An AI system that could run on a Raspberry Pi computer has comprehensively beaten a former fighter pilot in simulated air combat.
The system, ALPHA, appears to triumph over human abilities with two of the most common advantages of computers over people: speed and endurance. The system is said to be able to figure out and set up the best plan and response in combat 250 times quicker than a human can blink.
After successful testing in simulations against other systems, ALPHA took on retired US colonel Gene Lee. He’s a former pilot and trainer who has been taking on computer challengers since the early 1980s. In every simulated battle during “protracted” tests, ALPHA shot Lee down.
According to Lee, what makes the system different to other he’s faced is its ability to figure out the move or shot he was about to take and then counter it in time. He also noted that the lengthy tests, designed to replicate the duration of real missions, left him mentally exhausted.
It doesn’t appear such systems are likely to replace human pilots however. Instead, according to the University of Cincinnati, which worked on the ALPHA project, it’s more likely it would be used to power unmanned vehicles that would fight alongside piloted fighter planes.
The system runs through what’s described as a Genetic Fuzzy Tree approach. That means that although it still crunches the numbers on the various factors that influence the best course of action, it’s set up to only concentrate on a select set of variables that are relevant to the specific situation it has to deal with at any precise moment.