In 1997, IBM super computer Deep Blue won a chess game 3 1/2 to 2 1/2 against the reigning world champion and world N°1 for 20 years Gary Kasparov. It was a symbolic milestone, the first time the machine was wining over humans at such a complex game.
Twenty years later, in 2016, Google Deep Mind's AlphaGo won a Go game against the reigning and 18-time world champion, Lee Sedol by 4 wins to 1. That was the victory of artificial intelligence and deep learning at a game where there were even more possibilities and scenarios to handle than chess. Another milestone of the "rise of the machines".
In August 2020, we saw yet another milestone in this series of defeat of humans vs. machines. And again at one of the most complex tasks ever handled by humans: piloting a F16 fighter jet in a combat mission. DARPA (Defense Advanced Research Project Agency), the research branch of the U.S. Department of Defense, which is famous for the research that lead to the creation of Internet, organized the simulation called Alpha Dog Fight. A air-to-air duel between aircraft in the spirit of those of WW2.
One one side, the Heron AI algorithm, on the other side a qualified F16 Fighter jet pilot instructor. During the whole simulation, the human pilot could not get a hit against the jet piloted by the machine... and lost all of the 5 simulated dogfights!
DARPA's goal is to have unmanned aircrafts capable of engaging successfully in air-to-air combat missions. There is still progress to be made for the machine to pilot the latest jet fighter in use but we are getting a step closer. A victory based on machine learning and the digestion of 4 billion simulations.
(Source: DARPA)
תגובות