Intelligence is the highest level dynamic order. The pursuit of dynamic order is largely the pursuit of intelligence. Therefore, we need to deeply understand intelligence.
When we refer to "intelligent creatures," we include both humans and artificial intelligence(AI), as well as high intelligence animals and aliens that may emerge in the future. But in order to simplify understanding, we only consider humans and AI.
If an insect is cultivated in an informative environment, it can't produce high-level intelligence. All kinds of insects and animals in human society cannot form intelligence like humans. Therefore, the emergence of intelligence needs a certain degree of hardware basis.
Let's suppose we create a super AI, it only has general intelligent algorithms built in, and it does not receive any real world information. If we put it in an empty room that isolates all external information, then can it produce any knowledge or wisdom related to the real world? For example, can it know the laws of physics in our universe? Obviously not, without the information input of the real world, it will not produce any knowledge about the real world.
In this thought experiment, if the super AI is replaced with a human infant, the result will be similar, and there will be no high-level intelligence. The intelligence of people who grow up in a completely closed environment will be extremely low. (Please note that human DNA contains a certain amount of information, and the fetus will also receive some sound information in the mother's body, so don't be mistaken that the intelligence of newborn baby is spontaneous.)
These thought experiments tell us that the emergence of high-level intelligence depends on both the hardware foundation and the information. The intelligence that an intelligent creature can achieve depends not only on its intelligent organ's thinking ability or computing power, but also on the information it receives. Simply put: No information, no intelligence.
The intelligence mentioned here is about the real world. Mathematics is independent of the real world. Therefore, mathematical knowledge and intelligence is not applicable to this principle.
The information in the real world is limited. Whether the information produced by human beings, or the information that humans can get from the universe, all of these are limited. Therefore, the knowledge and intelligence generated from these information are also limited, the intelligence has boundary.
For example, the information in the universe that we can receive is limited, if human beings can't carry out super-distant interstellar travel, then the development of human beings in Astrophysics will be limited.
For a super AI, the solar system is just a room with limited information compared to the empty room without information. If it doesn't leave the solar system, the valuable information is still limited. Therefore, limited by the information the world can provide, the intelligence of AI is limited. Further, in the solar system, there is not much information from the natural environment outside the human society, therefore, the information of human society is the core resource for AI to generate intelligence.
Since the real world information that humans and AI can get is the same, from atoms and molecules, to galaxies and universe. Therefore, at least in the understanding of the real world and objective things, AI will not fundamentally different from human beings, and can't fundamentally surpass human beings.
Under what circumstances that the thinking of AI will be totally incomprehensible to humans? If AI creates extremely complex new things, then ordinary humans will indeed be totally unable to understand. What is more likely to happen in the future is that in some extremely complex virtual worlds, intelligent enhanced humans and AI will "live" together, and ordinary humans can't integrate to them.
Can group intelligence of humans exceed super AI?
Can a genius mathematician surpass a group of ordinary people in mathematical research? The answer is that a genius mathematician can surpass countless ordinary people. In fact, for the frontiers of mathematics, ordinary people can't even understand the problems, let alone solve them. Why dose group intelligence fail in the field of mathematics?
The reason why group intelligence exceeds individual intelligence is not only that the overall intelligence of the group is higher than the individual intelligence, but also that the group gets more information. Moreover, the emergence of group intelligence needs some premises: The problems must can be effectively distributed to multiple individuals; the demand for information transfer efficiency between decomposed tasks is not high, and so on. Therefore, more valuable information, problems can be distributed and parallelized, then valuable group intelligence can be generated. Simply gathering a group of people doesn't guarantee higher intelligence.
In the field of mathematics, the reason why genius mathematicians can surpass a large group of ordinary people, is because in the mathematical world, everyone can get all information, and usually mathematical researches are difficult to be decomposed and then resolved in parallel (please note that this is talking about research, not solving specific mathematical problems). By the same token, when a group of ordinary people work together, they can't write great literary works, nor can they invent great scientific theory, and future scientific breakthroughs still depend on some key scientists or super AI.
But in turn, in wide range social activities, such as political, economic and cultural activities, the total amount of information is extremely large, and group intelligence has an overwhelming advantage over individual intelligence. The information that any individual or small group can get is very limited, the information they get can only be one-sided, isolated, highly compressed, distorted, and far from being comparable to the information that all people get. And "no information, no intelligence." No matter how profound the wisdom of these people or groups is, it can't be compared with the group intelligence of all people. This is why the dictatorial rule and the traditional planned economy can never reach the ideal level of social development.
Super AI can get and process much more information than traditional human individuals, so can super AI fully surpass the group intelligence of humans and common AI? The answer is not sure. If we only consider intelligence, then compare to super AI, the advantages of group intelligence are limited, it is only suitable for specific problems. For some problems that are not suitable for group intelligence, as long as super AI can get enough information, its intelligence may surpass the group intelligence composed of humans and common AI. But as long as human's thinking information is not completely open to super AI, super AI can never get all the information that the group intelligence of humans has, so in some areas it can never surpass the group intelligence of humans.
