Chapter 286: The Shattering Myth

Wei Dongsheng: "The 42.2% winning rate is the absolute winning rate, which is limited to the absolute winning rate of Lanke Go's limited knowledge. Lanke Go assumes that the opponent's chess power is comparable to it. After the 78th move, the opponent's winning rate can be increased to 42.2%. That is, It is said that the fatal error of the 78th move has limited impact on the overall situation. Even if the opponent changes to Lanke Go, it is difficult to turn defeat into victory. Perhaps Lanke Go judged like this: The previous dozens of pieces layout, it is already incomprehensible by human players Under the circumstances, it quietly harvested at least 9 potential advantages."

"What is Lee Sedol's true winning percentage?"

"Lanke Go created several major data models for Lee Sedol: the Lee Sedol data model in 2010, the Lee Sedol data model from 2007 to 2010, the Lee Sedol data model at their peak, the enhanced Lee Sedol data model, the dynamic Lee Sedol data model, etc. ."

"The 2010 Lee Sedol model and Lanke Go include Lee Sedol's game history from May 2009 to May 2010. Compare it with Lee Sedol's general database and analyze Lee Sedol's recent state in detail. Perhaps affected by last year's suspension of work, Lanke Go judges Li Shishi’s chess power has declined significantly recently and is quite out of shape. Among the four major models, the Li Shishi model had the lowest winning rate in 2010. Even after Lan Ke Go’s 78th move, his winning rate was only 1.3%."

"From 2007 to 2010 was the peak period for Lee Sedol to dominate the world of Go. During this period, Lee Sedol's data model was relatively strong. After Lanke Go's 78th wrong move, its winning rate increased to 4.7%."

"At its peak, the Lee Sedol data model only collected high-quality games recognized by the Go world, and created a virtual Li Sedol who could play high-quality games every time. After Lanke Go made 78 wrong moves, its winning rate increased to 11.9%. "

"The enhanced version of Lee Sedol's data model, big data analysis of Lee Sedol's chess power evolutionary trajectory, to correct those imperfect Go concepts that are easily slain by dragons, and predict that he can make breakthroughs in certain aspects. After Lan Ke Go's 78th move, it The winning percentage increased to 18.6%."

"Dynamic version of Lee Sedol's data model. Lanke Go frequently compares and analyzes the current game and Lee Sedol's past games, and predicts Lee Sedol's next move in reality. If Lee Sedol wants to win the game by relying on past patterns that are effective but have loopholes in experience, Lanke Go It will quietly set ambush while ensuring the winning rate; if Li Shishi makes a self-breakthrough during the game, Lanke Go will immediately record those benign changes and use other computing resources to calculate the impact of the benign changes on the Canadian Li Shishi data model."

"The peak data model will help Lanke Go understand the past Li Shishi, the enhanced data model will help Lanke Go understand the future Li Shishi, and the dynamic version of the data model will help Lanke Go understand the current Li Shishi. Lanke Go Eyes. In the seven rounds of the game, in the first and second rounds, Li Shishi was completely out of state, especially in the second round, and his state sank to the bottom; in the third, fourth, and fifth rounds, Li Shishi gradually recovered his state and rose to the peak. Period; In the sixth game, Li Shishi was in a sluggish state; in the seventh game, Li Shishi was in a very good state, with a sense of self-breakthrough."

"Etc., etc."

"Lan Ke Go knows Li Shishi better than we thought."

"Someone may ask, for Li Shishi, is Lanke Go or the Go program? Do you need to re-adjust the data model against other players? If you change a player, you need to readjust again. Lanke Go is too mechanically rigid. Up."

"The answer, of course not."

"If it is so rigid that Spring and Autumn Search has the face to call itself artificial intelligence?"

"It is the enhanced Li Shishi data model and the dynamic Li Shishi model. It sounds complicated. In fact, in actual operation, Lanke Go is very simple and fast. All data models do not need to be entered manually."

"Lanke Go has a complete set of data model establishment methods. It can generate a series of data models in real time by scanning Li Shishi's past game records. If Lanke Go is authorized to connect to the Internet Go database, it can also use "Li Shishi + "Go" and other keywords search for Li Shishi’s information and automatically extract previous game records. In other words, even if the opponent is changed to Li Changho, Coulee, Kong Jie and other players, Lanke Go can quickly generate a series of enhanced versions of Li Changho and strengthen Data models such as version Coulee."

"We say Lanke Go is artificial intelligence, not only because it can win against Li Shishi and conquer the myth of Go, but also because it can automatically collect data, automatically process data, automatically build data models, and efficiently and specifically analyze opponents’ Go concepts and ideas. Go thinking."

"While we human chess players are looking for loopholes in Lanke Go, Lanke Go also uses cutting-edge technologies such as big data analysis to find loopholes in top players such as Li Shishi."

"Lanke Go can't be perfect enough to have no loopholes, but how can we humans be perfect enough that there are no loopholes?

