The world's first air battle between AI and human pilots

An F-16 fighter jet powered by artificial intelligence conducted the first ever air-to-air combat with a human-piloted fighter jet. This is the first footage from a September mock battle as part of DARPA's Air Combat Evolution (ACE) program.

Aircraft AI is nothing new, even for fighter jets. Indeed, military applications of AI are a major part of technological efforts aimed at improving conventional forces with robotic enhancements. For the United States and other air forces, this includes the creation of non-piloted loyal wingmen with the same performance capabilities as the pilot.

To accomplish this, DARPA launched the ACE program featuring the X-62A, also known as the Variable Flight Simulator Test Aircraft (VISTA). It is a Lockheed Martin F-16D Block 30-piece Marble II aircraft equipped with Block 40 avionics and other modifications that allow for the incorporation of AI. The ability to record flight time without human control has already been demonstrated, but it's still a long way from what the U.S. Air Force wants.

This is where fighting dogs come into play.

At first glance, DARPA appears to be working on an autonomous fighter capable of dramatic attacks. top gun-Type of engagement. But air combat is a bit outdated in the 21st century, and fifth-generation fighters are part of a global network of sensors, weapons, and command and control systems that the F-35 Lightning II and similar aircraft can detect and destroy. It is designed to be. You won't even see an adversary on the horizon.

The question is how to make AI more efficient.

Until recently, the main way to create AI for aerospace was to take an expert approach, where programmers relied on predefined rules written by experts for specific conditions. Simulations and flight tests were designed to confirm or correct these rules, with a primarily physics-based approach.

The ACE program aims to use machine learning as an alternative. In this regard, AI can adjust its behavior based on historical data and experience in nonlinear interactive systems. This can generate new insights and is especially useful in dynamic situations where the rules are unclear and outcomes unpredictable.

Although this is a useful approach, it also has its problems. When it comes to problem solving, it requires a much larger data workload and much more complexity. This puts more load on your computer. Additionally, there are challenges in understanding and validating what AI does and why. As far as the pilots operating the AI ​​are concerned, this can be summed up in one word: trust. Without it, pilots have nothing to do with machine learning AI.

That's why air combat was a big focus of recent tests at the Air Force Test Pilot School at Edwards Air Force Base, California. Dogfights may not be very relevant to modern combat, but such battles are complex and unpredictable, making them excellent for testing AI.

The AI ​​will also be required to follow the rules set for pilot training. Air combat training at altitudes up to 2,000 feet (600 meters) and speeds of 1,200 miles per hour (1,900 km/h) is expensive and tends to cause jets to collide with each other or hit bridges and buildings. Therefore, strict safety regulations are required.

The tests, which will continue throughout this year, aim to calibrate AI, establish ethical foundations for using such systems, and study how to measure and predict human trust in AI. Although the X-62A was operated by AI, there was always a safety pilot sitting in the cockpit.

So far, the program has made more than 100,000 lines of flight-critical software changes across 21 test flights.

If successful, ACE could lead to practical AI systems where fighter pilots are elevated to the role of mission commanders, overseeing more critical aspects of the mission, and where AI takes over flying and actual combat. Additionally, AI combat capabilities will also advance into over-the-horizon operations. This requires not only scaling up the technology, but also building the infrastructure to support it.

“The potential for autonomous air-to-air combat has been imagined for decades, but reality has remained a distant dream until now. In 2023, the X-62A will break through one of the most important barriers in combat aviation. ,” the chief said. Frank Kendall of the Air Force said he would personally fly in the X-62A. “This is a transformative moment, all made possible by the groundbreaking work of his X-62A ACE team.”

Source: DARPA