Artificial Intelligence Google Can Now Learn Like Human

Zatinet - A few months ago, DeepMind's artificial intelligence from Google managed to defeat one of the best players in the Go game. Artificial intelligence named AlphaGo was successfully hit the champion in a three-game series.

Latest, DeepMind successfully developed AlphaGo's successor with more capable capabilities. Named AlphaGo Zero, this artificial intelligence performs better because it does not require data from humans before.

So, this artificial intelligence really learns how to play go from scratch. Although no previous information, in a short time, AlphaGo Zero ability is equivalent to middle class players.

For information, go is a board game and is thousands of years old. Strategy games involving two people are usually popular in East Asian countries, such as China, South Korea, and Japan.

Quoted from The Guardian, Monday (23/10/2017), AlphaGo Zero began studying the go by putting the seeds to the board at random. But after that, he managed to develop the game and even found the best strategy to win.

"The system is stronger than the previous series from not using data including human skills, we chose to get rid of it and the program was able to learn to go alone," said AlphaGo chief researcher David Silver.

According to the research team, the method used in this artificial intelligence is called learning reinforcement. That is, the program can learn its ability from experience during play go.

Despite having more capabilities than its predecessor, the program from AlphaGo Zero was simpler. In fact, this program can master the game go in time faster with fewer exercises and smaller computers.

Share this post