Twenty years ago, the "Deep Blue" computer system defeated the chess champion Garry Kasparov in the public eye, and today, Google has developed an artificial intelligence system that has defeated the world's greatest Go game. Players. The rapid development of artificial intelligence has taken a more important step on the road to defeating the world's greatest super player.
According to Quartz, an artificial intelligence game developed by MIT student Vlad Firoiu has beaten some of the best players in the world. Although this artificial intelligence game did not play against the best player, one of the lost players admitted, "I'm not sure if anyone can beat it."
Although this achievement is very exciting, artificial intelligence still needs some improvement until it can overcome humanity in every game. Artificial intelligence is learning new ways to win games, such as remembering more game moves than humans. They are learning how to think strategically and how to collaborate with each other. This is a good performance of artificial intelligence and human competition, they will certainly go further.
New environment and new development
The "Nintendo Star Wars" is a Nintendo game developed in 2001. In the game, famous Nintendo characters such as Mario and Pikachu can fight each other. Although this game is very old and has already appeared in two sequel versions, it is still favored by a large number of players. "Fighters" continue to develop new fighting techniques to keep the game high.
Because Smash is a fighting game, artificial intelligence faces different challenges compared to chess. In chess, artificial intelligence can simply beat the player by remembering millions of moves. But in the upcoming game Smash, artificial intelligence cannot predict the future because it cannot reliably predict what human players will do in combat. In order to defeat human players, artificial intelligence will have to "learn" how to play Smash.
This is exactly what Firoiu and his colleagues taught about artificial intelligence. They have developed a neural network that can accurately determine the opponent's position and make combat instructions. The neural network will conduct AI training for close combat, allowing the AI ​​to attack itself. By playing with himself, AI learned how to beat the best Smash players in just a few weeks.
In addition to Smash, last month, artificial intelligence also used neural networks to defeat the world's best poker players. Just like Smash, poker can't just win through a calculation. It needs to understand the player's licensing habits, especially the ability to bluff. And just like the Smash artificial intelligence, before entering the tournament, the poker players will play games with themselves rather than compete against opponents. In this way, poker players have won trillions of games. These successes in poker and video games represent the rapid development of artificial intelligence.
But we cannot be complacent. Despite these successes, the development of artificial intelligence still has a long way to go. Because artificial intelligence reacts much faster than humans, Smash artificial intelligence is more aggressive than human players, and can avoid attacks in time. However, artificial intelligence has no ability to handle projectiles in the game, so it can't play against long-range characters. When an opponent hides in a corner, it will be at a loss and even jump to the cliff to commit suicide.
Therefore, some people may question whether Smash AI is really smarter than humans, or whether it just uses superior reaction speed to win. However, artificial intelligence only took a few weeks to eclipse human players who had been practicing for several years. Recent AI wins in areas such as Smash and poker have shown that as game parameters change, artificial intelligence can learn different types of game actions. The two games further emphasized one point: Different artificial intelligence systems can learn to cooperate, and can also compete with each other according to the different nature of the game.
In a sense, any AI program we write can be understood as a game. Whether it is to defeat the player in the game or perform tactical analysis, artificial intelligence is subject to certain parameters. Whether it is through the design of a programmer or through a neural network, it can constantly improve itself. Throughout human history, games are often viewed as a symbol of high intelligence. For example, there are many great leaders who like chess. Because in all kinds of games, artificial intelligence can defeat humans, it shows that artificial intelligence can obtain different types of intelligence through programmer programming. Neural networks and machine learning may revolutionize artificial intelligence, but to overcome humanity in every game still requires considerable progress.
(English source / engdaget compiler / machine Xiaoyi proof / rain eggs)
Fire Resistant Sleeve,Fireproof Sleeves,Fire Resistant Hose Sleeve,Fire Proof Sleeve
Longkou Libo Insulating Material Co.,Ltd. , https://www.liboinsulation.com