NCFU scientists create “smart” chips to work without the Internet and clouds

In the North Caucasus Information Center VGTRK in the framework of the joint project of GTRK "Stavropol" and the newspaper "MK-Caucasus" held a press conference with scientists of the North Caucasus Federal University (NCFU). The key topic of discussion was the opening of compact neural network architectures for edge computing at the University of Youth Laboratory.

Scientists have told how AI solutions help doctors make diagnoses and farmers save crops, and why the future of technology belongs not to giant servers, but to compact autonomous chips.

The laboratory was headed by candidate of physical and mathematical sciences Pavel Lyakhov. The main goal of the unit is to create methods of “compression” and hardware acceleration of neural networks. This will allow artificial intelligence to work directly on peripheral devices (phones, sensors, drones) without reference to cloud infrastructure.

The project has three phases. In the next two years, it includes the development of mathematical models of systems and neural networks and software implementation and experimental modeling. In 2028, we will enter hardware, i.e. the creation of specialized integrated circuits (ASIC) and programmable matrixes (FPGA).

Alexey Lyakhov stressed that the priority for Stavropol is autonomous AI for agriculture. Under the sanctions, Russian science has focused on compact solutions that do not depend on foreign graphics accelerators and work where there is no connection.

Being in a deep field where there is no Internet, the agrarian will be able to instantly get through a smartphone or drone an answer: are plants sick, do you need watering or fertilizers? Data processing on the end device is a huge economic effect for the economy of the region, the researcher said.

In addition to the agricultural sector, technologies are successfully implemented in medicine. So, in Stavropol, there is already an AI-based system that predicts the survival of dental implants, and development is underway to recognize arrhythmia and skin cancer.

Anzor Orazayev, Associate Professor of Mathematical Modeling at the Faculty of Mathematics and Computer Science named after Professor N.I. Chervyakov, Candidate of Technical Sciences, and Nikolai Nagornov, Senior Researcher, Associate Professor of Mathematical Modeling at the Faculty of Mathematics and Computer Science named after Professor N.I. Chervyakov, Candidate of Technical Sciences, focused on the problem of energy consumption.

“The further away the battery is from the outlet, the more important our technologies are,” Nikolai Nagornov explained. This is critical for sensors in the fields and portable medical equipment in ambulances, where the charge is limited and operating conditions are tougher than in the capital.

Anzor Orazayev added that their task is to “push” the most complex neural network into a tiny chip so that it does not overheat and works autonomously for as long as possible. The same applies to drones: AI can take on routine piloting tasks, but the responsibility always remains with the human operator.

The development of AI will inevitably lead to the intellectual stratification of society, experts say. Neural networks will become a powerful lever for those who understand the principles of their work, and only a crutch for those who put all the responsibility on them.

The changes will also affect the education system. Scientists have proposed a radical model: flip learning. The child receives factual answers at home with the help of AI, and comes to school for discussions, development of critical thinking and live communication with the teacher and peers.

Despite the fact that AI can replace up to 90% of employees in certain areas (in mines, logistics hubs, in primary diagnostics), Pavel Lyakhov urged not to give in to panic: The process of substitution is inevitable, but it opens the door to new professions: industrial engineers, data architects and AI systems. The main advice is not to be afraid, but to implement these tools as quickly as possible in your work and improve continuously.”.

Today, NCFU scientists actively cooperate with colleagues from China and leading research centers of St. Petersburg, confirming that the Stavropol School of Modular Computing named after Professor N.I. Chervyakov remains at the forefront of world technological progress.

Earlier it was reported that the NCFU opened the Youth Laboratory of Compact Neural Network Architectures.