The creation of neuromorphic artificial intelligence systems will allow mankind to enter a new stage in the development of computing technology, even closer to the capabilities of the biological brain.
As Denis Larionov, head of the artificial intelligence department at Tsifrum (Rosatom), noted during a report at the OpenTalks.AI conference on artificial intelligence, this will help to significantly increase the energy efficiency of computing.
Last week, the forum brought together in Yerevan (Republic of Armenia) iconic speakers of the artificial intelligence industry from Russia, the CIS countries and far abroad from among the leading universities and companies involved in the development of AI technologies.
“Neuromorphic systems borrow some of the principles of the human brain, such as computing in memory, the impulse nature of information transfer, analog computing, and others. Today, these systems are considered as the technological basis of a new direction and a new stage in the development of computer technology. Such systems have an unprecedentedly high level of energy efficiency – from 100 to 1000 times higher than classical solutions,” Denis Larionov said in his report during the forum.
Neuromorphic processors will become universal computing systems for solving a wide class of AI tasks; in addition to energy efficiency, they will be distinguished by high speed, almost unlimited scalability, and opportunities associated with continuous learning. However, at the same time, today they are inferior to classical processors in terms of the quality of problem solving and the development of the software and hardware ecosystem.
Today, dozens of foreign companies, as well as Russian researchers, are engaged in the applied application of neuromorphic computing. A promising application of neufromorphic AI systems is predicted in medicine (since the event computing model used in neuromorphic processors is fundamentally compatible with the human nervous system), in various industries, for example, the aerospace industry, in science and technology and, of course, in the nuclear industry.
“In production, using neuromorphic AI systems, predictive analytics and the direction of computing on end devices (edge AI) will develop. Neuromorphic processors will be widely used in solving intellectual problems in extreme conditions (space, ocean, radiation). This is primarily due to the high resistance of the impulse communication protocol to noise,” said Denis Larionov.
Research in the field of neuromorphic systems in the world is at its initial stage, however, the pace of development with the participation of global companies is gaining momentum. The tasks of ensuring the technological and industrial sovereignty of the country require the well-coordinated work of Russian scientists and developers in this direction.
The message Neuromorphic artificial intelligence systems can be a thousand times more energy efficient than classical computing systems appeared first on TEKNOBLOG.
