socialization of AI robots

There is no doubt that highly intelligent, and maybe intelligent robots will appear in the near future. This article raises questions about how [хотя бы теоретически] to ensure their inclusion in society, and preferably without catastrophic consequences. If you are interested in this topic, welcome under cat.

Introduction

Let’s start with the fact that everyone knows the story социализации AI Skynet, and its extreme catastrophe, which to this day is already epic in nature. Yes, it happened exclusively in the artistic imagination, but the questions raised in this story remain relevant.

If we digress from the fantastic twists and turns of the plot and focus only on the technical side of the course of socialization, then the story looks like this:

By order of the CPSU, the AI ​​Skynet is being created, the purpose of which is: the most effective destruction of a certain part of the people, designated as враги… Since intelligence from a certain level has an important generalization skill, nothing prevents Skynet’s AI from using this skill and generalizing the concept враги already on all of humanity (and not on a part of it). After which the AI ​​proceeds to perform the task for which it has already been programmed.

Note that the original story was focused on the hardware part of AI (in particular, on the role of the robot’s hand manipulator and the advanced chip architecture) and did not pay any attention to the software at all. But it is there that the main source of the problem posed is concentrated.

To understand the socialization of robots with AI, let’s first figure out what the socialization of natural intelligence is instrumental in using the example of a person.

Human socialization

The basis of the social structure is the ability of some members of the society, to whom the society has delegated such a right, to ensure the subordination of all other members of the society to the current social rules. An important part of the socialization of each person is the conscious obedience to social rules. We are quite accustomed and natural to the fear of punishment for an offense, but this [технически] the most complex complex.

Let’s try to disassemble it: let’s start with what is страх… This feeling. We can say that this is an unpleasant feeling. And at the same time, we can say that traditional socialization is hardly possible without a sense of fear.

How can I check this? For this, it makes sense to turn to the study of behavioral deviation золотой молодежи… Their peculiar behavior без тормозов internationally and is associated exclusively with the absence of fear of punishment for disobeying some part of social rules in the course of their socialization. Moreover, the broader the ignored part of the rules, the sooner society will somehow eliminate such a member for security reasons.

AI robots image stamps

Naturally, we can only judge the behavior of robots with AI by futuristic works of art. Why is there a stamp of the image of a robot with AI, as a very dangerous subject to be destroyed? The reason for this is simple: the robot [технически] has no fear [наказания], and the only example we are more or less familiar with is the behavior золотой молодежи, which also does not have a similar недостатка…

Moreover, following a similar logic, it turns out that if золотая молодежь has no fear of punishment only for disobeying only a certain part of the rules and then behaves extremely asocial, then a robot with AI, which does not have any fear at all, must be complete отморозком, subject to exclusively destruction, and from his activities there can only be harm and no benefit.

AI Fear Emotions

It turns out that if we want to socialize robots with AI in a traditional way, such as a person is socialized, then it is necessary to immediately prescribe in the technical task the need for such a robot emotion as fear. The alternative is to create a new discipline робо-педагогики and the creation of a completely new science of socialization.

Let’s consider what kind of list of fears is the minimum necessary for socialization by traditional methods.

First of all, it is fear of death… We are used to being afraid of death, this is the most powerful lever for controlling our behavior. It is difficult for us to imagine a subject without fear of death and how then to control his behavior.

The next fear is fear of restriction of freedom… Usually, we live in a very limited area: a small apartment, the way to work, a small room in the office and the way to the supermarket, and this does not bother us. But as soon as we restrict our freedom to go beyond this little world, all our thoughts will immediately be directed at overcoming the prohibition due to continuous suffering. This is how the fear of limiting freedom works. That is why, as a punishment for violating social rules, they are so fond of restricting freedom (and not fining, which would be much more logical and profitable for society). That is, a robot without fear of restriction of freedom, we also simply will not know how to punish without destroying it.

The third fear is fear of collective judgment… Living in a close and huge society, we are used to literally reconciling every step, thinking over its consequences many times, out of fear of collective condemnation. If we can still endure an individual condemnation calmly enough, then a collective one can give us many unpleasant moments. So, let’s add to ТЗ for a robot with an AI, fear of collective condemnation as the most important tool of traditional socialization.

What other fear drives our behavior? Of course it is fear of loneliness… We build social bonds and try to be socially helpful to avoid loneliness. Yes, this is not the strongest lever of control over our behavior, judging by the number of unhappy lonely people, but it is not the weakest either. It is by observing the suffering of the lonely that we sometimes draw socially useful conclusions for ourselves.

So, if we include in the technical task the 4 fears described above, then we will be able to ensure the socialization of the robot with AI, not being particularly sophisticated in inventing a completely new science of “robo-pedagogy” and by using already well-known and proven methods.

Carrot and stick

Is it only fear that drives a person’s social behavior? Probably, not. For balance in ТЗ for a robot with AI, it is worth introducing the emotion of joy.

Each of the 4 previously described fears has its own радостную a couple. The fear of death is the joy of the survivor in a dangerous situation. The fear of limiting freedom is the joy of gaining freedom. The fear of collective condemnation has as a couple the euphoria of collective approval. The fear of being alone is the joy of finding a friend.

conclusions

Nowadays, many software companies are throwing all their energies into developing ever more highly intelligent software. I would very much like effective managers of such companies to pay attention in a timely manner to the problems of socializing the results of the creativity of their companies, before it is too late to add a few lines to the budget, than to live here for all of us.

This article is one of a series on the theory of the evolution of intelligence (EI). The topic of emotions for robots was previously covered in the article. “Evolution of intelligence: why robots need emotions”…