A major trend in AI is the transition from computing to cloud computing. Rather than having AI devices perform remote computing over an Internet connection, they are increasingly processing things locally, using algorithms directly on the device. The benefits of this approach may include faster results, greater security and greater flexibility. But how far can you push this model?
Launching the Xnor based on Seattle is definitely right on the verge of bleeding. This week, the company unveiled a prototype of an AI camera that completely turns off solar energy – neither a battery nor an external power supply is needed. The camera has a resolution of 320 x 320, a FGPA chip for processing, and is loaded with an art recognition algorithm.
It is possible theoretically to stick a device like this somewhere out into the street and transfer data back to you indefinitely. This is compatible with several different low-power wireless protocols (since Wi-Fi takes too many batteries), which allows you to send information for tens of kilometers. And, says Xnor, if you put it with a battery, it can store enough energy throughout the day to keep it running during low sunshine hours and at night.
"We are investigating several cases of using these devices right now," says Mohammad Rastegari, technical director of Xnor The Verge . "From large-scale civilian projects to monitoring inside autonomous car cabin systems, to attach cameras to drones "
miniaturization. It was torn off from the Institute of Artificial Intelligence Alain in 2017 on the reverse side of the proprietary method of creating super-efficient systems of machine learning. (The key to this technology is the type of logic circuit, in home as the gateway to XNOR, hence the name.) This is also proven by the utility of its software by launching it on low-power, low-voltage devices like the Raspberry Pi Zero.] This sunny AI camera is not yet ready to sell though. Although it's completely self contained, there are certain For example, the number of frames that it can handle every second depends on how much the sun receives it. Xnor says that on a sunny day it runs at 32 frames per second, but it can be offset by larger solar cells.
It's clear that such devices will only become more widespread in the future. They are relatively cheap (the Xnor model costs $ 10) and is more convenient for operators. Since the photos and videos they accept never leave the device, they are potentially more private.
But the question remains: do we enjoy the world full of AI eyes that we always watch? This is a problem we are already faced with in the context of smart video surveillance, and as the Xnor company shows, this technology will be only less and more unpredictable.