(Original title: [study] Fang Jun's column: AlphaGo vs. Ke Jie: Watching the point is not the winner but the match match)
Starting today, Google’s AlphaGo will launch a series of matches with Ke Jie and other top Chinese chess players in Wuzhen. This game is not much suspense. Artificial intelligence (AI) has far surpassed humanity in the game of Go. Ke Jie himself lost two innings at the beginning of the year with the Alpha Dog online fast chess game, dubbed "Master."
From the perspective of the game system, the most interesting aspect this time is not whether AI can overcome humans, but what happens to the AI ​​Federation. This time the game was divided into three groups: Ke Jie and AI played three games; the human player was paired with AI. The two sides of the game were Coulee + AlphaGo and Lian Xiao + AlphaGo; five players teamed up against AI. Ke Jie's unique battle, team battle, human chances are very small. What really shines is the match between the human chess player and the AI. Matching matches can test a key question: Should people follow the judgment of AI teammates, or do they believe in their own instincts and analysis?
In the past few decades, computer-to-player brains such as deep blue computers played chess champion Kasparov, AI battles against humans such as Alpha Dog vs. Li Shishi, and Ke Jie, which have become hot news. But this kind of confrontation has never been the essence of the relationship between technology and people. For example, most people now think that people and cars run faster than anyone else. Even the value of entertainment is gone. How to make use of cars and use unmanned vehicles to make our lives more convenient is more valuable. The direction of exploration.
Now, we generally only remember that Kasparov lost in a deep blue computer game in 1997 and humans lost to computers. In fact, we should not forget that in 1998 he conducted a special experiment. He wanted to answer: What if the computer is not a human opponent but a human assistant?
He organized a special chess game called "advanced chess." Both sides of the game are "people + computers," that is, from the people and the computer to the dry, into the side by side with the computer. The benefits of teaming up with computers are obvious. If you have a computer to do the calculations, you can not only avoid mistakes, but also because the players do not have to do a lot of calculations, so that they can focus on in-depth strategic planning.
The combination of man and machine has both benefits and challenges. When people associate with computers or, as they say now, when people associate with AI, the question we may face is: Can you understand the process of AI selection? How does it differ when the judgment it makes is different from yours? Kasparov wrote in his autobiography, "Chess and Life," that "almost all activities in modern day-to-day life require the use of increasingly sophisticated tools that extend the issues raised above to all aspects of life."
Actually, artificial intelligence represented by machine learning is now a technique that we cannot yet understand its inner workings, and this may be the first time in the important technology invented by humans because the operational logic of machine learning is to make the machine autonomous. Learn. We learn algorithms and data for it. The results it gives are getting better and better. We can recognize people's faces, recognize cats, and do translations. But now we can't explain why clearly. Machine learning is now a "black box" that is becoming more and more powerful. We teach it a lot of learning algorithms, but we don't know what's going on inside.
Since the Alpha Dog raced with Li Shishi, people have analyzed many of its operating principles. For example, it uses two types of networks to form their own choices: to understand the "value network" of chess games and to select the "strategy network" for the next move. . However, we do not know what the exact reason for its choice at each step is. This time, when a human player joins up with an Alpha dog, when a player cannot understand the Alpha Dog's judgment, he cannot understand why it is so determined. What should a player do?
Humans have been thinking about the relationship between people and technology, but in fact most people did not think about this issue. In modern society, people will unconsciously believe in experts. When experts look like high-tech machines, it is very easy for people to unconsciously follow their instructions. The question is coming. Will you believe AI's judgment without thinking? Will you turn to AI if you have difficulty? We can take a look at this game and see what the top players will do.
We do not yet know the specific details of the match between AI and the Go in this Go game. Now we only see the match information. The key to match match is the player's understanding and cooperation with AI. But we have already seen that Kasparov's 1998 test was so conceived and proved that a better future may be the combination of man and machine, man and AI. There is one point that AI has surpassed Kasparov's idea. At that time, he thought that the combination of human and machine is better than that of humans or computers. "The combination of human strategic guiding ability and computer's strategic acumen is unbeatable." On Go, Alpha Dog has proven that he has strategic capabilities.
What is valuable in this competition is to inspire people to think and explore the possibility of “man + AI†more. The relationship between man and AI is not confrontation and transcendence. Humans have been inventing technology to replace some of their own functions. The mission that Google's brain sets for itself is, "How to make machines smarter and how to make people's lives better." I think we can add another one. How do people associate with the machine? Of course, this third article cannot be added because it has always been the deepest relationship between people and technology.
(The author is an internet observer)
Outdoor low power LED lamps,Aluminum Alloy LED lamps,LED point light
Kindwin Technology (H.K.) Limited , https://www.ktl-led.com