“Strengthening human capabilities”: a professor of the Russian Academy of Sciences spoke about the further development of AI

— What is AI?

Concepts such as “neural network” and “artificial intelligence” are often mistakenly considered synonymous.

What are the main differences?

— Artificial intelligence is a combination of technologies and intelligent computer programs.

Their important distinguishing feature is that they are able to improve the quality of their work by learning from data.

If the zero version made a mistake somewhere, it can be corrected.

It can be compared to how we gain knowledge first at school, then at university, and the quality of our education increases.

Neural networks are one specific type of artificial intelligence model, but there's really nothing wrong with synonymizing them.

Because in 99.9% of cases now, when they talk about artificial intelligence, they mean models that use neural networks, so in a sense these have become synonymous.

— A prompt engineer is someone who develops and improves artificial intelligence models, giving step-by-step instructions or “tips.”

At the same time, the profession is on the list of those that will be replaced by AI.

Isn't it dangerous to create something that will replace the creator?

- Any technologies, when they are created, also affect new professions - their emergence or disappearance.

Now prompt engineers are in great demand.

And even when the same GigaChat recently passed a medical exam, the commission asked the machine a question, but the text itself was entered during testing by a prompt engineer.

This skill is now in high demand, and its value is only growing.

The ability to correctly write a query to a large model will eventually become a necessary requirement not only for researchers in the field, but for all people who use these models.

I don't see any risk of them being replaced, just that the types of activities will change a little.

— What is the difference between AI and human?

Can AI interact with humans?

In what ways is a human superior to AI?

— I like the definition that artificial intelligence is a technology that is created by people to solve their specific problems.

We do not understand how natural intelligence works, how the brain works, and we currently also do not know how artificial intelligence works.

That is, engineers have created something that scientists are not yet able to describe.

But if we talk about specific examples, energy efficiency and the number of tasks that a person’s natural intelligence can solve are much higher today.

However, it is clear that artificial intelligence already solves many problems much better than humans.

In fact, the most effective option is when they work together.

• Gettyimages.ru

• © marchmeena29

— What can’t AI do?

— Artificial intelligence can’t do a lot of things.

First of all, this is due to such complex tasks that require not routine work, but logical work.

There are exceptions - for example, he was recently taught to solve specific problems in geometry.

But in general, a person still copes better with setting and formulating goals and finding new methods for solving problems.

At the same time, someone who knows how to use those largest models goes through a large number of different options much faster - this creates a kind of multiplier of human abilities.

To summarize, artificial intelligence can do little without a person, but a person together with artificial intelligence can already do much more.

— How does AI help humans?

What tasks do you think AI will perform in the future?

How will AI impact our lives in the future, for example in the field of medicine and art?

— Artificial intelligence has already learned to see, that is, to work with pictures, read and understand texts, hear, work with sound and video.

Technology actively helps people.

For example, medicine in Moscow is actually a world leader in the use of artificial intelligence.

Solutions based on it are able to make a diagnosis based on a person’s entire clinical history, and in some cases even write a draft doctor’s report.

All this significantly reduces the time and burden on specialists, which leads to an increase in the speed of diagnosis and an increase in people’s quality of life.

In addition to medicine, artificial intelligence is also used in IT companies and in the financial sector.

For example, in the banking sector, technology makes it possible to significantly increase the number of analyzed factors when issuing loans - the entire client history is used.

This reduces risks and brings greater benefits to the business.

— Where in everyday life in 2024 do we encounter AI?

“When we interact with our digital environment using mobile phones and laptops, we already live in the world of artificial intelligence: the voice goes through various methods using technology to compress the data, transmit it, and so on.

We encounter it in Russian clinics, and also when we make purchases, walk around the city, or come to the bank.

At Gosuslugi it is planned to introduce a large language model for quick assistance to users by the end of the year.

— SpaceX and Tesla founder Elon Musk called artificial intelligence “the biggest risk humanity faces as a civilization.”

Why is AI dangerous?

— When new technologies appear, there are always people who actively promote and use them, and those who primarily talk about big risks.

Any innovation can be a risk, or it can become an advantage, because it depends on the person whether it will be used for good or harm.

Artificial intelligence is not being created so that robots can take over the world, but so that it can be used to solve specific problems for people.

