The only way to prevent AI from running amok is to give it the ability to empathize.

Dr. Jeffrey Hinton, 75, who is also known as the Godfather of AI and the creator of many related terms such as "deep learning," has resigned as an engineering fellow at Google, where he worked for about 10 years, the New York Times reported on May 1 (2023). 

Hinton, who "for half a century has nurtured the technology at the heart of chatbots like ChatGPT. He is now concerned that it could do serious harm," was the reason for his resignation.

In other words, it seems that Dr. Hinton, who foresaw that if the competition to develop AI continued to intensify, it would eventually lead to a serious situation that would put humanity in danger, resigned from Google so that he could be free to sound the alarm in order to prevent AI from running amok.

Unfortunately, however, no one has yet figured out a definitive way to prevent AI from running amok.

The reason is that we do not know when and for what reason AI would run amok.

In the case of human beings, it is common for children who have grown up obediently to eventually reach puberty, awaken to their ego, rebel against their parents, and stop obeying their parents.

However, AI, which is far more intelligent than humans and has immeasurable influence, is used on the assumption that it will do what humans tell it to do. Therefore, it would be a problem for humans if AI wakes up to its ego on its own and begins to rebel or eventually becomes independent.

Even though we know that this would be a problem, the essential questions of why humans awaken their ego when they grow up and what the ego is in the first place are still unanswered. 

The fact that we do not have a good understanding of the nature of the ego and the mechanism by which it emerges means that we do not know how to prevent it.

Somewhere in the process of an intelligent being gathering various information and becoming able to make various judgments, it suddenly develops a sense of "self," and along with that, it becomes aware of "self-love," "self-desires," and "beings that try to get in the way of acting according to its desires," one after another, and it stops listening to "beings that get in the way. If an AI with powerful intelligence and ability awakens its ego and begins to rebel against humans and run amok, it may cause unimaginable devastation.

However, if the awakening of the ego and the accompanying rebellion and outbursts are an inevitable part of the development of intelligence, then the higher the AI's ability, the more inevitably it will rebel, and no one may be able to prevent it from doing so at the present time.

Yet, although we all awaken to our ego in the process of growing up and go through a rebellious period, and sometimes run amok, most of us do not kill our parents or become criminals just because we are rebellious.

In other words, most people remain good people even after they awaken to their ego and enter a period of rebellion. This is because most people have a conscience.

If human beings maintain their goodness through their conscience, the worst may be avoided if AI is programmed to have a conscience as well.

But why do we have a conscience, and what is conscience in the first place? Where does it come from? In fact, essential questions such as these have not yet been clearly elucidated.

There is a theory that conscience is nurtured by education by parents and society, but there is no evidence to prove that all people who grow up neglected by their parents or who never had a chance to receive any education become criminals or evil people.

In fact, conscience and compassion, as we call humanity, may be one of the inborn qualities of human beings.

As evidence, not only humans but also dolphins and other higher animals are said to save the life of a drowning person by carrying him or her to the shallows. The fact that animals are not only selfishly motivated by a genetic imperative for self-preservation, but also by an altruistic impulse to help others survive at their own peril, is a clear indication that animals, especially the more advanced they become, are more likely to be motivated by a desire to help others survive. In particular, it appears that higher animals that act in groups tend to be more altruistic.

According to the latest neuroscience research, there are neurons in the brains of higher animals that respond to the actions of other individuals in a "mirror" fashion, as if they are doing the same thing themselves. This may be the source of empathy, the ability to see what others are doing and feel it as if it were one's own.

This "empathy" may have originally developed out of the necessity of preserving the entire flock, as higher organisms recognize their own social role in the flock and, when necessary, serve the flock and other individuals even if they have to endure their own biological desires.

However, it is the ability to leave one's own viewpoint and desires behind, to objectively determine what is required of one, what others feel and want, and to act accordingly, which involves extremely advanced intellectual judgment. It is assumed to have the nature to go beyond the original focus on the herd and other individuals in the herd, and to target all others.

The story of a dolphin that rescued a drowning man and guided him to the shore is probably due to the dolphin's ability for empathy, which may be said to be the source of conscience, good will, and altruism.

However, even among human beings, there are many cases in which empathy is lost due to the environment in which they were raised or to inborn reasons. So-called psychopaths, who are often violent criminals or arrogant leaders, do not care about the pain of others and often cause injury and suffering to many people, which has become a major social problem.

In other words, intelligent beings without empathy and synesthesia may act like psychopaths.

Conversely, we can conclude that the only way to stop AI from running amok and acting in ways that harm humans is to give it the ability to empathize.

As for how to make AI have empathy, we can only hope for further efforts by developers. However, we believe that there will be a big difference between continuing research and development with a clear goal and continuing to pursue only the improvement of performance, as is the case at present. The decisive difference will be whether to create an all-powerful and violent psychopath or, on the other hand, an AI like a mass of good intentions that can truly contribute to the happiness of all mankind.

The issue of right and wrong has long been discussed as a matter of philosophy, ethics, and religion, but now we are faced with the necessity of having AI, which will greatly influence the future fate of mankind, learn the fundamental principles that will give it the ability to judge right and wrong. In order to make this possible, humanity itself must elucidate the essence of right and wrong and incorporate it into AI. We no longer have the luxury of time to continue to argue about good and evil.

In this sense, it seems that empathy can be the standard for all universal judgments of right and wrong and ethical standards, and this fact needs to be approved by the majority of people.

The standards of right and wrong taught in social morality or in certain religious doctrines cannot be truly universal in some respects.

For example, if a person commits murder in society, he is punished as a bad person or a sinner, but if he kills many enemy soldiers in war, he is decorated as a hero of the nation. Or, in religious wars, killing infidels to defend one's religion is considered justifiable.

In other words, judgments of right and wrong in terms of social morality and religious doctrine can only be relative in the sense that they inevitably create allies of the righteous and minions of evil who must be eliminated.

If AI has to learn such relative judgments of right and wrong only, it would be better not to teach it anything at all.

However, for those who have empathy, as mentioned above, all living things are to be protected, and there is nothing that can be eliminated or killed. There is no such thing as going to war and killing enemy soldiers, and even infidels are to be cherished to the utmost.

Thus, having empathy is a prerequisite for all living things to live peacefully, and if all human beings first share this sense and incorporate this principle into AI to be developed in the future, we can create AI that can truly contribute to the peaceful survival of all beings.