Ukraine Developing AI for Passive Reconnaissance of the Enemy

The Boryviter Center of Research is developing a neural network for the passive collection and analysis of publications in enemy media resources.

Pavlo Musienko, head of the analytical department, shared this with Militarnyi in an interview.

The ODCR Assistant neural network project aims to develop a powerful tool that will assist the Ukrainian military in automating the collection of information from open sources and enable the independent separation of potentially useful data for further analysis.

ODCR is an acronym for the key steps of experience capture: Observation, Discussion, Conclusion, and Recommendation.

“This module is currently being tested. We have connected it to some unclassified data streams and are passing all publications through it. Artificial intelligence already recognizes the main criteria of experience, which it uses to filter messages and generate observations ready for further analysis,” the head of the analytics department notes.

This includes social networks of the Russian military, as well as technical specialists involved in the development or use of military equipment on the front line.

The system monitors about a thousand messages a day from various sources, from highly specialized blogs to general channels.

Algorithms have already been trained to perform a two-stage sorting of information into those that may be valuable and those that are not, using a system of “red” and “green markers” – keywords that allow verifying the type of message. If there are “green” phrases and no “red” ones, the message is passed on.

After selecting useful publications, the ODCR Assistant analyzes the content to determine if the information is already known, and, if necessary, systematizes the data.

“It’s too early to talk about full machine analysis and automatic generation of ready-made data – this still requires human intervention. In addition, we work only with open sources, and this also imposes certain limitations on us. As soon as the functionality is finalized, we will use it to strengthen the capabilities of combat experience analysis specialists,” Pavlo Musienko says.

Boryviter is also working on another language model for the Ukrainian military. This project is also based on a neural network and has similar functionality to ChatGPT: it will communicate with experience carriers and simplify the search for answers to questions such as “what happened,” “what are the root causes of the situation,” and “what to change next time,” etc.

The developers said that the organization of the work was well established due to a well-selected team. The main difficulties are related to resources – you need a lot of hardware, because the more data you have, the more you need to process it. We already have large servers that constantly process the collected data.

Currently, the work is carried out exclusively by Boryviter. The organization is open to cooperation and support, but it is vital for it to focus on supporting the Ukrainian army and remain non-profit.

Read more on the tools and methods of studying and implementing military experience in the Ukrainian army in the article “Artificial Intelligence, Forums and Airborne Officers: How Does the Ukrainian Army Collect Combat Experience?”.