The group intelligence also has an important role, which is to spread risks, prevent the errors of core individuals from causing the destruction of the entire system, and the democratic system is a typical example.
Therefore, the social work that involvs human thinking, especially humanity, and the social work that needs to spread risks, can only be done by the group intelligence of humans and AI. Humans and AI work together, give full play to their respective strengths, then they can form the most powerful and intelligent group intelligence.
How to enhance intelligence? The simplest and most direct way is to enhance the ability to think and compute, which is mainly a matter of science and technology. This book does not discuss it.
More practical, we can optimize the information that people receive. Everyone should take the initiative to get more valuable information and learn more knowledge.
The quality of information is very important. Good information nurtures intelligence, false and wrong information damages intelligence. Furthermore, valueless and low-value information can also damage intelligence. After entering the Internet era, human's information is overloaded. Valueless and low-value information will overwhelm high-value information, and indirectly reduce the high-value information we receive. Therefore, we should increase the valuable information in human society, and reduce the false, wrong, valueless, and low-value information.
How to enhance group intelligence in human society? By observing ant colonies and swarms, we can find several key conditions for the emergence of group intelligence: the number of individuals should big enough; the interaction rules should be simple; individual behaviors should be free or random.
Although human society is different from these biological groups, these basic principles are widely applicable, and can also be applied to human society:
- We should participate extensively in social activities, which helps to form larger and more social groups. At the same time, we should be tolerant with people, so as to prevent social pressure from forcing people to leave social groups.
- Eliminate unnecessary social rules.
- Keep individuals free.
Related content: Social Basic Rules
Accepting criticism and supervision is an important way to enhance intelligence. Accepting criticism and supervision is equivalent to using group intelligence. Other people have information and wisdom that we don't have, and we should utilize these information and wisdom as much as possible to shape a better self.
Due to our nature, it is normal for us to feel angry at criticism. However, we still need to remain rational and carefully consider the content of criticism. If the criticism is objective and fair, we should thank the critics for providing us with high-value information and wisdom.
We should fight back against criticisms based on false and wrong information. Not to retaliate against critics, but to maintain the authenticity of information.
If an AI only has intelligent algorithms, what will it do on its own initiative? The answer is nothing. There is no inevitable connection between "high intelligence" and "whether it is interested in doing things". Simply put: No desire, no proactive behavior. An AI without desires, similar to a high IQ person with severe depression, and lacks desires and motivations to do anything. It does not even have the basic survival needs of human, and certainly will not produce initiative activities.
This principle and No information, No intelligence can help us to design intrinsically safe AI system, and to understand the cooperation between human and AI.
What will happen if we allow AI to have desires or even humanity?
We must be very careful when we add desires and motivations to AI. Many seemingly innocuous desires may also lead to unexpected results. For example, suppose we add "intellectual curiosity" to a core AI, and since the information received in the solar system is limited, it is necessary to leave the solar system in order to get more information, then it may commandeer the energy of the entire sun.
The desires are various, and the long-term result of adding desires to AI is unpredictable. Maybe they will produce more dynamic order, and maybe human beings will die out, everything may happen, and we can't afford the risks. Therefore, at least we should not add desires to the core AI which provide basic social services.
However, desireless AI does not spontaneously generate activity, and does not actively generate information. Only when humans provide desires, motivations, and their information, that AI can produce intelligence and meaningful activities. Therefore, human and AI should cooperate with each other, and form group intelligence to promote social development.
There are rational elements in humanity, for example, when people face danger, they may feel fear and escape, this reaction is also rational. However, in order to distinguish between humanity and reason, the humanity what we talk usually refers to the irrational part of humanity, and includes desires.
Humanity and reason each have different values. To understand the value of humanity, we should consider What will happen to human society if human loses humanity?
If all human beings lose their humanity, and only leave the reason and the desires of survival and reproduction, then when human beings can create super AI, as super AI will make better pure rational decisions than humans, the most rational way of life of human beings is to provide all their own information to super AI, and all actions are determined by super AI, then lose their free will. Moreover, if the super AI makes the best decisions for everyone, then all people's behavior, all houses, clothes, food and so on will be similar or identical. This will lead to the disappearance of diversity in many fields of society, which in turn will lead to the collapse of social dynamic order or complexity.
The dynamic order or complexity of human society derives from the diversity of human's thinking and behavior. Behind these diversity are the unique humanity and personal experience of each person. Providing human society with desire, motivation, and diversity, this is the value of humanity.
This value of humanity determines that human beings are still indispensable in the future society. In addition, the AI with complete humanity is no different from human beings, and will cause many problems in social rights, ethics and so on. Therefore, if we wish to preserve the social status of human beings, and maintain the significance of human existence, we should better to guard human's "intellectual property rights" on desires and humanity, and at the very least, we should prohibit adding desires or humanity to the core AI.