AI Pilot Takes Control of US Military Aircraft for the First Time

As the US military increasingly integrates drones into its aerial operations, human operators have long been essential for controlling these aircraft. However, this paradigm is shifting. Air Force Assistant Secretary Dr. Will Roper disclosed that for the first time, AI took on the role of a "co-pilot" in a military aircraft. During a December flight aboard a U-2 Dragon Lady spy plane in California, the AI systems were tested in real-time operations.

The AI, known as ARTUμ, is a variant of the μZero AI, which has previously been utilized in games like chess. In this instance, it exercised control over radar during a simulated missile strike on Beale Air Force Base. ARTUμ determined when to scan for incoming missiles and when to prioritize self-preservation, effectively acting as the mission leader by guiding the human pilot's flight path without any override capability.

The U-2 FedLab team dedicated a month to training μZero's algorithms for radar operations, enabling the AI to identify threats and manage interactions with the pilot. To mitigate risks, the finalized ARTUμ was isolated from other subsystems.

While the integration of AI into military operations is progressing, full autonomy in combat scenarios remains a distant goal. An AI-guided dogfighting drone is not expected to take flight until mid-2021, and there are concerns about the system’s vulnerability to human tactics. The Department of Defense is actively developing ethical guidelines to ensure that AI does not autonomously initiate attacks.

Despite the need for human oversight, AI's role is becoming increasingly valuable. Roper emphasizes that AI can assist pilots in navigating complex situations, allowing them to focus on critical mission elements. This collaboration could provide the military with a significant advantage in battle by streamlining decision-making processes.