When two U.S. Air Force fighters were recently sent into a combat demonstration, only one pilot was at the controls. The other fighter was piloted by artificial intelligence. The demonstration showed how far the U.S. military has come in terms of technology. But it was supposed to be just a taste of what one of the most powerful militaries in the world has at its disposal.
Superpowers are arming themselves with AI
The United States is competing with China on artificial intelligence and its use in weapons systems. “Whether you want to call it a race or not, I think it is,” says Christopher Grady, vice chief of staff at the U.S. General Staff. “Both we and the Chinese recognize that in the not-too-distant future, artificial intelligence is going to play a key role on the battlefield. That’s why everyone is working on it with equal intensity,” he explains.
At the same time, there is a growing fear among the public that future wars will be fought by machines. However, the American armed forces assure us that this will not happen, at least not on the US side. How far a potential adversary can go remains to be seen. And the United States is willing to meet any challenge.
From Machine Learning to Autonomous AI
The key to using AI in the military is combining machine learning and autonomy. In the former, a computer analyzes data and sets of rules to draw conclusions. Autonomy occurs when all the data is used to take specific actions without further human intervention.
The precursor was the development of the Aegis Combat System air and missile defense system in the 1960s and 1970s. It was trained on a series of “if-then” rules, allowing it to recognize and intercept incoming missiles on its own—far faster than a human could. However, the system’s responses were limited only to predefined rules, and the system itself was not designed to learn from its decisions.
The “Big Bang” of Artificial Intelligence
In 2012, AI took a giant leap forward when the combination of large amounts of data and advanced computing power allowed computers to process information and write rules on their own. Experts refer to that breakthrough as the “big bang” of AI.
The Pentagon currently has several hundred AI projects underway. One of them is based on analyzing thousands of hours of cockpit conversations. From the flood of words and messages, researchers at the Massachusetts Institute of Technology (MIT) have created a dataset that AI can use to learn to distinguish critical information from ordinary conversations. This will allow AI to quickly alert to dangerous situations.
For the US, security is a key issue
The fundamental issue in all American projects is safety. In the case of an AI-controlled fighter, experts emphasize that it is necessary to precisely control what data is fed into the simulator from which the AI learns. For each flight, the AI has only one specific set of data, which is systematically exchanged before each subsequent flight.
In addition to safety, the AI also places a huge emphasis on data utility. “When it comes to cockpit conversations, the AI needs to understand how the communication is happening,” Admiral Grady explains. “That it’s largely a reflection of the command and control system and the way the pilots think. But we shouldn’t feed the AI with vulgar language. It doesn’t need to learn to swear,” he adds with a smile.
Source: ZDFheute