As a species, we are at a point where the technology we create has the ability to surpass us humans at many tasks we perform on a daily basis. And while that is a welcome development in most fields, trusting technology such as AI to control weapons systems is something most humans would be very wary of.
blame terminator Whether it's a movie or something you like, most of us can't help but admit that every time we hear about an AI doing something it couldn't do, our heart rate increases a little. yeah. Especially when it involves the mind of a computer piloting a fighter jet in a real-world dogfight against a human enemy.
I'm not sure how many people know this, but several years ago the Defense Advanced Research Projects Agency (DARPA) launched a program aimed at teaching artificial intelligence how to beat human pilots in air combat. It is running from.
The program is called Air Combat Evolution (ACE), and although it sounds like a video game title, it's not. Because the ultimate goal of this program is to provide humans with another tool to accomplish their mission.
Why? It's really simple. If the AI knows what to do in an air combat situation, and the humans flying with her trust her AI enough, the computer's brain will be able to handle small-scale combat and outdo the human. Your brain can stay focused on the bigger dogfight.
For example, while the AI engages in dogfights to eliminate targets within visual range, pilots can focus on remotely flying swarms of drones operating in the area.
DARPA has been testing AI algorithms since announcing the program in 2019, but last year it tested the system in real life and a specially modified F-16 fighter jet called Variable. Moved to world testing. In-flight simulator test aircraft (VISTA or X-62A).
Photo: DARPA
This plane is a derivative of the F-16D and is the work of General Dynamics and Calspan. This aircraft is regularly used at the Air Force Test Pilot School and can accommodate two pilots. The General Electric jet engine produces 28,600 pounds of thrust when the afterburners are hot, giving it a top speed of Mach 1.2 (915 miles per hour/1,460 kilometers per hour).
It is this VISTA plane that DARPA used to test the AI. Although details of the systems used during the flight were not disclosed, it is important to note that computers were not the only ones on board the aircraft during the flight. A human pilot was on board, ready to take over if anything went wrong.
In the visual range combat scenario being conducted, the AI's opponent was another fighter controlled solely by human hands and brains.
DARPA did not say specifically when this sequence of events would occur. “First-ever aerial test of an AI algorithm that allows a fighter jet to fly autonomously against a human-piloted fighter jet” The flight took place, but the video released last week was shot in 2023. It is known that the flight took place over Edwards Air Force Base in California, but its results remain a secret.
The clip below primarily shows members of the ACE team weighing in on the reasons for the DARPA program.How it represents a transformative moment in aerospace history, unlike other aerospace autonomy projects. ” A short example of an air battle is shown, the first time such images have been published.
The biggest challenge DARPA appears to face here is not the technology itself, but getting pilots to trust the plane's AI to save their lives. This is a key aspect of his ACE program, currently keeping the range of AI pilots within set boundaries.
Photo: DARPA
According to Air Force Lt. Col. Dan Javorsek, “Air engagement scenarios become more difficult and realistic only after human pilots are confident that AI algorithms can be trusted to handle limited, transparent, and predictable behavior.”
What this means, at least for now, is that the tests being conducted this year are not prohibitively complex. However, AI may end up being involved in more difficult scenarios, as this is the only way AI can evolve.
To date, the AI has flown 21 times, during which time Air Force programmers have modified more than 100,000 lines of code. And no wonder, considering that air combat training is probably as dangerous as air combat itself.
According to U.S. Air Force statistics, from 2000 to 2019, more than 32 aircraft were destroyed and 23 were damaged during air combat training flights. There were 27 mid-air collisions, resulting in the loss of 12 air force personnel.
Ideally, with the advent of AI, humans could of course learn how to confidently place their lives in the hands of computer programs during the most difficult missions military aviators have ever had to face. Everything changes.
