While the US military has been flying autonomous planes for decades, machine learning has historically been prohibited due to its high risk and lack of independent control.
“We have to be able to trust these algorithms to use them in a real-world setting,” said Lt Col Ryan Hefron, the programme manager at DARPA.
Key steps forward
Since DARPA founded its Air Combat Evolution (ACE) programme in 2019, it has made several key steps forward in rendering AI-powered aircraft safe for use.
In 2019, AI software was used against a human F16 pilot in a simulator for the first time, trouncing them 5-0 in a series of head-to-head dogfights.
Since the inception of the VISTA jet, a series of safety instructions have been implemented in order to ensure it avoids collisions and protects human safety where possible, making it suitable for real-life manoeuvres.
Source link