AI: Rage against the machine

A computer called Deep Blue that shocked the world by beating Garry Kasparov at chess game in 1996 could today lose against your iPhone. Now artificial intelligence is giving computers even greater powers to vanquish humans in computer games.

 

AI versus Humans

When the best players in the world competed in “The International”, the Dota 2 championship, in Seattle this year, the prize money was close to $25 million, with the winning team taking $10.8 million. The game allowed players to choose between 113 characters with specific abilities. This made strategy very difficult to predict. Teams of five players fought each other with the aim of destroying their opponents’ towers. Elon Musk's OpenAI learned how to play a simplified version of Dota 2. It managed to easily beat the best players by inventing strategies that had never been seen before. For example, one player fled from the action several times. This proved a very bad move: players later discovered that fleeing from pursuers triggers a more aggressive behaviour and allows for powerful counter attacks. Human players have adopted this strategy allowing them to beat the OpenAI since then!

 

 “Computers revisit old games and discover mind-blowing moves”

Computerised Grandmasters

In Garry Kasparov’s new book, “Deep Thinking”, the chess Grandmaster explains how, by taking care of repetitive tasks, machines allow humans to focus on complex missions where they can add value. Chess is a very special game in this regard: players can access all the past games of their future opponents with a click. Computers then take seconds to calculate a strategy to defeat potential adversaries: a task that would take a team of Grandmasters several days. Computers revisit old games and discover mind-blowing moves that no one had previously thought of playing. In this way computers are making humans better chess players and, by saving them days of calculations, allowing them to focus on the bigger picture.

 

Stopping GOs

The ancient Chinese game, Go, is certainly a most instructive game when it comes to AI. Last year, the computer programme, AlphaGo beat one of the world’s best Go players, Lee Sedol. However, AlphaGo lost a few months later against an updated version of itself: 0-100! This was made possible by radically changing the approach: the first version analysed millions of human games, while the second was only given access to the basics moves of the game. AlphaGo therefore thought it was playing against itself. Programmes like this can play a game in the blink of an eye but can also handle many games simultaneously. To remove the confusion between algorithms and AI, the latter is often described as the ability that a computer has to write its own code. But AI can cause conflict between humans. After Facebook’s Mark Zuckerberg was forced to halt an AI experiment because he considered it was dangerous, Elon Musk Tweeted: “I've talked to Mark about this. His understanding of the subject is limited”. End of the conversation for now!

 

 

 

 

 

 

 

 

© 360Crossmedia

Cover.png