Artificial intelligence defeated a US Air Force pilot with a score of 5:0 [video]
WASHINGTON, (BM) – The Defense Advanced Research Projects Agency (DARPA) of the US Department of Defense has summed up the results of the AlphaDogfight Trials virtual air combat tests of neural network algorithms and they are very disappointing for humans: artificial intelligence was able to defeat a professional US Air Force pilot with a score of 5:0, learned BulgarianMilitary.com.
The AlphaDogfight Trials ended yesterday with the victory of the neural network algorithm in all five virtual battles of the third stage of the test. In the finals, an experienced US Air Force instructor pilot with the callsign Banger competed against artificial intelligence.
The results of the competition will form the basis of the Pentagon’s project to develop a trainable artificial intelligence system Air Combat Evolution, which in the future will be able to take control of a combat aircraft and conduct maneuverable close air combat.
The ultimate goal of the ACE program is to create an artificial intelligence system for combat drones, including unmanned wingmen who will cover manned fighters.
The new neural network system will have to conduct air combat faster and more efficiently than humans. In the first two stages of the AlphaDogfight Trials, neural network algorithms controlled the F-15C Eagle heavy fighters, and in the third, the F-16 Fighting Falcon medium fighters.
At the third stage of testing, neural network algorithms conducted air battles with each other. In all battles, the winner was a system developed by Heron Systems. Air battles were fought at close range using only cannon armament. Then the system was exposed to a virtual highly maneuverable air battle with a real pilot, but what type of neural network the developers used was not disclosed.
It is known that during training, the neural network performed four billion simulations. The second result in virtual aerial combat was shown by an algorithm developed by Lockheed Martin.
Eight companies were admitted to the third stage: Aurora Flight Sciences, EpiSys Science, Georgia Tech Research Institute, Heron Systems, Lockheed Martin, Perspecta Labs, PhysicsAI and SoarTech. They took place in the FlightGear simulator using the JSBSim algorithm.
Follow us everywhere and at any time. BulgarianMilitary.com has responsive design and you can open the page from any computer, mobile devices or web browsers. For more up-to-date news from us, follow our YouTube, Reddit, LinkedIn, Twitter and Facebook pages. Do not miss the chance to subscribe to our newsletter. Subscribe and read our stories in News360App in AppStore or GooglePlay or in FeedlyApp in AppStore or GooglePlay
Subscribe to Google News.