's new program works through the same tactic that allowed programs like Deep Blue to defeat human players in chess. Known as "tree search," the tactic consists of a simple search of possible outcomes from a given move. But where a chess program sorts through dozens of moves for each possible board position, a Go program must consider hundreds of possible moves, a computing task that grows exponentially with each further step in the decision tree. Google researchers solved that problem by incorporating neural network technology, building "value networks" to evaluate board positions and "policy networks" to evaluate individual moves. That lets the program prioritize specific possibilities, cutting down on the need to explore every branch of the tree.

But while the technology revealed today is conceptually impressive, it's a long way from being fully tested. DeepMind's new system is a clear improvement over previous Go programs, beating 99.8 percent of tested programs, but its record against human opponents is much thinner. The only human opponent so far has been European Go champion Fan Hui, and the two only faced off over five games. It's an impressive record, but a short one, which will make the system's upcoming games even more interesting to watch. The system is scheduled to take on Go world champion Lee Sedol in March of this year.

Verge Video: Quantum computing might be closer than we think

Be sure to subscribe to The Verge's YouTube channel for more videos