Researchers have developed an artificial intelligence that can defeat human pilots in combat. This new AI, known as ALPHA, is designed for military drones or ‘unmanned aerial vehicles’ (UAVs).
ALPHA was tested in simulations and the results have been published in the Journal of Defense Management. The first versions beat AI systems previously used by the U.S. Air Force Research Lab. Advanced versions were then assessed in a flight simulator by retired Air Force Colonel Gene Lee, an expert in combat tactics.
After his first fight against ALPHA, Lee said, “It seemed to be aware of my intentions and reacting instantly to my changes in flight and my missile deployment. It knew how to defeat the shot I was taking. It moved instantly between defensive and offensive actions as needed.”
According to Lee, most AI opponents can’t keep up with the pace and pressure of real combat-like scenarios and can be beaten by experienced human pilots. But during simulated aerial engagements with ALPHA, Lee couldn’t score a single kill and was repeatedly shot out of the air.
ALPHA has defeated a few other experts too, even when given handicaps in speed, turning, sensors and missiles. Colonel Lee, who has been testing systems since the early 1980s, says it’s “the most aggressive, responsive, dynamic and credible AI I’ve seen to date.”
The AI is based on an algorithm created at the University of Cincinnati by aerospace engineers Kelly Cohen and Nick Ernest. ALPHA itself is being developed by the company Psibernetix. Ernest, now CEO of Psibertentix, says, “The goal is to continue developing ALPHA, to push and extend its capabilities, and perform additional testing against other trained pilots.”
ALPHA can currently process sensor data and plan combat moves for four drones in less than a millisecond, or over 250 times faster than the eye can blink — reaction times far beyond human abilities.
The AI isn’t powered by a super computer — it runs on an ordinary $500 desktop PC, which means high performance at low cost. This is achieved through efficient algorithms.
Although we usually expect machines to process all the available data before making decisions, ALPHA takes a more human approach to solving problems, simplifying variables to only consider the most relevant information — to work out whether an opponent is better, for example.
Instead of using numbers for precise parameters, ALPHA’s algorithms are based on language or ‘fuzzy logic’ — it makes decisions via if-then rules. That reduces the number of branches in a decision-making tree, which lowers the computing power required to find the best strategy.
ALPHA is built on a so-called ‘genetic algorithm’ — rules inspired by how genes are inherited from one generation to the next. Genetic algorithms are a type of ‘evolutionary algorithm’, mimicking the process of adaptation by natural selection (survival of the fittest). In this case, bits of computer code are ‘bred’ to create new combinations, with the most efficient versions chosen by the selection process.
Drones powered by artificial intelligence could evade attacks while continually adapting to enemy tactics. While a drone could be a wingman for manned aircraft, it could also be used to fight aerial battles without humans at all.
This article was written by JV Chamary from Forbes and was legally licensed through the NewsCred publisher network.