Artificial intelligence (AI) is continuing to advance and has now defeated a human fighter pilot in a virtual combat simulation.
This result was achieved in the US Army’s AlphaDogfight competition, which aimed to demonstrate the feasibility of developing autonomous agents capable of defeating enemy aircraft in aerial combat.
Eight teams participated in a series of tests and in the final last Thursday, Heron Systems was declared the winner after two days of aerial combat using only nose-mounted guns.
Heron Systems then faced a human fighter pilot seated in a simulator and wearing a virtual reality headset, winning five rounds to zero.
The artificial intelligence also demonstrated its effectiveness in deep reinforcement learning, where algorithms can repeatedly test a task in a virtual environment until they develop an understanding.
Lockheed Martin’s Vice President of Artificial Intelligence, Matt Tarascio, and their Chief AI Architect, Lee Ritholtz, explained that teaching an algorithm to perform well in aerial combat is very different from teaching it to “fly”.
The software starts with a complete lack of understanding, even of the most basic flying tasks, which puts it at a disadvantage compared to any human at first.
However, this aerial combat system doesn’t require a lot of processing power and can run on a chip similar to the NVIDIA Tegra. It’s worth noting that this isn’t the first time that AI has fought against a human fighter pilot in a competition.
In 2016, a demonstration showed that an AI agent named Alpha could beat an experienced human combat flight instructor.
The DARPA simulation was possibly more significant, as it pitted a variety of AI agents against each other and then against a human in a highly structured framework. This demonstrates that AI is continuing to advance and its use in different fields is becoming increasingly effective.
