‘It’s not science fiction’: US Air Force must build trust in AI, says expert

Defense News | Claire Barrett

The United States Air Force needs to quickly win the trust of operators regarding artificial intelligence if it is to effectively deal with threats posed by large numbers of unmanned aerial systems. At the Air and Space Cyber Conference, industry experts stressed that the rapid development of drone swarms reduces the time available for defensive decisions to just a few seconds, which means that automated systems are essential for battle captains.

The change from human-in-the-loop to human-on-the-loop systems is intended to address the problem of magazine depletion by using low-collateral directed energy weapons rather than the scarce physical interceptors. Yet trust still stands as the main barrier. The development of these autonomous capabilities is made more difficult by the absence of clear industry policy regulations. Nevertheless, the panel cautions that not advancing will leave the military at a disadvantage when faced with adversaries who deploy uncontrolled AI agents. In the end, it is considered essential to transform these technologies from the realm of science fiction into a reality in order to retain a competitive advantage in future high-speed conflicts.

Comment

The shift to using humans as an integral part of the targeting process in directed-energy systems such as the Epirus Leonidas greatly changes the speed at which tactical decisions are made when carrying out saturation raids. Since the Leonidas system automates the threat-prioritisation process, it avoids the cognitive bottleneck that would otherwise arise from having to carry out manual threat assessment in the case of drone swarms coming from multiple directions. Yet serious operational weaknesses are introduced by relying on algorithmic execution if the radar feeds linked to USAFE defences are disrupted by adversary electronic warfare.

As a result, placing these autonomous counter-UAS systems at forward bases such as Ramstein Air Base leads to a structural decentralisation of the authority to engage. In response, the command structures pre-assign the algorithmic firing solutions, which causes the battle captain's role to change from that of an active tactical controller to one of a post-engagement auditor. This change finally alters the tactics-related liability and risk management procedures within the USAFE integrated air and missile defence system.

Strategic Question for Discussion

How can the command structures of USAFE maintain effective oversight if the need to pre-delegate algorithmic firing solutions when deploying the Epirus Leonidas to forward bases such as Ramstein Air Base arises and adversary electronic warfare deteriorates the main communication links?

The pattern of operations indicates that USAFE will gradually depend more on edge-computed, decentralised decision-making units which will function autonomously when communication links are cut. I believe that this change will call for previously defined engagement areas in which the Epirus Leonidas can fire without the need for real-time human involvement, even though this involves a higher risk of friendly-fire incidents. In the end, this course of development shows that tactical control will move from active command to one based on strict airspace separation procedures.

Share your assessment in the comments below.

To read in details click here