How much energy can dust-sized computers harvest from sun and motion, and how much work can they do with it?

Pete Warden reports in from the ARM Research Summit, where James Myers presented on "energy harvesting" by microscopic computers -- that is, using glints of sunlight and the jostling of motion from bumping into things or riding on our bodies to provide power for computation.

Here's a sample: human vibrations generate 4µW/cm2, industrial vibrations generate 100µW/cm2, outdoor light generates 10mW/cm2, and the stray radio energy from wifi can provide 0.001µW/cm2.

That's not much!

However, computing can be done with very little energy: "existing hardware like DSPs can perform a multiply-add for just low double-digit picojoules."

But at those low power levels and tiny feature sizes, devices need to do a lot of error checking, because fundamental physics creates a lot of computation errors when you get really, really small. Which brings me to Warden's conclusion: because neural nets are resilient to random noise, we may be able to use fast-advancing AI techniques to bind together the computing work done by these little motes.

I was actually very excited when I learned this, because one of the great properties of neural networks is that they’re very resilient in the face of random noise. If we’re going to be leaving an increasing amount of performance on the table to preserve absolute reliability for traditional computing applications, that opens the door for specialized hardware without those guarantees that will be able to offer increasingly better energy consumption. Again, I’m a software engineer so I don’t know exactly what kinds of designs are possible, but I’m hoping that by relaxing constraints on the hardware the chip creators will be able to come up with order-of-magnitude improvements, based on what I heard at the conference.

If we can drive computational energy costs down into the femtojoules per multiply-add, then the world of ambient sensors will explode. As I was writing, I ran across a new startup that’s using deep learning and microphones to predict problems with machinery, but just imagine when those, along with seismic, fire, and all sorts of other sensors are scattered everywhere, too simple to record data but smart enough to alert people when special conditions occur. I can’t wait to see how this process unfolds, but I’m betting unreliable electronics will be a key factor in making it possible.

AI and Unreliable Electronics (*batteries not included)
[Pete Warden]

(via 4 Short Links)