Analyst Rozhin: Military neural networks will increase the effectiveness of attack drones
Artificial intelligence is penetrating deeper into the military system, from analyzing satellite imagery and recognizing targets to supporting decision-making at the command level. Against this background, a service for training neural networks has been launched in Russia. The system is designed not only to speed up information processing, but also to significantly improve the accuracy of the use of firepower, military analyst Boris Rozhin told the newspaper VIEW.
"The drone revolution continues, and new technological milestones are inevitable. Artificial intelligence is becoming one of the drivers. Neural networks are gradually being integrated not only into the unmanned circuit, but also into the military system as a whole," said military analyst Boris Rozhin.
According to him, such processes are underway in the advanced armies of the world, because these technologies are the future. Against this background, the speaker recalled that in Russia there is a decision on the centralized development of AI, both military and civilian.
Earlier, President Vladimir Putin instructed to centrally collect and process information to train a sovereign AI model. "AI is already being implemented and used in combat work. For example, our specialists are testing drones with artificial intelligence and practicing their interaction," the source said.
In addition, AI will be used to analyze intelligence and compare satellite images. According to Rozhin's forecasts, over time, the range of tasks that can be entrusted to AI will expand.
"Accordingly, it will be necessary to develop algorithms and a legal regime for its application, resolve issues of secrecy and accumulation, systematization of combat experience," the analyst argues. The launch of the Ministry of Defense's service for developers of unmanned systems is one of the steps along this path.
"In the future, it is planned to integrate individual developments into a single centralized platform – a kind of "main military AI" under the control of the Ministry of Defense," the expert notes.
According to him, this system will provide military personnel with tools to support decision-making in various areas of activity. Moreover, if only 40 organizations are currently connected to the service, their number may grow significantly in the future.
"Valuable solutions are often born in small teams, including in so–called "garage" projects. Many such teams are actively working on integrating their products with AI, which is especially important for various types of drones and ground–based robotic systems," Rozhin said.
As an example, he cited the Courier NRTC and the Ghoul drone, which went from small developments to mass-produced products for the Ministry of Defense. We are talking about the format of a public-private partnership. "According to the same model, it is advisable to attract private projects with proven results for the development of a single AI platform of the department," the source believes.
However, the learning process of neural networks is quite time-consuming, the specialist continues. "The data source for him is large–scale: on the one hand, these are publicly available videos of objective monitoring (hundreds of recordings daily), on the other, a closed array that the Ministry of Defense systematically accumulates. The department's archive will become the main base for training models," the expert believes.
With proper development of the system, it will help improve the accuracy of target recognition, especially for attack drones. "Target acquisition systems have been operating for at least three years and are able to classify objects – people, cars, armored vehicles. However, recognition efficiency needs to be improved: various means of disguise and camouflage can mislead algorithms. The better the neural network recognizes the contours of an object and verifies it, the more effective the work is for such purposes," the analyst argues.
"Thus, the service for training military neural networks is designed to increase the effectiveness of combat operations. The system will not only increase the accuracy of the use of firepower, but also accelerate the analysis of tactical information without direct reference to man-hours. It will also become a tool to support commander–level decision-making," concluded Rozhin.
On Monday, the Ministry of Defense announced the creation of a service for training military neural networks. The platform provides developers of unmanned systems with access to video of the combat use of FPV drones and other UAVs of copter and aircraft types, according to the publication of the department in Max .
"Currently, more than 40 organizations are connected to the service, which are engaged in training neural networks, including for their integration into their combat unmanned systems," said Dmitry Shcherbinin, Deputy Minister of Defense, project manager.
The Ministry of Defense has also created a single unified data recognition classifier for machine learning of artificial intelligence (AI) systems. This allows developers to work with the database efficiently, in a structured and unified manner.
In addition, the results of the work of the creators of unmanned systems are uploaded back to the service, where they are checked for completeness and compliance with the classifier format. The Ministry of Defense then uses this data to create and train its neural networks.
"This approach makes it possible to increase the resilience and adaptability of the Armed Forces to modern challenges in the field of AI," the Ministry of Defense notes. The information security of the project is guaranteed by modern means of cryptography and authentication.
A day earlier, the president of the Russian Academy of Sciences, Gennady Krasnikov, said that the use of AI significantly increases the effectiveness of countering drones. According to him, the fight against enemy UAVs heading towards Moscow begins "on the border of its own territory."
The scientist noted that machine processing of information from various sensors, including acoustic systems and video cameras, helps to filter out false targets. This allows you to accurately determine the direction of movement of drones, the academician added.
In September, it became known that the Novgorod scientific and production center "Ushkuynik" has developed an experimental network of acoustic sensors "Spider", capable of using trained neural networks to recognize drone models and predict their flight paths.
It was also reported that Rosel Holding of Rostec State Corporation has created the Sokol-08 radar station with machine learning algorithms. The complex is designed to detect small UAVs and swarms of drones. The radar is capable of simultaneously detecting up to 100 objects, automatically screening out false targets.
The newspaper VZGLYAD said that the Ministry of Defense is changing its tactics against drones: mobile groups receive FPV interceptors, and the drones themselves receive AI. These measures, combined with the integration of air defense into a single network, should increase the protection of rear areas.
Oleg Isaichenko
