70 years ago, at the Massachusetts Institute of Technology in the United States, there was a famous conversation that revealed our current predicament. At that time, Marvin Minsk, a pioneer in artificial intelligence research, declared: "We want to give machines intelligence and make them self-aware." Douglas Engelbart, who invented word processing and the mouse, responded: "You want to give machines so many good things. ? Then what are you going to do for humanity?"
To this day, this seems to have become a "dark cloud" hanging over all human beings: from "Terminator" to "The Matrix" to "I, Robot" and "Westworld", countless science fiction movies have described a An extremely worrying future - machines that have gained self-awareness will in turn take over human beings or even eliminate them. In recent years, the rapid development of artificial intelligence (AI) has made many people increasingly worried about whether humans will be enslaved by artificial intelligence?
Not just “another tool”
Yuval Harari's new book "Beyond Homo Sapiens" is quite wary of this and firmly believes that this is very possible: humans have created forces that they do not know how to control. Once these new things escape human control, will in turn enslave or exterminate humanity. In his view, the non-human intelligence spawned by computer networks has brought about the first technology in history that can make decisions on its own and create new ideas, which may have a subversive impact on the order and system of current human society.
Disruptive impact? Of course there must be. Sunstein has long warned in his book "Network Republic" that the Internet is not just a technology. Its characteristics will cause "collaborative filtering" and "group polarization", allowing more and more people to hear only themselves. The response will over time put the operation of society into crisis. The information cocoons we are seeing today have confirmed this, we all know it’s bad, but the question is how bad is it?
In Harari's view, the challenges of artificial intelligence go far beyond that, because it is not just "another tool", but has the potential to master the information network for human communication. No matter how powerful new tools were in the past, the decision to use them rested with humans. However, artificial intelligence has changed this for the first time. This is precisely the fatal weakness of human civilization: "Although humans can establish large-scale cooperation networks to gain tremendous power, the way these networks are constructed means that humans are often unwise in their use of these powers." As a result, once Our failure to wisely control such networks risks our own destruction.
The emergence of human civilization relies on cooperation, and cooperation must rely on information exchange, otherwise no large society will be possible. From this perspective, all organizational forms in history mostly rely on a set of words that members collectively abide by. This is an "intersubjective reality": as long as members believe it to be "true," it is true for them, and cooperation and trust are all based on this tacit consensus.
Now here’s the problem: Artificial intelligence not only provides technical support for the centralized management of information, but it can also create new ideas on its own. Through independent learning, the algorithm can learn things that are not written into the program, and can also make decisions on its own. Something executives didn’t foresee. The more we talk to computers and the more information we reveal, eventually robot programs will be able to feed them opinions, sway our opinions, and manipulate human society. “Computers don’t need to send killer robots to shoot people, they just need to Just manipulate a human being to pull the trigger.”
To avoid such a terrible scene, the most critical thing is to have a strong self-correction mechanism to constantly identify and solve problems. However, as is common in computer programs, revising bugs often leads to new bugs: "Although the self-correction mechanism is conducive to the pursuit of truth, it will greatly increase the cost of maintaining order. The self-correction mechanism is too powerful. Doubts, disagreements, conflicts and rifts often arise. "Artificial intelligence may not change the pattern of social evolution, but it may make this evolution go straight in a dangerous direction.
At first glance, this is talking about artificial intelligence and information networks. In fact, the concern is whether and how the current social system can cope with such an impact. Harari highlighted the urgency of the crisis but may have exaggerated the danger. He is very worried that the "silicon curtain" will split human computer networks and create information cocoons that are difficult to communicate with each other. But on the other hand, in a pluralistic world, isn't it also naive to believe that "the world is flat" Faith? The information cocoon certainly brings closure, but it also provides a shelter for diverse cultures. Otherwise, we may witness the disappearance of a large number of niche cultures. Although that kind of concentration of information seems to be blessed by technology at first glance, over time, can this kind of self-isolation really have a stronger endogenous power and eventually withstand a vibrant open network?
Will artificial intelligence get out of control?
It is undeniable that his warning is indeed necessary, especially since technology often precedes social consciousness, and too many people are unable to realize what artificial intelligence actually means. Thirty years ago, U.S. Congressman Edward Markey said this: “The good news from Washington is that everyone in Congress supports the idea of an information superhighway. The bad news is that no one knows what it means. "This seems to be mocking the stupidity of politicians, but in fact, any Any new thing has this characteristic. When printing was first born, people never imagined that it would spawn countless books, archives, and invent banknotes. How many people could imagine that the chat tools of social software have evolved to this day and have so many features? Their multifunctionality makes our lives inseparable from them.
In other words, the impact of new things emerges gradually, and no one can guess from the beginning what complex and far-reaching social consequences it will eventually have. Harari is right about one thing: the key lies in the self-correction mechanism of society, rather than overreacting when there is a problem, and wishing to strangle this dangerous new invention in the cradle to prevent it from causing greater disasters; but in his opinion We are very worried that the self-evolution of artificial intelligence is out of human control. We originally wanted to get A, but ended up getting B - however, why can't "unintended consequences" be good?
Worrying about the eventual backlash of human creations has a long tradition in Western culture, stemming from the logic of the binary opposition between subject and object. The "Frankenstein Monster" depicted in the pioneering work of modern science fiction is such a story. This concern is certainly necessary, but does this kind of "controlled evolution" also mean that the upper limit of human intelligence will set a ceiling for technological progress?
Kevin Kelly also foresees in "Out of Control: The Final Fate and End of All Humanity" that humans may lose control of machines: "In releasing the power of life into the machines we create, we lose control of Their control. After they gain wildness and gain some surprises and surprises, there is a dilemma that all creators must face: they will no longer fully own their best. A proud creation.” However, he does not believe that such a future is worth fearing. Instead, he believes in the need to relinquish some control and let machines simulate “natural evolution” on their own: “Evolution can enable us to surpass our own planning ability; We can carve out things that we cannot; evolution can reach a more perfect state; evolution can take care of the world that we cannot take care of. But... the price of evolution is loss of control."
Learn to coexist with each other
Yes, artificial intelligence may have a profound impact on the current social system, but it needs to target society itself rather than technology, because the characteristics of computer networks often do not depend on technology, but on social networks. own characteristics. To put it bluntly, what kind of society will give rise to what kind of information networks and artificial intelligence applications. We might as well be frank and not blame technology.
This does not mean that we do not need to worry about artificial intelligence getting out of control, but it means that at least at this stage, its application and development still depend on society itself. This is like cell division in our bodies, allowing organisms to grow, adapt, recover and repair, allowing us to survive. But once this mechanism is hijacked by cancer cells and abnormal cell growth occurs, it will be fatal to the human body. We neither need to talk about cancer, nor can we sit back and relax. Instead, we need to find a way to get along with cells, eliminating mutant cells without damaging normal growth.
Since we cannot destroy artificial intelligence, nor do we want to be eliminated by it, then instead of seeing this as a struggle for control, we should learn more about the logic of biology and learn to coexist with each other.