In the future, different AI systems could be delegated to carry out multiple different subroutines aboard the same aircraft.

Northrop Grumman has teamed with multiple AI firms, including Shield AI, Accelint, and Applied Intuition, to demonstrate that multiple AIs can control one aircraft mid-flight. The program, known as Talon IQ (and formerly known as Beacon), is designed to demonstrate that modular AI control can redefine how aircraft are flown and fought. 

Northrop Grumman Is “Hot Swapping” AI Systems

For the test platform, Talon IQ uses the Scaled Composites Model 437 Vanguard, an experimental testbed. For the tests, the aircraft was manned with a safety pilot onboard. But the pilot stayed mostly hands off, with the aircraft serving as a flying lab for autonomous flight. 

The AI being tested was arranged in a layered architecture. The base layer related to flight controls—keeping the aircraft stable, safe, and airworthy. The upper layer had the more sophisticated purpose of carrying out the mission. These layers functioned separately; flight was handled by the core system, while the mission was handled by AI. The architecture allows for a plug-and-play scenario, allowing for different AIs to be swapped in at the controls mid-flight.

The swaps were executed without disruption, and real-time control was maintained throughout. A rough analogy would be apps running on a phone’s operating system; the underlying system runs continuously, but only one app can run on the screen at a time. And through decoupling flight controls from mission controls, flexibility was enabled without sacrificing basic flight control.

Northrop’s “Prism” system acts as the baseline system, and would hand off control to Shield Hivemind or Applied Intuition or Accelint for task-specific controls. The tasks executed included Combat Air Patrol, simulated engagements, and similar missions. Notably, through AI specialization and swapping, a division of labor was created amongst different AI systems. 

Why Does AI Swapping Matter?

The tactical implications of modular AI are that different AI can be specialized and applied to specific functions mid-missions. One AI system can handle navigation, for instance, while another can handle engagement. Real-time adaptability allows for swapping mid-mission—reducing the pilot workload in instances where the aircraft is manned, and potentially pushing the human into more of an oversight role. While the human is relegated, the aircraft does gain multi-mission flexibility; a single aircraft can more effectively execute multiple roles thanks to AI swapping mid-flight. Drone teaming becomes easier, too, given that the future “loyal wingman” drones are expected to be governed largely by AI. In essence, AI swapping allows for the aircraft to become a platform for interchangeable tactics. 

The Talon IQ testing program is further proof of the trend towards software-driven warfare, where capability is defined by code. Software updates faster than hardware, meaning iteration cycles will accelerate. This allows for significant upgrades without new aircraft, which in turn allows for rapid adaptation. Modularity improves scalability, enabling large autonomous fleets that can overwhelm enemies with volume, potentially at low cost. Still, for now, a human pilot remains onboard the aircraft, meaning the autonomous systems are not fully trusted. The testing phase is still ongoing; the program has risks because the integration complexity is so high.

For the time being, the program is transitional. In the future, however, expect modular AI integration in the NGAD and the Collaborative Combat Aircraft. In the future, AI is almost certain to fully replace the pilot, making autonomous aircraft the standard. 

About the Author: Harrison Kass

Harrison Kass is a writer and attorney focused on national security, technology, and political culture. His work has appeared in City Journal, The Hill, Quillette, The Spectator, and The Cipher Brief. He holds a JD from the University of Oregon and a master’s in Global & Joint Program Studies from NYU. More at harrisonkass.com.

The post Can Multiple AIs Fly the Same Aircraft? Northrop Grumman Thinks So appeared first on The National Interest.