The Thinking Drone: Gyroscopes to GPUs
How battlefield AI is turning flying machines into autonomous weapons and reshaping the future of war in Ukraine and beyond.
July 2025 - Written by David Hambling and PDW CEO Ryan Gury
The evolution has been a century in the making. Now it’s reaching terminal velocity, playing out around the world. Drones are getting smarter – faster.
Elmer Sperry’s Automatic Airplane which first flew in 1917 featured the latest in autopilot technology: a gyroscopic stabilizer, a directive gyroscope, and a barometer to adjust altitude. After catapult launch, Sperry’s programmable attack drone would ascend to a given height, fly in a specified direction for a given time, then crash into a target – or at least into the general area of the target. Sperry admitted the Automatic Airplane was not accurate enough to hit a ship, and it never went into production.
Photo: Mike Mareen/AdobeStock
Subsequent drones relied heavily on remote human operators. For decades they were referred to as Remotely Piloted Vehicles or RPVs and had little or no ability to fly unaided. Even high-end USAF MQ-9 Reapers could not take off or land without a human pilot at the controls until a few years ago.
Everything changed with the consumer drone revolution of the 2010s, when makers like DJI harnessed technology developed for smartphones and other mobile electronics. This enabled drones to fly without human input, even tracking and automatically following the operator for ‘hands free flight’.
“We are seeing rapid adoption of modern commercial tech by militaries around the world,” said PDW CEO and cofounder Ryan Gury. “Our adversaries are already employing full AI on the battlefield – the decision is up to us if we want to lead or fall behind.”
Photo: PumpedVisuals/AdobeStock
In Ukraine, many FPV attack drones are now AI-enabled. These use commercial hardware, such as Raspberry Pi processors with customized machine vision software, to lock on to a target designated by the operator. Such drones can complete an attack even if the signal is lost due to jamming or terrain. According to Ukraine’s SBU, which carried out Operation Spiderweb, some of the attack drones “switched to performing the mission using artificial intelligence…After approaching and contacting a specifically designated target, the warhead was automatically triggered.”
Some drones in Ukraine now feature even higher levels of on-board intelligence and can recognize and classify objects on the ground. In the reconnaissance role they can scout areas where radio does not reach and bring back data. This also means human operators do not need to look though hundreds of hours of video but can concentrate on sections where the drone’s AI has flagged an item of interest.
Photo: PumpedVisuals/AdobeStock
At the highest level of automation, which we are already seeing on a small scale in Ukraine, drones like the Russian V2U (Using an NVIDIA Jetson processor) are searching, finding and classifying targets, then attacking them autonomously. The technology is currently limited and unreliable but is likely to improve fast. Both the AI software and affordable AI hardware are developing rapidly.
Fully autonomous weapons pose serious ethical questions as well as operational ones, but the capability now exists and is likely to be deployed increasingly in Ukraine.
“The future of warfare will undoubtedly be robotics on the front line,” continued Gury. “To maintain cross-domain battlefield superiority, America and its allies must invest in companies capable of rapid iteration and delivering reliable systems at scale.”