WASHINGTON (AP) – Two Air Force fighter jets recently clashed in a dogfight in California. One of them was flown by a pilot. The other one wasn’t.
The second jet was controlled by artificial intelligence and the Air Force’s highest-ranking civilian was sitting in the front seat. It was the ultimate demonstration of how far the Air Force had come in developing a technology that dates back to the 1950s. But this is just a hint of the technology yet to come.
The United States is racing to stay ahead of China in its use of artificial intelligence and weapons systems. The focus on artificial intelligence has raised public concern that future wars will be fought by machines that select and shoot targets without direct human intervention. Officials say that will never happen, at least on the US side. But there are questions about what a potential enemy would allow, and the military sees no alternative but to quickly deploy U.S. capabilities.
“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. “We both recognize that this will be a very critical element of the future battlefield. China is working on this as hard as we are.”
A look at the history of military development of artificial intelligence, what technologies are on the horizon and how to keep them under control:
From MACHINE LEARNING to Autonomy
The roots of AI in the military are actually a hybrid of machine learning and autonomy. Machine learning occurs when a computer analyzes data and sets of rules to arrive at conclusions. Autonomy occurs when these consequences are applied to action without further human input.
This took initial form with the development of the Navy’s Aegis missile defense system in the 1960s and 1970s. Aegis was trained through a set of human-programmed if/then rulesets to be able to detect and intercept incoming missiles autonomously and faster than a human. However, the Aegis system was not designed to learn from its decisions, and its reactions were limited by its rules.
“If a system uses ‘if/then,’ it’s probably not machine learning, which is a field of artificial intelligence that involves creating systems that learn from data,” said Air Force Lt. Col. Christopher Berardi, who is based at the Massachusetts Institute of Technology. Technology that will aid the Air Force’s artificial intelligence development.
Artificial intelligence took a big step forward in 2012, when the combination of big data and advanced computing power enabled computers to begin analyzing information and writing rule sets themselves. This is what AI experts call the “big bang” of AI.
New data created by a computer writing the rules is artificial intelligence. Systems can be programmed to act autonomously based on conclusions drawn from machine-written rules, a form of autonomy enabled by AI.
TESTING AN AI ALTERNATIVE TO GPS NAVIGATION
Air Force Secretary Frank Kendall got a taste of this advanced warfare this month when he flew Vista, the first F-16 fighter jet controlled by artificial intelligence, during a dogfight exercise over California’s Edwards Air Force Base.
While this jet is the most visible sign of ongoing AI work, there are hundreds of AI projects ongoing at the Pentagon.
At MIT, service members worked to sift through thousands of hours of recorded pilot conversations to create a data set from the flood of messages exchanged between crews and air operations centers during flights; so the AI could learn the difference between critical messages, such as the closure of a runway. and casual cockpit chatter. The goal was to let the AI learn which messages were critical to amplify to enable controllers to see them faster.
In another important project, the army is working on an artificial intelligence alternative to GPS satellite-based navigation.
In a future war, high-value GPS satellites would likely be shot down or interfered with. The loss of GPS could blind U.S. communications, navigation and banking systems and reduce the ability of the U.S. military’s fleet of aircraft and warships to coordinate the response.
Last year, the Air Force flew an artificial intelligence program loaded onto a laptop strapped to the floor of a C-17 military cargo plane to study an alternative solution using the Earth’s magnetic fields.
It has been known that planes can navigate by following the Earth’s magnetic fields, but until now this has not been practical because each plane produces so much of its own electromagnetic noise that there is no good way to filter only the Earth’s emissions.
“Magnetometers are very sensitive,” said Col. Garry Floyd, director of the Department of the Air Force-MIT Artificial Intelligence Accelerator program. “If you turn on the strobe lights of a C-17, we can see it.”
Floyd said the AI learned through flights and reams of data which signals to ignore and which to follow, and the results were “very, very impressive.” “We’re talking about tactical airborne quality.”
“We think we might have added an arrow to the quiver of what we can do if we work in an environment where GPS is blocked. We’re going to do that,” Floyd said.
So far, AI has only been tested on the C-17. Other aircraft will also be tested and, if working, could offer the military another way to operate if the GPS fails.
SAFETY RAILS AND PILOT TALK
The AI-controlled F-16 Vista has significant safety rails, as the Air Force trains. There are still mechanical limits that prevent the learning AI from performing maneuvers that would endanger the aircraft. There is also a safety pilot that can take over control from the AI with the press of a button.
The algorithm cannot learn in flight, so each time it only has the data and rule sets it has built up from previous flights. Once a new flight is finished, the algorithm is transferred back to a simulator, where it is fed with new data collected during the flight, from which it learns, creates new rule sets, and improves its performance.
However, artificial intelligence learns fast. Because of the supercomputing speed the AI uses to analyze data and then use these new rulesets to fly in the simulator, its speed at finding the most efficient way to fly and maneuver has led it to beat some human pilots in dogfighting exercises.
But security is still a critical concern, and officials said the most important way to consider security is to control what data is reintroduced into the simulator so the AI can learn. In the jet’s case, ensuring the data reflects safe flight. Ultimately, the Air Force hopes that a version of the artificial intelligence under development could serve as the brains for a fleet of 1,000 unmanned fighter jets being developed by General Atomics and Anduril.
In an AI training experiment on how pilots communicate, soldiers assigned to MIT scrubbed records to remove classified information and the pilots’ sometimes salty language.
Learning how pilots communicate “is a reflection of command and control, how pilots think. If they’re going to be really good, the machines need to understand that, too,” said Deputy Chief of Staff Grady. “They don’t need to learn to swear.”