Go: Post-Mortem Humanity Go: Post-Mortem Humanity

By Antti Hahto


Historically, gaming has been the second of two big facilitators, progressing digital technology onward. The other; yeah, you already guessed it.

In this sense Go (aka Baduk, Weiqi), an ancient board game having history of thousands of years, has been somewhat peculiar holy grail for cutting-edge A.I. researchers. Among Western world, it is considered every civilized human being to know at least something about Chess. Yet meantime, not too many have heard of (not to mention tried!) the more popular of the siblings.

Some time ago, a historical 4-1 victory over humanity happened. Being in, no less than, a game of Go. Why was it so important? Short answer: imitating human intelligence, instead of brute force, “cold” calculation.


Do Androids Dream on Newest Apps? How do we know when a machine can be considered as intelligent? What is intelligence, anyway? To be honest, we kind of don’t know. There’s a lot of, by all means very intelligent, classic definitions, but eventually through pure narcissism it boils down “to behave such as a human would”.

Even though a chess computer nowadays beats humans by left… circuit, we don’t consider it to be very intelligent anymore. The same things happens all over again; when machines gradually take over on human jobs, the goal of intellect behavior always conveniently moves one step forward. “But a machine is never capable of doing art, to dream and to reason!” Well, it’s already more or less doing so, and at increasing speeds.

Main problem in SETI, search for extraterrestrial intelligence, is, that we only recognize features by comparison to something already familiar with us. Instead of going space, we could visit our backyard (if having one), sit down and observe. Ants can build rafts, bridges, utilize tools and practice agriculture. A simple fungus or parasite can command host in a very complex ways, affecting behavior to reach its own sinister goals. Plants communicate with each other and if threatened, can release a “cry for help” for example to predators being natural enemy of caterpillars. A slime mold can find shortest path inside a labyrinth. List goes on and on, kudos to David Attenborough!


Apart from building bridges to nature-inspired methods, back to Go — a game used in training military strategies. Main reason for using it as a machine learning sparring opponent mainly being, it hasn’t been possible to make a very good Go player with traditional approaches by (more or less) trying to simulate all the possible outcomes of every move. A professional Go player has an “intuition”, “feeling”, of good moves, essentially doing pattern recognition task of the board state. It happens, humans have traditionally been very effective on pattern recognition; ask from any website using I’m-Not-A-Robot verification (CAPTCHA). It also happens, soft computing techniques such as neural networks excel in this very same area compared to more traditional software methods (reasons for this being a topic of another time). Maybe, as good pattern recognizers we are, there is that intuition — being it false or not, a glimpse of something very human out there.

We at Intopalo see recent advances and course of direction exciting, but not really that unexpected. Most of the methods have been there for a very long time, but only being gradually enabled by the rise of parallel computing and Big Data collecting. Many companies have already collected a huge amount of data, be it sensor readings, ERP or marketing statistics, and are waking up to find some good use for it.


Go is a game of perfect information, where everything is visible on board. When the information is not there, being a very realistic scenario in business world, some human help is still needed. Until next version.