post-AI era

Original link: https://yufree.cn/cn/2023/07/19/post-ai/

If you want to know the impact of a certain technology, you must first observe the fields that adopt the technology first. For artificial intelligence, this proposition is Go.

Go is a complex system under simple rules, which can be compared to cellular automata. The difference is that Go is a finite game, that is, there is a clear goal and there must be a result within the maximum period. This kind of problem is actually the favorite type of computer, and the calculation that limits the final result is an optimization problem. In reality, many occupations can also belong to Go problems. The difference is that the rules are more complex, and the system or computing space is more open. But as long as your purpose is clear, there is always an algorithm.

The 391-frame game of Go pays attention to the three stages of layout, middle game and officials. If it is an analogy to life, it is equivalent to the three stages of pre-employment, work and retirement. Just like in Go, many people just cast in the middle of the game, and many people win the game at the official stage. If you insist on giving a report card, you will find that most games cannot reach the official stage. Although many people naturally set their goal on a good retirement life, middle-aged people in the mid-range will end up early because of the unpredictable behavior of their opponents or the times, and live a life that can be seen at the end.

However, this does not change the good hopes of most people for the layout stage. They all hope for a good start. Many parents even sum up some fixed routines to pave the way for their children. This is called a fixed pattern in Go. In the enlightenment stage of Go, in addition to doing life and death problems, you have to memorize the set formula. A traditional set formula may not be successful in the final game, but because the players are also memorizing the set formula, at least everyone will not lose or win before the middle game. The comparison of life is that everyone receives education. Some may learn quickly and some slowly, but at least what they learn is the same. Few games end at the formula stage. At this time, everyone follows the routine step by step. Even if someone knows more formulas, it is impossible to accurately predict how effective they will be.

The fight in the mid-game of Go sometimes benefits from the predictability of the formula, and sometimes gets caught in the ever-changing chess game and can’t figure it out at all. Comparing life is that the knowledge you learn is sometimes useful, sometimes it is not enough at all, and your perspective is even limited by what you learned at the beginning. If your opponent is this world, everyone will play common hands and good hands, and some people will be slaughtered without paying attention in the middle game, and a large number of accumulated eyes will be reset to zero in an instant. For professional chess players, most of them will admit defeat at this time, and at worst, restart the game. But for specific people, this may mean the death of a family or a heavy debt or a prison sentence. There is no way to restart life. At this time, it is actually painful every step of the way. However, as long as you are still playing, you may encounter the opponent playing the spoon, that is, the opponent has made a bad move or made a bad move. At this time, you can seize the opportunity and fight back.

When it comes to the official stage, although the rules of each country are different, both sides of the game have begun to nod secretly. In fact, the end of the official chess stage is not far away. Many experienced professional chess players already know the result at this stage. Of course, there are many cases of playing in the official chess stage. At the end of the game, if Caution is as good as the opponent, the outcome should be within half an eye. At this time, it is difficult to say who wins, it is purely luck and rules. The retirement stage of life is actually similar, and I don’t care about the final victory or defeat when I am alive, I only care about whether the game of chess is wonderful.

The above is about Go before 2016, and the following is about the current situation. First of all, artificial intelligence is the same as human beings. It also learns the fixed pattern and classic game first. However, after training to a certain level, artificial intelligence introduces enhanced learning, that is, it plays against itself. Many people began to review past chess games and let artificial intelligence evaluate the quality. It turned out that many players who used to be considered bad players were good in the eyes of artificial intelligence. The most famous of these is the dot-three-three layout. Now it is common to see dot-three-three in the opening game, but before 2016, few people started with dot-three-three. Without him, Go education at that time believed that 33 is not conducive to future development. In terms of formulas, Shen Zhenzhen, the current number one Go player in the world, is called “Shen Gong Intelligence”. It is said that he often trains with artificial intelligence, and he may have memorized countless artificial intelligence formulas in his mind, and every step of the game is consistent with the first choice of artificial intelligence. His opening like this is a bit of a dimensionality reduction blow to professional chess players who are not familiar with artificial intelligence. However, Ke Jie, the last world champion before the era of artificial intelligence, commented that this is very boring, and it is purely a memory game.

To put it simply, it is impossible to win the game of Go today in man-machine games, and even many artificial intelligences are specialized in guiding chess online. What’s more interesting is that human beings not only no longer pursue their own dignity, but instead study the methods and principles of artificial intelligence, and even sum up the classic methods of artificial intelligence, and start to learn from machines.

