US races to develop AI-powered, GPS-free fighter jets, outpacing China

Following a recent dogfight between an artificial intelligence (AI) piloted F-16 and a regular piloted one, the United States Air Force (USAF) has reaffirmed its commitment to stay ahead of potential competitors, like China, in this field. The dogfight took place on May 4 and featured the USAF’s advanced X-62A VISTA AI-piloted F-16 with the Air Force’s highest-ranking civilian riding along in the front seat.

This was a significant milestone in the development of AI for the USAF, but, as the Associated Press (AP) reports, it is only a hint of its potential.

The United States is racing with China to advance the use of AI in weapon systems. However, this has raised concerns among the public that future wars could be fought by machines that select and target without human involvement.

While officials claim that such a scenario will never happen on the US side, there are worries about what a potential adversary might do. Therefore, the military feels it has no choice but to prioritize the rapid development and deployment of US capabilities.

The US is racing to master AI

“Whether you want to call it a race or not, it certainly is,” said Admiral Christopher Grady, vice chairman of the Joint Chiefs of Staff.

“Both of us have recognized that this will be a critical element of the future battlefield. China’s working on it as hard as we are,” he added.

The dogfighting AI is currently trained using a combination of machine learning and autonomy. Rules are set offline, uploaded to the aircraft, tested, and dumped at the mission’s end. The machine learning loop then repeats ad infimum to refine the AI pilot’s capabilities.

Apart from basic flight and combat, the military is also developing an AI-based alternative to satellite-dependent GPS navigation.

In a potential future conflict, high-value GPS satellites risk being hit or interfered with. If this happens, the loss of GPS could have severe consequences for communication, navigation, and banking systems in the United States.

Additionally, it could make it harder for the US military to coordinate a response, affecting the effectiveness of its fleet of aircraft and warships.

To this end, last year, the Air Force flew an AI program on a laptop strapped to the floor of a C-17 military cargo plane to work on an alternative solution using the Earth’s magnetic fields. The results were very interesting.

Throughout the flights and extensive data analysis, the AI learned to distinguish between signals to disregard and those to follow, and the results “were very, very impressive,” Floyd said. “We’re talking tactical airdrop quality,” he said.

“We think we may have added an arrow to the quiver in terms of what we can do should we end up operating in a GPS-denied environment, which we will,” Floyd added.

Safety must also come first

Safety is a critical concern when it comes to training AI systems, especially in the case of unmanned warplanes. Air Force officials have emphasized that the key to ensuring safety is to control the data fed into the simulator for the AI to learn from.

Therefore, ensuring that the data used is based on safe flying practices is essential. The Air Force envisions a future where a version of this AI could control a fleet of 1,000 unmanned warplanes currently under development by General Atomics and Anduril.

In the experiment, the service members assigned to MIT cleaned up the recordings of pilots’ communication to remove classified information and inappropriate language.

“Understanding how pilots communicate [indicates] their command and control abilities and thought process. If machines are to achieve their full potential, they need to understand this aspect as well,” said Grady.

“However, [machines don’t need to learn] how to use profanity,” he added.