The Russian Academy of Sciences explained what artificial intelligence really is

To begin with, we must objectively note that the authority of the Russian Academy of Sciences is growing. This was noted by all its members with whom I had the opportunity to talk on this topic, and this was also proven by the large list of Russian ministers and political figures who graced the General Meeting on Tuesday with their presence.

The first scientific report on microelectronics was, naturally, read by the President of the Academy himself. First he told me “how it is with them”:

The roadmap is accelerating and moving forward. Moreover, it is ahead of schedule: they planned (at Intel) to make 2-nanometer technology by 2025, but are ready to present it already in 2024…

Then Gennady Yakovlevich moved to our microelectronics production. Let’s omit its complex details, noting only Krasnikov’s positive expectations regarding the formation of the necessary technological chains in our country: for photoresistors, technological equipment, electronic engineering…

“We need to restore this situation, as it was in the Soviet Union,” he said. – I have proposals for all departments. <...> — I believe that the work will develop, and in the next 4-5 years we will see qualitative changes.

One of the following reports is from the director of the Institute of System Programming of the Russian Academy of Sciences. V.P. Ivannikov Harutyun Avetisyan was dedicated to the creation of trusted artificial intelligence.

The term “trusted” is used in a number of international and Russian documents and means transparency and security for the user. In general, this is an ethical topic that reliable technologies should stand behind as guardians.

Mathematician Avetisyan reassured us a little, saying that artificial intelligence, which many people use to scare humanity today, is essentially not scary, and Russia has rich potential in the development of technologies capable of taming reckless AI.

“Now we live in a world of weak artificial intelligence, and there is no question about its subjectivity,” the academician told those present and promised to explain this thesis during the break. The MK correspondent took advantage of this.

If you look at what is happening in the world (by the way, several articles by guys from Google have recently been published about this), then all existing generative models still work only in the segment and on the data on which they were trained, explains Harutyun Ishkhanovich. – Behind this there is a simple thing: in fact, everything that we now call AI is machine learning. It may be very cunning, but it’s just machine learning.

– Well, is a “miracle” in the form of AI possible in the future?

– There is hope, but this will not happen earlier than in 10 years. This cannot be achieved using supercomputer resources – you need more sophisticated mathematics inside… However, someone says that there are 2 years left before the advent of artificial intelligence…

– Where do you see the imperfections of AI?

– It is, in part, clearly visible when the degradation of the “miracle” occurs, when a car suddenly runs into a pedestrian and for some reason, instead of stopping, begins to behave inappropriately: it continues to move, but in reverse. It’s impossible to prepare AI for every eventuality…

– What other dangers do you see in generative systems?

– Unlike conventional information systems, which are checked, in AI it is often not clear what needs to be checked. At the same time, the benefits from its use are high – labor productivity increases everywhere. Everything is being implemented, but without taking into account the component that the system is not fully trusted. Moreover, it’s not just about bookmarks…

– What is this?

– In other words – poisoned data that can be slipped in by scammers. Your machine will learn from them, and after that your system can be hacked. For example: a person creates a computer vision structure that should control the perimeter around a house so that intruders do not enter the territory. But he was given such data that a person puts on a certain (specified in the bookmark) badge or cap, and the system lets him through.

– What do we have to protect ourselves from this?

“We can be proud because we were the first to sound the alarm; five years ago we created a new specialty in universities: “Cybersecurity.” We are developing at a normal level.

…The next speakers were experts on the most advanced laser technologies in the country, climate change issues, and transformation of the world order.

Closing the first part of the meeting, the scientists seemed to be pleased, because even at the beginning of the speech, Gennady Krasnikov announced the beginning of support from the Ministry of Finance for the sixth, military subprogram of the FNTP, noted that from 2024 the categorization of RAS institutes, which prevented the development of institutes 2 and 3 categories, and also pleased the members of the Academy of the Russian Academy of Sciences, whose scholarships will be increased from January 2024.