This reminds me of a similar but fundamentally different scene. In red and white machine game racing, there is a gameplay called TAS. When playing some ancient games, it will analyze frame by frame to perform perfect command operations. Many games have made a qualitative leap in real-life racing after the TAS gameplay has been made, and many of them are also particularly entertaining. However, the biggest limitation of TAS is that it can only find the best of known routines, and will not actively discover outrageous new routines. At present, artificial intelligence may discover new routines because of its self-play enhanced learning method. At this time, we can only learn from machines.

So, is the new routine of the machine the best? I don’t think so. The form of life has developed to carbon-based intelligence, but silicon-based intelligence is no longer something we can judge but can only learn. At most, we just unplug the power. Stronger Go players should be made by the intelligence behind the machine intelligence. What methods they use may no longer be comprehensible to us. Human beings can create intelligence that is smarter than human beings, and then this smarter intelligence can further create things that are smarter than it. As long as we accept the progressive education concept that the blue is better than the blue, then we may have to give up the premise that humans are the most intelligent species. After all, the education of carbon-based organisms is still too slow and too conservative.

Another revelation of Go lies in its evaluation of the current set. As mentioned earlier, in the previous education system, the winning moves may actually be losing. Even the so-called Go theory has been impacted by artificial intelligence. Ke Jie mentioned in a live broadcast that he was invited to teach chess to low-level professional players. It may be useless now, because he thought it was right or good at the time, but it was actually not good for artificial intelligence to evaluate. His own theory of Go or the so-called sense of chess may actually be wrong. That is to say, a good application scenario of artificial intelligence is to judge whether the experience summarized in those experience-based disciplines is reliable. It is very likely that the classic practice in some fields is limited to local solutions and does not see the optimal solution. In fact, artificial intelligence is gradually sinking into traditional industries such as manufacturing, and there should be a lot of room for improvement. Artificial intelligence after deep learning does not even require humans to provide sample features or statistics. The model can learn by itself to form features, and then discover more general patterns or formulas through enhanced learning. Of course, this may still not be the optimal solution, but this is no longer a problem that humans need to solve.

In addition, we can see that there are still people playing Go, and there are still competitions. In other words, although we can’t beat machines, we can still roll with people. However, hacking behaviors have already appeared in the current Go game. When artificial intelligence becomes popular, the first people to use it can also be regarded as hackers. Therefore, like the Go world, governments around the world are discussing how to control artificial intelligence. The core is not to touch the foundation of social fairness, otherwise the existing stable order of existence will be destroyed. From another perspective, in the post-artificial intelligence era, we don’t need to worry too much about how to get along with machines. After all, ants don’t need to worry about how to get along with humans. What humans need to worry about is still how to get along with their own kind. Higher intelligence has always been indifferent to lower intelligence, and what lower intelligence needs to care most has always been the competition of equal intelligence.

In the post-AI era, if life is a game of chess, then don’t regard machines as opponents. It is more likely that we will improve the rules of the world based on the suggestions of machines, so as to at least ensure that the game of life is relatively fair. In addition, stop discussing whether you will be replaced by machines. As the carrier of the next generation of intelligence, machines don’t even bother to replace a human job. A more likely scenario is that the machine as a babysitter develops a distribution system without an economic system suitable for the sustainable survival of human beings on the earth. Your food, clothing, housing and transportation are provided free of charge on demand, but the machine will not define the meaning of victory and defeat and survival for the giant baby group of human beings. At that time, human beings need to find a game scene for themselves to survive. What about artificial intelligence? They may be the exploration entities of the galaxy ocean and the practitioners of human curiosity.

Human beings are always the only enemies.

ps Don’t mistake me for knowing chess. When I was a teenager, my Chinese teacher said that his classmate had a cerebral hemorrhage because of playing chess. Therefore, I have a shadow and don’t want to learn chess. I even feel uncomfortable when I see this word. Now I want to open it, but my computing power can’t keep up. If I pass the 101 test, I will be unable to pass the 12th level. I’m afraid I’m not qualified for the enlightenment class.

This article is transferred from: https://yufree.cn/cn/2023/07/19/post-ai/
This site is only for collection, and the copyright belongs to the original author.