In recent years, as the field of artificial intelligence (AI) has exploded, many scientists, including Stephen Hawking, have repeatedly expressed concern about the potential risks associated with AI. In this regard, endowing AI with a rich palette of emotions seems to be a rather dangerous undertaking. However, according to the professor of the Department of Cybernetics, NRNU MEPhI, all these risks are completely controllable and in the future, "emotional" work will become reliable helpers of a person.
A universal robot needs human emotions
Samsonovich says that a man long ago invented all kinds of devices: a camera, a camcorder, a calculator, a telephone, today all these functions are integrated into one small smartphone. Currently, a new era has come in the development of computer programs, a complex program is needed that can independently perform almost all the work, as well as communicate effectively with a person and accurately understand our requirements. To achieve this goal, you need to create an "emotional" AI.
According to the scientist, not sharing AI with emotions is a complex and ambitious project. First, it is required to establish constant contact between a human and a robot, to make sure that the AI can accurately understand anthroposemantics, and also use a similar language system for feedback. Secondly, to create such a virtual assistant who can independently study in the real world, understand human emotions and feelings, and enjoy the trust of a person. In the end, it should be an assistant robot that carries out human commands, moreover, it is an extension of the human body, when with the help of thought it is possible to exercise effective control over the robot.
For artificial intelligence, as well as for humans, "love is stronger than money"
In 1942, Isaac Asimov, in his story "Round Dance", put forward the "Three Laws of Robotics". Can these three laws actually be implemented to regulate AI behavior?
Promotional video:
Alexey Samsonovich thinks that the answer is obviously negative. In the future, AI will have to independently face complex and unpredictable external circumstances, using data analysis to independently make the best choice. In addition to being full of vulnerabilities and vague in and of itself, The Three Laws of Robotics can make it impossible for AI to take a step in the real world. However, according to the scientist, people can use other methods to train AI so that its behavior is favorable to humans.
Currently, Samsonovich's team is conducting an extremely interesting experiment: they have developed one simple computer game in which three virtual AI characters are isolated on one planet, it is necessary to carry out a rescue operation using two teleports. These two teleporters can only be activated together, in the end, only two characters will be saved, and the rest on the planet will have to wait for one of his comrades to return for him. Thus, a test is carried out whether the "emotional" AI will be able to adapt to relationships in human society, establish trust with partners, show sympathy for a friend in a difficult situation and consciously help him. The research team will use this experiment to collect data and optimize the AI emotional model.
The threats posed by artificial intelligence originate in humanity
After the computer program AlphaGo defeated Li Sedol in the battle between man and machine in the game of go, people began to worry that in the future, AI could gain intellectual superiority over man. In this regard, Samsonovich says that AlphaGo's victory means the rapid advancement of AI in the field of deep learning, but go is an area in which it is relatively easy to produce mathematical abstractions. Today, the superiority of AI over humans is still based on the computational ability of the former, in essence, this kind of advantage is still determined by humans. The existence and development of AI in the future will be fully limited by humans, the growth of its capabilities will depend on the continuous optimization of algorithms and models by humans.
According to the scientist, if it happens that the development of AI will pose a threat to the existence of humanity, then the reason will still need to be sought in humanity itself. He said: “I have no doubt that there will be people with malicious intent who will try to use AI technology to create dangerous robots and even robotic weapons. In fact, a new type of offense is constantly and relentlessly following the development of electronic computing technology. For example, viruses and malware are human-dependent.
People can, of their own choice, endow AI with positive emotions so that it becomes an effective assistant for our work and life, and not at all a disaster that will wipe us off the face of the earth."