< br />
MOSCOW, May 14. Artificial intelligence has already learned how to perform basic tasks and can replace humans in recognizing fraudulent sites or links, but it is not capable of predicting the future, said a leading data scientist at the Laboratory Kaspersky» Vladislav Tushkanov.
“Artificial intelligence, in my opinion, is not applicable in what is called the prediction of human social behavior. There have been many such studies that are devoted to, for example, whether a person will study well at a university or what profession he should choose. All data on this is taken human, are transferred to artificial intelligence, but the results are very controversial. Predictions about human behavior is where artificial intelligence does not need to be applied, because the consequences will be unpredictable, «he shared his opinion.
Answering the question where artificial intelligence can and is being used now, Tushkanov said that neural networks can be entrusted with some basic tasks currently performed by junior specialists.
“There is one simple principle to understand where machine learning can be used. If a specialist can perform this task instantly, for example, to recognize a phishing link or a fraudulent site, then in such cases it is quite rational to use neural networks, because they are easily automated. In our In practice, there is also routine work: you cannot make out all phishing and malware with your hands — we detect 400,000 unique malicious objects per day,» he gave an example.
At the same time, the expert explained that the neural network can help only in the case of the simplest threats, while recognizing more complex threats, human participation is still necessary.
At the same time, Tushkanov warned against perceiving neural networks as a panacea. «There is a lot of hype around ChatGPT, and now a lot of people think that this is a silver bullet that will solve their problems quickly and easily. But this is a special area, it has been developing for a very long time, and specialists who know how to work with it should work with it contact who understand how to make these decisions reliable, how to make sure that they are not wrong or wrong in a controlled number of cases,» he explained.