[infurl]

https://techxplore.com/news/2020-10-breakthrough-energy-efficient-artificial-intelligence.html

This seems like something to get excited about.

Thanks to a mathematical breakthrough, AI applications like speech recognition, gesture recognition and ECG classification can become a hundred to a thousand times more energy efficient. This means it will be possible to put much more elaborate AI in chips, enabling applications to run on a smartphone or smartwatch where before this was done in the cloud.

Under supervision of CWI researcher and UvA professor cognitive neurobiology Sander Bohté, researchers developed a learning algorithm for so-called spiking neural networks. Such networks have been around for some time, but are very difficult to handle from a mathematical perspective, making it hard to put them into practice so far. The new algorithm is groundbreaking in two ways: the neurons in the network are required to communicate a lot less frequently, and each individual neuron has to execute fewer calculations.

[HS]

I don't have much of a frame of reference. Is this comparable to AC/DC currents, with the non continuous math wasting less energy on unnecessary things? Is it more like only pedaling uphill, and coasting the rest of the time? Could those analogue neurons from the "Memristor Breakthrough" (picture below) be made to approximate this style of function, where they pick their moments to communicate? Can any mathematical formula be created as an analogue structure? Interesting stuff.

[infurl]

A closer analogy might be like the difference between switching and linear amplifiers or power regulators, but what it amounts to is not wasting energy on things that aren't changing, just responding to events and remaining at rest when nothing is happening. That would be complicated to model mathematically, much like integer programming is much more complicated than linear programming in operations research.

https://en.wikipedia.org/wiki/Integer_programming