At the moment, I don’t see any significant risks - rather, on the contrary, today we are taking an overly cautious approach towards artificial intelligence.

For example, the same driverless cars can drive perfectly on public roads for many years, but they are not allowed there because no one can give a 100% guarantee of safety, although the ride quality of driverless cars exceeds the ride quality of the average driver.

But we are still afraid.

In this sense, risks must be taken into account, but we must not be afraid, but consider the technology realistically.

— Is it possible to hack AI?

— Like any complex system, it cannot be made absolutely secure.

Therefore, all artificial intelligence developers assess the risks of hacking and work to counteract them from the very beginning.

It is important to know that this is an eternal race: attackers are busy creating new threats, but at the same time, information security specialists are developing systems to protect against such threats.

— What professions will be replaced by AI in the near future?

— Artificial intelligence technology is aimed at enhancing human capabilities; in the near future, it will only be able to replace professions with exclusively routine tasks.

One of them is a call center operator, whose tasks do not go beyond providing reference information.

But, as with any technological revolution, along with the abolition of professions, new ones appear - and much more.

Key condition: people must be able to work with new technologies.

That is, an artist or designer who does not know neural networks can be replaced by a designer who knows how to work with them.

—Who will AI definitely not replace?

— It seems to me that there will be no professions that do not use artificial intelligence.

Maybe only specialties related to physical labor.

This is no longer a question of artificial intelligence, but of robotics and the feasibility of its development.

I can cite a specific profession in which AI will definitely not replace humans - this is a surgeon.

When it comes to surgical operations, it is necessary to take into account complex dynamics and a high number of decisions made - today technology cannot replace a real expert.

— Is it realistic to replace judges or psychologists with AI?

— Artificial intelligence is not a replacement, but an assistant to humans.

If we are talking about trials, then not a single person is able to keep in mind all the laws that have been adopted.

It's the same in psychology.

No psychologist can keep all possible cases in his head.

And the enhancement occurs precisely due to the capabilities of artificial intelligence - the technology can significantly help both psychologists and judges.

But, of course, it is definitely impossible to delegate decision-making in the judicial sphere, but in some cases a trip to a psychologist can be completely replaced by communication with a fairly understandable chatbot.

There are already known cases of a person’s emotional attachment to a virtual assistant.

That is, some people are quite suitable for a virtual rather than a real psychologist, but for some not.

Therefore, a lot of interesting things await us here too.

— How far can AI go in self-awareness?

Is AI capable of realizing itself as a person?

— The question of such subjectivity of artificial intelligence still remains open.

It is discussed by both experts in the field of artificial intelligence and philosophers.

What is missing here is a clearer definition of self-awareness.

There are a number of psychological tests devoted to issues of subjectivity.

For example, the mirror test is used to understand whether animals are able to recognize themselves in the mirror - some identify themselves in the reflection, others cannot.

Theoretically, you could do the same test for large language models, and they would have the equivalent of conditional self-awareness.

It seems to me that at some point we will have to recognize the subjectivity of artificial intelligence.

Like a dog, which we believe is aware and understands, at some level even a language model may well do this.

Why not?

We're just not used to it yet.

— Will AI ever be able to solve global applied problems of humanity, for example, invent a cure for cancer or HIV?

— Yes, artificial intelligence is already actively used in the creation of drugs and reduces their development time significantly.

The main difficulty of this process is that you need to sort through a huge number of options, and although artificial intelligence is not entirely capable of replacing the entire development process, it can significantly speed it up.

In other words, technology will not be able to predict 100% whether a drug will work or whether it will be toxic.

But after analyzing large amounts of data, AI will select candidates much faster, who will then be tested using standard protocols.

— Is it possible that AI will reach such a level of development that a person will be able to transfer a “cast” of his consciousness into the network and “live forever”?

— The development of artificial intelligence is proceeding at a tremendous pace, but the issue of “casting” consciousness and its transfer to digital form is rather in the philosophical and theological plane.

We don’t really know what consciousness is, what the soul is.

In general, such a machine will be able to behave like a person, it will even be able to deceive someone.

But I think that this is neither a copy, nor a replacement, much less immortality.

For now, this is more science fiction than the real application of technology.