The expert appreciated the possibility of transferring important decisions to AI

Experts believe that in the future, artificial intelligence systems will be able to participate in making collective decisions affecting the interests of millions of people. However, this will require retaining a key role for humans in determining the rules of operation of such systems and controlling their decisions, Elena Kantonistova, associate professor and academic director of the Artificial Intelligence Graduate program at the HSE Faculty of Computer Science, told Izvestia.

In short: chatbots and robots will be taught to work without the Internet

Solving a complex mathematical problem will help to "compress" neural networks without loss of quality.

According to the expert, AI is able to work effectively in areas where there are clear evaluation criteria and formalized goals. In such cases, algorithms can analyze large amounts of data and help choose optimal solutions.

At the same time, the researcher emphasizes that the most important condition remains the presence of predefined rules and transparent decision-making algorithms.

"Models should not come up with rules in the process of work. They must act within the framework of clearly defined principles, constraints, and priorities set by the individual. This is especially important in systems that affect the interests of a large number of people," Kantonistova said.

As the expert explained, it is not yet possible to fully formalize complex social, political or ethical issues. Therefore, human involvement in such areas should remain mandatory.

Simultaneously with the development of technology, it is necessary to create mechanisms for monitoring and regulating the use of AI. According to the expert, today one of the main issues is the distribution of responsibility for decisions made with the participation of algorithms.

"It is necessary to determine in advance who is responsible for the consequences of decisions made using artificial intelligence: the developer of the model, the company that implemented the system, or the operator who uses it. Without clearly defined legal responsibility, even the most advanced AI systems will remain a source of increased risks," Kantonistova emphasized.

According to the expert, in addition to legal regulation, technical control mechanisms are also important. These include the transparency of algorithms, the ability to explain the logic of decision-making, auditing systems, logging their actions, as well as strict restrictions on the powers of models.

For more information on how artificial intelligence can change the management system and what risks accompany its implementation, see the Izvestia article.