The Alphabet of Stones

What a computer program’s victory over a human in the game of Go could mean for the future of geopolitics.

Written by Rohit Gupta | Published:March 27, 2016 3:00 am
Google, AlphaGo, Go, artificial intelligence, Deep mind, Google artificial intelligence, Go game, Chinese game, Uber, Ola, technology, technology news What a computer program’s victory over a human in the game of Go could mean for the future of geopolitics.

The consequences of Korean player Lee Se-dol’s historic defeat against a computer program in March 2016 will be both global and political. One reason is that the ancient and revered board-game Go (some claim it was invented by a Chinese emperor around 2300 BC) — in its very essence, is a profound meditation on the art of war.

There are only two types of stones in Go, black or white — reminiscent of zeroes and ones in digital computers. Contrary to the hierarchical pawns, bishops and kings in chess, the pieces in Go are identical and theoretically equal in value, somewhat analogous to people in a communist regime. The aim is to capture territory and annihilate the enemy stones by surrounding them.

When you compare the presence of black cabs on a map of Uber with the grey-white cabs of Ola, the shadow of this ancient game is staring back at you in the form of a corporate conflict. Black and white do not represent just two colours, but any two forces of nature in opposition with each other — countries, religions, poems and nuclear weapons, love and hate, poverty and plenty, light and darkness. (Too many black Uber cabs in the same area are not good either because they tend to suffocate each other). One of the most potent elements in Go is empty space: when placed wisely, the absence of a single stone can radiate the geometry of an impenetrable castle throughout the 19×19 grid.

The utter simplicity of its rules is in sharp contrast to the bewildering complexity of Go. Dutch computer scientist John Tromp noted that comparing Go to chess is “not even like comparing the size of the universe with the nucleus of an atom”. As the game progresses, the smallest error made in this dynamic universe of Yin and Yang can magnify surreptitiously into an irreversible cataclysm. A butterfly flutters its wings, slaves mutiny on a ship, corporations go bankrupt, the Soviet Union breaks apart, a black asteroid strikes the earth and dinosaurs go extinct.

Artificial intelligence too, like this ancient boardgame, goes back to the very dawn of human civilisation. A Roman tutorial on rhetoric for orators called Ad Herennium (86-82 BC) says this about memorisation technique, “…and now we will speak of the artificial memory.” Many Greeks, including Socrates, were against the invention of the written word, because they feared it would destroy the ability of human beings to remember. Millennia later, the rise of computers has released the art of memory like a gigantic djinn from Aladdin’s lamp, just as telescopes opened up the horizons of astronomy.

Our world presents designers of artificial intelligence (AI) with some thorny and diverse challenges. A self-driving car requires fluid control of feedback loops from its sensors; a chess algorithm tries to see ahead of the opponent by mapping all possible moves; and Go requires neural networks or the “Monte Carlo tree search” (MCTS), since the possible outcomes are far too many to anticipate, even for a supercomputer. These algorithms are probabilistic, in that they sacrifice certainty for overall performance.

Nonetheless, Google’s DeepMind made an explicit demonstration of its superiority in this technology with the March 15 game. The stock value of its parent company Alphabet jumped from $713 by 5.19 per cent to $750 after the event, and this digital “arms-race” has only just begun. (The common argument that computers will take away human jobs is easily annulled when you show how many people across the world are today employed as software engineers.)

Lee Se-dol’s defeat was, in fact, a victory for Go, because the match was followed online by its largest audience in history. Both Fan Hui and Lee later testified that they learned new moves, one of which helped Lee defeat DeepMind’s algorithm in the fourth game. This symbiotic relationship between AI and human beings may well be the next step in the evolution of society.

The scenarios in which artificial intelligence can now be deployed is staggeringly vast in terms of infrastructure — the management of vehicular traffic, personalised teaching assistants, medicine and health, high-speed stock trading and eventually, the automation of entire economies and ecosystems. Who can say that AI, in a not too distant future, will not replace democracies with more intelligent and dynamic constitutions? We are already witnessing a world where populist elections tend to result in despotic and fascist governments.

Imagine cybernetic swaraj instead — a self-regulated democracy where we vote perpetually for ideas, laws and amendments. The constitution itself is an operating system, ensuring equality for everyone. All citizens are members of its parliament, but there is no fallible human at the apex, no one at all in the driving seat. After all, if a car can drive itself with artificial intelligence, why not a country?

Rohit Gupta is an independent historian of science and mathematics. He tweets as @fadesingh