The RAS discussed the development and criminalization of artificial intelligence

Scientists see pros and cons in AI

To what depths of our psyche can artificial intelligence (AI) be able to reach, and why we may need digital twins in the near future, academics discussed on Tuesday at a meeting at the Presidium of the RAS

Academician Dmitry Ushakov Photo: Still from video

Artificial intelligence can bring great help to a person, but it can also turn into a big problem, depending on whose hands it falls into.

Director of the Institute of Psychology of the Russian Academy of Sciences, academician Dmitry Ushakov told the audience what just recently seemed like science fiction : It turns out that now, with the help of artificial intelligence (AI), it is possible not only to evaluate a person as an individual, but also to predict his behavior in a given situation. How? Having recognized his train of thought…

Let's figure out how this is done, step by step. At the very first level, an assessment of a person’s properties is made from three sources: from a video image, from the acoustics of his speech, from the content of speech. Using a program that psychologists are developing together with the Institute of System Programming (ISP RAS), AI can make a complete “portrait”, determining whether you are an introvert or an extrovert, the presence or absence of hidden fears, and so on.

“Introverts, according to the conclusion made by the neural network, are more prone to using first-person pronouns in speech,” Ushakov gives an example. “People with a high level of conscientiousness use links… We can understand which people on social networks can refer to topics of drugs and weapons — these are more neurotic individuals.”

Then more interesting things begin to happen. The psychologist says that the next step in the implementation of AI in the field of population support will be digital doubles of people.

Knowing all the strengths and weaknesses of an individual, scientists can work with his digital copy, calculating, for example, which educational path will be more effective for him. “For example, one person needs to solve 20 problems in order to master a mathematical topic, another needs 100,” explains Ushakov. – With the help of a double, you can generate a task that will be the most developmental for a given student. But it is possible, by analogy with calculating chess moves, to predict the specific actions of an individual in a certain situation.”

According to him, it will be possible to analyze every next communicative action of a person. The Institute of Psychology has already developed the so-called intent analysis (from the English intent — “intention”), which allows you to understand what the purpose is behind what a person says. For example, he may strive to share information, show himself off, or maybe he seeks to make his interlocutor laugh, flatter, humiliate, ridicule, etc. “We can automatically study communication thanks to intention and teach a neural network to predict the course of each partner in a specific dialogue,” says the psychologist.

This, he said, makes it possible to analyze not only the course of communication, but also the state of people, to understand the complex vicissitudes of their feelings. And this, in turn, will allow us to provide them with very high-quality psychological assistance. For example, during a lesson, such a neural network that deciphers moods and intentions will tell the teacher which of the students is already tired and needs rest, and another, perhaps, on the contrary, can be given an additional task so that he does not get bored.

“The same smart speaker that now communicates with people on the same emotional level could in the future adapt to the owner,” says Dmitry Ushakov. “We could automatically analyze the sentiments of participants in business negotiations.”

In the future, he said, technology could become an indispensable assistant to a psychologist in resolving interpersonal conflicts.

“That is, we would get a digital “angel,” says the academician. – But we understand that where there is an angel, there is, as a rule, a fallen angel. The attackers who have taken possession of the technology can, thanks to the same algorithms, put pressure on people, manipulate, even waging psychological wars.”

He cited as an example the words of the American psychologist Francis Fukuyama, who said that the development of technology will lead to the appearance of two breeds of people — one with saddles on their backs, the other with spurs on their legs, with which they will spur the first…

Academician of the Russian Academy of Sciences Harutyun Avetisyan, who heads the ISP RAS, added fuel to the fire by citing real examples “from the life” of foreign AI:

“Recently, in Hong Kong, a banker was forced via video conferencing (which was arranged by scammers using artificial intelligence) to allegedly transfer two million dollars to his management in London. Another example: in California, an incident occurred — a person was hit by a car driven by an autopilot, and the autopilot, not realizing that it had hit a person, tried to drive off him (in addition, driving over the unfortunate person). The victim survived, but many questions arose about the advisability of using driverless cars in cities.”

The academician also spoke about the latest “achievements” of foreign computer game developers: “Our counterparts have gone so far as to create heroes in them who, through the game, find out more about you than you know about yourself… Then these “game images” can influence you through online channels and even offline: a person may come up to you, say a couple of phrases (triggers), and you will suddenly throw yourself under a car or, for example, rush somewhere instead of your business.”

Academician Mikhail Kirpichnikov asked what proposals AI developers have to reduce such risks.

“If it is impossible to completely abandon them (even if we do this, there is no guarantee that they will do the same on the other side of the world), we must “Now, without delay, resolve the issue of how to turn AI in the right direction,” answered Dmitry Ushakov.

According to Avetisyan, experts so far see only one way — the inclusion in the criminal code of new crimes related to the use of AI, in other words – the criminalization of artificial intelligence. Of course, this may require additional control systems. If the content is without watermarks (special certification marks), then it is illegal. According to the scientist, a depository of trusted solutions should be created at the international level.

President of the Russian Academy of Sciences Gennady Krasnikov added in this regard that the Presidium of the Russian Academy of Sciences has already created a Cybersecurity Council, which includes law enforcement agencies.