AI computing power of go
Publish: 2021-03-25 19:56:37
1. This year is a historic year for go. In January, afar dog first appeared with mixed reputation. March shocked the world and the first world war shocked people. In the face of the great changes that have not happened in three thousand years, it is not beneficial to escape. All we can do is try to understand new technologies and adapt ourselves to these changes
I learned to play go from my father when I was young. So far, I still remember how I was brought into this pit. My father, who was over 30 years old and still in grade two, said to me, "I don't think you can learn to learn. (I bought a watch last year. Which eye of yours can tell that I can't learn? Is it biological. I think you can try to be an athlete (Nani?). I checked, the highest bonus is three items (Bonus! Bonus. The first one is too expensive for us to learn, It's tennis. We can't afford that. Second, you can't learn (and I can't learn?), It's boxing. The third is go. You can learn this. (OK, I'll learn go. Wait, something's wrong. At that time, there was no net, where did you find the bonus ranking? Routine, it's all routine. "
from then on, I began to learn to read books by myself, to see the lines of CCTV5, to see some hanging disk explanation. I started to improve quickly, and soon my father couldn't get rid of me. But to a certain extent, it can not be improved. Reading books with dead and alive topics is either too easy and worthless, or too difficult and confused. I still have a lot of questions in my mind about the reference map that I don't have in the book. Without the guidance of teachers in the layout and operational direction, it is difficult to make a breakthrough after all. Finally, he gave up go and went to study honestly. Now, if I had afar dogs to play with when I was a child, even if it could not reach the professional level, it would not be just today's level
when I was in primary school, I came into contact with a go game by chance. It has been more than ten years since I began to pay attention to computer go. The well-known game programs played are GNU go, silver star go, crazy stone and Zen. The proceres to solve the problems are Yokohama problems. In particular, Zen, from zen4 to zen6, has witnessed the growth of the program. At the same time, he has some experience in using Go program to improve his level more efficiently. The main purpose of this paper is to talk about how to use the "other" in the world to better serve the amateurs who want to improve their level of go< Zen6 is a computer combat software developed by Japanese engineers. Running on a better PC, it can reach the level of 8 or even 9 segments of Yicheng. It can be said that apart from afar dog, zen6 is the strongest
although zen6 is not as strong as afar dogs, zen6 is more suitable than afar dogs for ordinary fans. First of all, they use similar algorithms, both are neural network and Monte Carlo tree search. Secondly, in terms of cost, most people can't afford to play Alfa dog, and zen6 is quite close to the people. Finally, afar dog is too strong, for the general amateurs, the strength of zen6 is more suitable for some.
I learned to play go from my father when I was young. So far, I still remember how I was brought into this pit. My father, who was over 30 years old and still in grade two, said to me, "I don't think you can learn to learn. (I bought a watch last year. Which eye of yours can tell that I can't learn? Is it biological. I think you can try to be an athlete (Nani?). I checked, the highest bonus is three items (Bonus! Bonus. The first one is too expensive for us to learn, It's tennis. We can't afford that. Second, you can't learn (and I can't learn?), It's boxing. The third is go. You can learn this. (OK, I'll learn go. Wait, something's wrong. At that time, there was no net, where did you find the bonus ranking? Routine, it's all routine. "
from then on, I began to learn to read books by myself, to see the lines of CCTV5, to see some hanging disk explanation. I started to improve quickly, and soon my father couldn't get rid of me. But to a certain extent, it can not be improved. Reading books with dead and alive topics is either too easy and worthless, or too difficult and confused. I still have a lot of questions in my mind about the reference map that I don't have in the book. Without the guidance of teachers in the layout and operational direction, it is difficult to make a breakthrough after all. Finally, he gave up go and went to study honestly. Now, if I had afar dogs to play with when I was a child, even if it could not reach the professional level, it would not be just today's level
when I was in primary school, I came into contact with a go game by chance. It has been more than ten years since I began to pay attention to computer go. The well-known game programs played are GNU go, silver star go, crazy stone and Zen. The proceres to solve the problems are Yokohama problems. In particular, Zen, from zen4 to zen6, has witnessed the growth of the program. At the same time, he has some experience in using Go program to improve his level more efficiently. The main purpose of this paper is to talk about how to use the "other" in the world to better serve the amateurs who want to improve their level of go< Zen6 is a computer combat software developed by Japanese engineers. Running on a better PC, it can reach the level of 8 or even 9 segments of Yicheng. It can be said that apart from afar dog, zen6 is the strongest
although zen6 is not as strong as afar dogs, zen6 is more suitable than afar dogs for ordinary fans. First of all, they use similar algorithms, both are neural network and Monte Carlo tree search. Secondly, in terms of cost, most people can't afford to play Alfa dog, and zen6 is quite close to the people. Finally, afar dog is too strong, for the general amateurs, the strength of zen6 is more suitable for some.
2.
