Russia has overtaken Ukraine in the large-scale use of AI on board drones, radio tech expert says – photo

Russia has overtaken Ukraine in the widespread use of neural networks directly onboard drones.

Source: Serhii "Flash" Beskrestnov, an adviser to the president of Ukraine and an expert in radio technology, spoke about this in an interview with DOU.

Quote: "We use neural networks to analyse images from reconnaissance UAVs to understand the whereabouts of the enemy's hidden assets .... In other words, neural networks are currently helping us to locate targets – and we started doing this a long time ago, even before the Russians. But the Russians have overtaken us in the technology of applying neural networks directly on board UAVs…"

Details: The expert explained that Russian neural networks have already been trained to recognise various types of targets, including tanks, cars, railway tracks and locomotives. Thanks to this, the drone can navigate visually without satellite navigation signals and autonomously search for a specified type of target.

Quote: "The Russians are already demonstrating this – they already have UAVs that fly autonomously. This is the V2U, this is the new Molniya."

More details: According to Flash, Ukraine started using neural networks even earlier than Russia, though mainly for analysing images from reconnaissance UAVs and detecting targets. He does not yet foresee the widespread use of neural networks directly on board Ukrainian strike UAVs.

Quote: "We have similar applications – we're not any less clever. But for the Russians, this is a standardised solution that has long been operating at the scale of the entire front line. Anyone who knows what Molniya is understands the danger: they have started applying artificial intelligence solutions to the cheapest Russian UAV, that is, on a mass scale."

More details: According to Flash, at this rate, Russia could scale up this solution to thousands or even tens of thousands of UAVs. Ukraine currently has prototypes of such systems, but lacks UAVs that could be produced on such a scale.

Flash reported that Russian drones use, amongst other things, NVIDIA Jetson Nano single-board computers and the YOLOv5 neural network for visual target recognition.

He also highlighted the increase in storage capacity in Russian V2U drones. While before, such drones could offer up to 128 GB of free storage space, now they have 2-4 TB. He believes this indicates an increase in the size of the neural network models and the volume of data used to train them.

More details: Flash said that the role of artificial intelligence in UAVs and drone control will continue to grow.

Quote: "I think it is only a matter of time before their Shaheds also start using artificial intelligence and neural networks for target recognition. This is our future."

According to Flash, Russia is already using AI-equipped Molniya systems to strike Zaporizhzhia.

Background: Ukrainian defence forces are already using more than 70 systems incorporating AI and computer vision for navigation, object recognition and target engagement.

Subscribe on Patreon to back our journalism long-term.