AI system defeats US military pilot in F-16 aerial combat | Scientific and technical news


An artificial intelligence (AI) system defeated a US military pilot in a five-round challenge to demonstrate whether autonomous computer systems might be able to fly fighter jets.

The U.S. Defense Advanced Research Projects Agency (DARPA) conducted its last AlphaDogfight test this week as part of a program to see if autonomous systems could defeat opposing aircraft in simulated air combat.

It is a precursor to DARPA’s Air Combat Evolution (ACE) program, which will eventually fly planes live using AI algorithms to demonstrate that the technology will work in the real world.

One of AlphaDogfight’s simulated battles. Pic: DARPA

On Thursday, the AI ​​developed by Heron Systems defeated an experienced F-16 pilot in each of the five rounds – although the success of a simulation is only a step towards fully autonomous systems used on F fighter jets. -16.

The timing isn’t quite as big as when IBM’s computer Deep Blue beat the former world chess champion Garry Kasparov in 1997, effectively confirming that computers were capable of playing chess better than humans.

Chess is a game with a finite number of possible moves, and accommodating all possible moves to defeat any human opponent has proven to be well within the capacity of a modern computer.

Although the computer simulation for the air combat program is much more complicated than the game of chess, it is also constrained by the number of moves and variables that have already been coded in the simulation – and there is no such constraints on air combat in real life.

The simulation also featured a World War II– style scenario for airplanes, where the air battle was based on the idea that the fighter plane only had cannons facing forward to attack each other.

Modern air combat can be much more complex, especially when it includes missiles, which were not used in simulations.

U.S. Air Force Lt. Col. Dan Javorsek, who is DARPA’s ACE program manager, was cautious about the limitations of simulation to prove AI superiority over a human pilot.

Lt. Col. Javorsek said: “Just because the outcome turned out the way it did today doesn’t mean the AI ​​didn’t grab onto a strange digital artifact we weren’t thinking about. not or that we didn’t think of, and that’s what gave them an advantage. ”

“There were definitely some peculiar behaviors in engagement,” he added, before pointing out that the result was always a human triumph, as the humans developed Heron’s successful AI system.


Please enter your comment!
Please enter your name here