The 2018 China go Conference opened in Nanning, Guangxi on August 8. There are 26 go competitions in the conference, including the national go competition, the national go team championship, the national go King competition, the world intelligent go open competition, the nine way go King competition, the marathon go competition, and the Traditional professional competitions such as parent-child double competition and interesting competitions close to people's life. Among them, AI go competition and Instrial Application Exhibition become the focus of this conference“ The current competitive level and popularization of Chinese go are in the forefront of the world
in this process, "artificial intelligence + go instry application" will be an important direction of go instrialization in the future“ Like all sports in China, go will develop towards instrialization in the future. " Qin Yonggang, general manager of Hua City Weilian Sports Instry Co., Ltd., said that as the first fully market-oriented "city go League" in China, after more than three years of development, the members of the city go League have expanded to 32 clubs, with teams from 31 cities in seven countries. It has become a brand game with a certain scale and high influence
3. AI has realized the pursuit of human beings for the realm of go: to be simple, to win without fighting, to make the best use of the situation and to make no mistakes. Because of the lack of computing power, it's no use learning any more. You can be a companion to broaden your mind.
4. It is mainly strong in two aspects:
1. Powerful violent computing power, that is, mathematical computing power. The computer surpassed human beings one or two hundred years ago, let alone now. This ability enables the computer to have an advantage over human beings in local computation, and the local exhaustive method can be implemented. In this way, in the fight, especially fast chess, there will be a significant advantage over humans
2. Super high learning efficiency. This is the highlight of AFA go and what Google wants to show. If I remember correctly, AFA can play thousands of chess games a day. The number of disks in a week may exceed that of Ke Jie's life. Even though the thinking and summarizing ability of each game is much weaker than that of human beings, the efficiency is far beyond that
in my opinion, if Ke Jie can play half or even one tenth of AFA's sets and keep such a young state, he will definitely win. Unfortunately, according to people's life expectancy and physiological needs, this is impossible.
1. Powerful violent computing power, that is, mathematical computing power. The computer surpassed human beings one or two hundred years ago, let alone now. This ability enables the computer to have an advantage over human beings in local computation, and the local exhaustive method can be implemented. In this way, in the fight, especially fast chess, there will be a significant advantage over humans
2. Super high learning efficiency. This is the highlight of AFA go and what Google wants to show. If I remember correctly, AFA can play thousands of chess games a day. The number of disks in a week may exceed that of Ke Jie's life. Even though the thinking and summarizing ability of each game is much weaker than that of human beings, the efficiency is far beyond that
in my opinion, if Ke Jie can play half or even one tenth of AFA's sets and keep such a young state, he will definitely win. Unfortunately, according to people's life expectancy and physiological needs, this is impossible.
5. There is a time length T, beyond which the human can calculate more steps than the computer. On the chessboard, that is to say, people's computing power is stronger than machines. In fact, this is a very natural inference, but we are shocked by AI's recent achievements, and no one has mentioned it. Of course, if the length of time is one step a year, then there is no practical significance, but I don't think it is. My inference is supported by a little data. In the paper published by the alpha go group at the beginning of this year, the distributed alpha go with more than 20 times of computing power won 77% over the stand-alone version. If two people fight each other, should the 20 times time difference be more? In addition, there is now a version of Zen. Although it is far from powerful, the reasoning used to verify my inference is the same
6. In my opinion, it's just special computing software, not intelligence. In fact, go players don't compare intelligence and intelligence. Go ability is just an abstract computing ability in intelligence. When it comes to intelligence, it should be a comprehensive ability to solve practical problems.
7. The single version of go AI intelligent software needs good computer configuration to give full play to the best performance. At present, the highest level software can be found on the Internet is Tianding go 6. Youye 5
8. No special configuration is required
there is no special requirement for local configuration of AI go like Jueyi
all data are sent from the cloud.
there is no special requirement for local configuration of AI go like Jueyi
all data are sent from the cloud.
9. Go is available on the mobile phone. If there are some software, it can download go, and then it can be used by employees or smart phones
10. The previous AI of go imitated another chess player, who chose the easiest way to win according to the previous experience
alpha go is a simulation of the whole history from the invention of go to today. From the first game of go (which must have been very elementary), it has been graally improved to today. Now it has been simulated to the level of professional rank, and it will soon surpass today. It is equivalent to simulating the development of go with a machine and planning the formula of the whole game, just like the formula of go, In fact, it's the best way for human beings to set up. There are dozens of steps in the grand formula. Will the computer give you a 250 Hand formula Let's wait and see. The day when the 250 Hand formula comes out will be the day when the full value of the game of go will be explained.
alpha go is a simulation of the whole history from the invention of go to today. From the first game of go (which must have been very elementary), it has been graally improved to today. Now it has been simulated to the level of professional rank, and it will soon surpass today. It is equivalent to simulating the development of go with a machine and planning the formula of the whole game, just like the formula of go, In fact, it's the best way for human beings to set up. There are dozens of steps in the grand formula. Will the computer give you a 250 Hand formula Let's wait and see. The day when the 250 Hand formula comes out will be the day when the full value of the game of go will be explained.
Hot content