AI Learns to Play Go by Watching Humans

DeepMind’s AlphaGo Zero Achieves Breakthrough in Artificial Intelligence

An artificial intelligence (AI) system called AlphaGo Zero has learned to play Go by watching human players for just a few hours. This is a significant development, as it suggests that AI could eventually be used to create artificially intelligent agents that can compete with humans in various tasks.

Go is a complex board game considered one of the most challenging games for humans to master. It is played on a grid with 19x19 intersections, with more possible moves than atoms in the universe.

AlphaGo Zero is a deep learning system developed by Google’s DeepMind artificial intelligence research lab. It was trained on a dataset of 30 million Go games played by humans. However, AlphaGo Zero did not use any human knowledge or training data. Instead, it learned to play Go by playing against itself repeatedly.

After just 40 days of training, AlphaGo Zero beat the previous state-of-the-art AI system, AlphaGo Master, which had been trained in millions of human games. AlphaGo Zero has also beaten the world’s top human Go players.

The success of AlphaGo Zero is a breakthrough in artificial intelligence. AI systems can learn to master complex tasks without human intervention. This could affect many applications, such as self-driving cars, medical diagnosis, and robotics.

The development of AlphaGo Zero is a reminder of the rapid progress made in artificial intelligence. AI systems will likely improve exponentially in the years to come. This could profoundly impact our lives as AI systems become increasingly capable of performing tasks that were once thought to be the exclusive domain of humans.