Ukraine Collects 2 Million Hours of Battlefield Footage To Train AI For Military Purposes

Ukraine gathered a large database of around 2 million hours of battlefield footage from over 15,000 drone crews since Russia’s invasion in 2022.

According to Reuters, the footage gathered—equivalent to 228 years of video—includes content from Russia and Ukraine’s frontiers and can be used to train AI models for military purposes.

“This is food for the AI: If you want to teach an AI, you give it 2 million hours (of video), it will become something supernatural,” said Oleksandr Dmitriev, founder of the non-profit digital system OCHI, to Reuters.

Dmitriev explained that the footage could be used to improve combat tactics and optimize weapons’s performance with AI-powered systems.

The use of artificial intelligence for military tactics is not new. Both Russia and Ukraine have used this technology during their conflict to scan images and identify targets more effectively than a human could, or to fly to certain regions without human piloting.

After realizing that the drones deployed in 2022 could provide valuable information for AI training, Dmitriev’s team has been gathering around 5 terabytes of new data daily.

The Ukrainian government also has another system called Avengers which collects and processes its own data. The Ministry of Defense revealed that with this AI-powered tool, they have been able to identify 12,000 Russian targets.

Experts agree that the large battlefield footage could be very powerful depending on the quality of the images, but Dmitriev didn’t disclose more details.

The use of AI for military purposes has been increasing in multiple regions of the world. A few weeks ago it was revealed that Chinese researchers have been using Meta’s AI model Llama to develop military systems, and Google Deepmind workers have expressed concerns over Google’s contracts with military organizations.