"As long as Lanke Go can beat Li Shishi, it will win!"

The Lanke Go of Spring and Autumn Search and the Google AlphaGo in 30 years of memory, although both are Go programs, are completely different in nature.

Although he cannot compete on the same stage, Wei Dongsheng believes that the Lanke Go he programmed himself can easily defeat Ke Jie's AlphaGo in May 2017, because the two Go programs are not at the same level. Wei Dongsheng overlooks the concept of deep learning with the computer intelligence life knowledge system, and his understanding of deep learning far exceeds the Google team; even if he does not use the black technology of the computer intelligence life knowledge system, his practical ability is ten times stronger than the Google team.

In addition to the gap in chess power, Lanke Go's artificial intelligence is renamed to be veritable.

When you are looking for its loopholes, it is also looking for your loopholes.

Hearing Wei Dongsheng's famous saying, which was somewhat alarming, the audience could not help but be shocked.

Some people are slow to wake up to the cruel truth: we think of wisdom too illusory, and we set artificial intelligence standards too high.

Perhaps the threshold of wisdom is not as high as imagined.

Perhaps the term wisdom is not so philosophical.

Does a computer program autonomously search for loopholes in human chess players, is it intelligent?

Deep learning and other related technologies are used to improve their ability to determine the value of their own, big data analysis and other related technologies are used to observe and find the weakness of the target, the new computer program represented by the Lanke Go program has been able to carry out a certain degree of offensive and defensive operations against humans . It is so weak, but it brings a possibility in the future.

As Wei Dongsheng showed, Lanke Go has established countless data models to analyze Li Shishi’s weaknesses and even predict the direction of Li Shishi’s self-breakthrough. How does Li Shishi plan to win it?

Some people may say that Li Shishi's first line of life is constantly seeking innovation.

Li Shishi broke through the limit and reached the unknown territory of Lanke Go, and he was powerless.

However, this naive idea is obviously optimistic.

Lan Ke Go can be trained to play 200 million chess games in just half a year, and can analyze more than 700,000 chess games against Li Shishi in just half a month. The efficiency of Lanke Go is so high that its speed to adapt to the new game is bound to be ten times, a hundred times, a thousand times, and ten thousand times faster than Li Shishi's continuous self-breaking speed.

think about it.

Li Shishi worked hard for a year and finally ushered in a self-breakthrough. Kolan Ke Go quickly followed him, efficiently deduced hundreds of thousands of games and analyzed the value of Li Shishi's self-breakthrough, and integrated Li Shishi's self-breakthrough achievements into its own value network system in just one day.

The vision expanded to the entire Go world.

Lanke Go closely follows all the top players such as Lee Sedol, Lee Changho, Gulei, Kong Jie, Park Yeonhuan, Ke Jie, and every top player's interpretation of the unknown field of Go can capture its value network system in time. Chess players such as Lee Sedol, Park Yeon-hwan, Ke Jie want to fight Lanke Go, like a struggle between individuals and collectives. They not only have to win the deep learning results of Lanke Go’s self-play, but also the entire Go world against Go Century contribution.

It will learn by itself and is better at learning from humans.

While humans are scattered, artificial intelligence is one.

This is the advantage of artificial intelligence, and the horror of artificial intelligence.

If artificial intelligence can learn human knowledge, even if it “lack” self-learning and innovation ability, it can become a collection of human beings like Gaia consciousness and Alaya consciousness in fantasy novels. Fighting with an artificial intelligence is to compete with all previous achievements of the entire human society, unless the human race invents new weapons that can destroy artificial intelligence in a very short time, and destroys them before the artificial intelligence understands, learns, and absorbs the new weapons. Artificial intelligence, otherwise humans have little hope of defeating artificial intelligence with an infinite number of advantages.

This may be imaginary, maybe still far away.

However, the crisis has begun to breed.

Lanke Go has been able to find the weakness of a specific target by analyzing the game with big data. If you change the scene and logic, can it find the "weakness" of the specific program by analyzing the source code of the big data program?

The program code looks very deep, but in fact it is very simple.

Programmers should have a deep understanding of this, after all, dealing with code is much easier than dealing with people.

Especially for those non-innovative codes, it is just like farming as a manual labor. The term "code farmer" is really both vivid and appropriate. Theoretically, if Lanke Go can deep learning to discover the unknown Go concept of human players, it will inevitably also deep learning to increase the data analysis method to analyze the source code of a specific program, small it can quickly find bugs, and large it can optimize The efficiency of program execution.

Just like the Go myth, the myth that humans can write programs will eventually shatter.

Of course, Lanke Go certainly cannot update its source code now.

But what about perfecting it?

What about new theories similar to deep learning and big data analysis?

If you like to re-take the road of the future, please collect it: (lightnovelpub.net) Re-take the road of the future. The literature is updated fastest.