All that being said, the MIT researchers have exposed a potentially concerning aspect of a widespread autonomous fleet that hasn’t been considered up until now.
“These vehicles could actually be using a ton of computer power,” said Sertac Karaman, associate professor of aeronautics and astronautics. “They have a 360-degree view of the world, so while we have two eyes, they may have 20 eyes, looking all over the place and trying to understand all the things that are happening at the same time.”
An example was given where one autonomous vehicle has 10 deep neural networks. These process images from 10 cameras, and over one hour of commuting, that’s 21.6 million inferences in a day. That rises to 21.6 quadrillion inferences (1 quadrillion is 1,000 trillion if you can wrap your head around such numbers) for one billion cars. Consider that Facebook’s global data centers make a few trillion inferences each day, and that puts into perspective the emissions price that we may have to pay for a large autonomous fleet.