Open: How does artificial intelligence "roll down" humans?

However, in the field of competition, the superiority of artificial intelligence compared with human intelligence has become more and more obvious. In May 1997, Kasparov lost to IBM's computer program "Deep Blue", which opened up the public eyesight of artificial intelligence in the field of chess and athletics.

However, with the rise of the Internet, AI gradually entered a period of silence after the new millennium. Until 2006, big data and deep learning broke out again, and it once again entered the public eye. In the field of athletics, this artificial intelligence has chosen the more difficult areas of Go, Defo, and games.

So, in these areas, how does artificial intelligence "crush" humans?

"Deep Blue": My crushing of human wisdom is only the beginning of AI

Many people think that in 1997, Deep Blue won Kasparov as the first time artificial intelligence defeated humans. But in fact, it's not. As early as 1988, the deep blue brother "thinking" has already won in the game against the chess master. It is symbolic that Deep Blue is the first artificial intelligence to win the world championship in chess. In fact, in the game of February 1996, Deep Blue lost to Kasparov. He also stated that the computer would have to wait until 2010 to win the world championship. As a result, this moment actually only waited for one year and three months, showing that the ability to learn artificial intelligence is not judged by human intelligence.

"Deep Blue": My crushing of human wisdom is only the beginning of AI.

AlphaGO, great art, DeepZenGo: In fact, we have disdain to compete with humans.

After "thinking" and "dark blue", the artificial intelligence broke out again and shifted the focus to Go. This time, unlike the "deep blue" dominates the world, the artificial intelligence of the world of chess is "a fierce battle."

In fact, initially AlphaGO was anonymous in the network until it finally challenged Li Shishi and eventually won the public. As for why AlphaGO can defeat human beings, in addition to the data processing capability similar to Deep Blue, its deep neural network can predict the possible situation after the longest 40 steps, and can judge the optimal result at any time by virtue of computing power. Even if the start is unfavorable, or in the case of a child, you can control it globally.

In Asia, the most interesting thing in 2017 is the “Tear of the World” by Deep Art and DeepZenGo. Among them, the two most attracting public attention occurred in March and August respectively. In March, the 10th UEC Cup World Go AI Competition was held in Japan. In the final match against DeepZenGo in the finals, the game was unexpectedly reversed in the overall situation and eventually won the UEC Cup for the first time. In August, the first World Smart Go game, the original winner of the championship, was defeated by the half-way in the quarter-final, the "CGI" from Taiwan, and finally won by the old rival DeepZenGo. Although AlphaGo, fascinating, and DeepZenGo have not yet come to an ultimate showdown, their "river status" is still promising. After the launch of the advanced version of AlphaGoZero, it passed the self-matching game within three days, surpassing the strength of AlphaGo and winning the 100-game winning. After that, DeepMind published a curtain call for AlphaGo. Indeed, there is no rival in the world of Go, and DeepMind, who is still in the cold, has targeted its next opponent to breast cancer.

Defu: artificial intelligence opens the way to open new fields

In December 2017, Science published an article saying that American experts have developed a new artificial intelligence system that beats four top players in Texas Hold'em. According to Noam Brown, an expert at Carnegie Mellon University, “The technology we developed when we created this artificial intelligence system was not designed to play poker. It can also help solve other problems under 'imperfect information'. And this kind of problem can be seen everywhere. We hope that the concept adopted in the development of Libratus will play a decisive role in the further development and promotion of artificial intelligence."

However, why artificial intelligence can play the slap, which is not only different from the big data analysis and processing ability of chess and Go. The main methods used are game theory and operations research. The key needs to understand four concepts: Nash equilibrium strategy, Counterfactual best strategy, abstraction, and finality. Taking Nash equilibrium as an example, how to estimate the probability of the betting strategy of the opponent in the case of information asymmetry, so as to obtain the dominant strategy in the game matrix.

"Deep blue", "Go chess hegemony", "Depu" give human thinking:

The reason why Deep Blue can make the world shocked is because its outstanding performance has raised a philosophical question for human beings - what is true thinking? The average operation speed is 2 million steps/second, and the exponential super human ability for the chessboard. But without the sensibility of volatility and intuition, it is completely calculated through data. Is this really a reflection? In addition to excellent speed, what is the value of this thinking ability? Should humans also let their intelligence develop in such a scale?

These thoughts have continued to "go chess hegemony."

On the other hand, the application of artificial intelligence in the spurs shows that in addition to its good performance in the field of statistical thinking such as data calculation, it also has excellent performance in social and economic aspects such as game equilibrium.