In the April 18, 2024 issue of The Telegraph, journalist Cameron Henderson published an article titled “History made as U.S. Military conducts first ever human vs AI dogfight”, and reported something that profoundly disturbs me. He wrote that in September 2023, a USAF F-16 fighter pilot went through a series of engagements with another F-16 controlled by Artificial Intelligence (AI). He said: “Travelling at speeds of up to 1,200 miles per hour, the two jets practised both defensive and offensive scenarios as well as within-visual-range combat, known as dogfighting. At one point they came within 2,000 feet (610 metres) of each other.” This is the first time that an AI pilot has flown in the air against a human pilot, and sadly I think it just could be the beginning of the end for human fighter pilots.
People have predicted that humans would eventually be replaced by computers in the cockpit for decades, but until now it has only happened in simulators. Henderson said that back in 2019, a dogfight between a human USAF fighter pilot and an AI pilot in a simulator resulted in 5-0 victory for the machine. After that, much alleged progress has been made: “Since it was first built in December 2022, the [AI] jet has been taken out on at least 21 test flights, totalling more than 17 hours of flight time and the first-time machine-learning has been used to pilot a fighter jet. During flight, the AI algorithm on the jet analyses data and makes real-time decisions, a process called machine learning, that mirrors the way in which fighter pilots hone their instincts over years of practice. Carrying out a dogfight between an AI-powered jet and a human marks a ‘transformational moment in aerospace history’, DARPA [Defense Advanced Research Projects Agency] said in a statement.”
The article did not actually come out and say the AI pilot won any engagements this time, which could be very good news, but nevertheless the USAF is apparently very pleased with this development. This is disturbing to people like me, who see good reason to be sceptical about AI, especially the potential to replace human beings in various occupations. On the other hand, the fact that the USAF apparently did not release any information about who won the dogfight could be very telling. Perhaps the man or woman won, and if so, they can keep their jobs, for now.
If it turns out that the AI pilot won, it will be a pivotal moment for the USAF and all the world’s air forces for that matter. However, if I had my way, this would be the moment that the USAF abandons AI in the cockpit, as a duty to its pilots and humanity, and increases the flying hours for human pilots. And I would really like to see a treaty with other powers that bans AI in combat now, before humans become totally secondary to machines. Some might think I am an alarmist, but in addition to protecting jobs, I see AI in combat as unethical. For example, we should ponder what will happen to an AI pilot that mistakenly shoots down a friendly aircraft. Who will be held accountable? I do not have the answer, do you?
Mark my words, the day we abandon human capabilities, reasoning, judgement, and morality in favor of AI will be the end of us all. AI is simply too dangerous and there are too many unknowns to proceed with these experiments. I will be following the developments on this closely and you should too if you care about people more than machines.
Roger Thompson is a research fellow at Dalhousie University’s Centre for the Study of Security and Development, the author of "Lessons Not Learned: The U.S. Navy’s Status Quo Culture", and a former researcher at Canada’s National Defence Headquarters.
