AI at the edge. A brief explanation.

AI at the edge. A not so brief and simple example.

At my home we use Alexa for everything. I chose Alexa after conducting many tests with Alexa, Google and Apple Homekit.

Each one has its detractors. But Alexa has the worldwide cloud server base that outperforms the others.

At my home Alexa controls most things. Lights, Aircon, Heaters, Door sensors, Music, TV's, Garage doors, Air purifiers, electrical switches etc. We control them by voice in nearly all cases. I rarely use a light switch at home, and I have routines for turning things on and off.

I have over 60 IOT devices connected.

All of them have one thing in common....they are all cloud based....meaning I can control them even if I am not at home as long as they are connected to the internet and hence the cloud. I can also set routines to turn lights on an hour before sunset etc and that runs locally. But everything else relies on the cloud which can cause a lag in operation, but it works.

That is all great until one day when the house loses connection to the internet and the cloud is not available. Oop's.

AI at the edge will eliminate that problem.

A device with AI at the edge will not need the cloud. It will learn to use the internal network to control devices so that the internet is not required as long as the home or office wifi is running....no problems. In a blackout who cares....nothing works anyway.

AI in edge devices will process the commands required on the device instead of going to the cloud for instructions.

If I say "Alexa turn on the bedroom lights", Alexa goes to the cloud...it interprets what I said....then goes to Lifx cloud and tells Lifx to turn on the bedroom lights. This all takes time...approx 2 seconds actually. 2 secs is not a long time right now....but for ADAS its 120 meters of travel or the difference between stopping the car in time and killing some folks.

AI on the edge will not need to do that....it will know that Lifx is on my network and work out the command to send to it direct...no lag and no cloud required. In ADAS that means control of the car’s operations in an instant.

I did mention Wyze camera's in an earlier post. They had xnor.ai built into them...on the edge. These cameras could distinctly identify humans versus anything else, including my cats, almost instantly because the code was based on a chip internal to the camera. I have not seen a camera do the same thing since. Sorry to say Apple bought xnor.ai and Wyze was forced to remove the AI from its cameras. They replaced it with standard person detection which is rubbish because by the time the camera sends the image to the cloud and back...its all-over red rover. Person has gone out of view. Lots of false positives or no capture at all. This is what AI on the edge is capable of.

In ADAS this is the holy grail...and one that BRN is most suited for.

There will be thousands of edge-based AI devices in the future...most likely millions. Food for thought.

I could go on and on with my research in this space....from traffic lights to baby monitors....but I will leave that for another time.

It would seem the time is so very right for BRN IMHO.

Sorry for the long-winded post.