Researchers at the Institute of Biology and Biomedicine of Lobachevsky State University of Nizhny Novgorod were the first in the world to create a neural network based on the principles of interaction of brain cells, which made it possible to increase its performance by 20%, according to the National Center for Physics and Mathematics (NCPM).
According to the author of the study, Susanna Gordleeva, digital analogues of auxiliary brain cells – astrocytes – were used in artificial intelligence for the first time. This is a new step in the development of spiking (impulse) neural networks, she explained.
It is they who process information in the way that the human brain does, solving dynamic problems, learning and memorizing data
Digital astrocytes control neurons, implementing the function of working memory. It is possible that this is the missing link in solving the problem of impulse neural networks, Gordleeva suggested.
Neural networks of this type can explain the mechanisms of information processing in the brain, the development of neurodegenerative diseases during human aging. On their basis, it is possible to create neuromorphic platforms for use in medicine, IT.
AI technologies have been actively developing in Russia since 2023. At "LETI", a neural network was trained to detect diabetes mellitus by human eyes. And "Comrade Major" finds administrators of anonymous Telegram channels. YandexGPT has implemented a video retelling function.