Building autonomous vehicle prototypes checks off all these boxes.
The fact that the real payoff of this R&D lies years -- if not decades -- down the road makes it all the more daunting for established automakers and startups alike.
That's why finding means of testing autonomous systems outside of cars is an attractive alternative. Microsoft recently announced its researchers' development of autonomous gliders. While the main focus of the news centered on planes controlled by artificial intelligence up in the air, the technology could be utilized in areas with significantly lower elevation, namely autonomous cars.
"What makes Microsoft's sailplane project so useful to all autonomous vehicle development is that it provides a relatively cheap platform for testing and training AI agents to independently operate a machine," Neel V. Patel wrote in an article for Inverse. "Although the sailplane is just one specific type of vehicle, the algorithm that makes up the AI system is designed to physically navigate a machine around in three dimensions. There aren't many factors that separate what an algorithm must do to correctly operate a glider versus a drone, or land vehicle like a car."
While the average new passenger car weighs over two tons, the sailplanes tested by Microsoft resemble model airplanes at 16.5 feet long and 12.5 pounds. The artificial intelligence software that powers the sailplanes still represents millions of dollars in development. But the fact that it is housed in a physical body that is cheaper by an order of magnitude than a car is significant.
There are many overlapping functions of a self-piloted glider and a self-driving car.
Each must perceive and understand their physical surroundings. For most self-driving cars, this primarily means using laser systems to detect other cars in their immediate vicinity, as well as unexpected objects in the road. Microsoft's gliders, on the other hand, are more concerned with identifying things like wind direction and air temperature. The decision-making algorithms that both AVs use to interpret this received information, though, are similar.
Similar, but not interchangeable, according to AI researcher Peter Stone.
"Many people think of AI as one single entity that can be sprinkled onto all sorts of things," Stone told Inverse. "And that's just not true. [That said,] the algorithms developed to deal with decision making and perception can be generally applied."
While they're doing different work, developers of artificial intelligence, regardless of vehicle type, are looking to solve the same underlying issues. If a solution is discovered in self-piloted aircrafts, you can bet that a similar breakthrough in self-driving cars will follow soon after.