• United States

Edge-chips could render some networks unnecessary

News Analysis
Dec 06, 20183 mins
Data CenterNetworking

Neuristor circuits will perform super-edge computations similar to how the brain does them, reducing network loads and possibly making some networks redundant.

virtual brain / digital mind / artificial intelligence / machine learning / neural network
Credit: MetamorWorks / Getty Images

Hardware processing should replace a device’s dependency on networks, some scientists say. Making machines more efficient, saving power and resilience are behind the reasoning.

“Devices like drones depend on a constant Wi-Fi signal. If the Wi-Fi stops, the drone crashes,” an article about researchers at Binghamton University in Binghamton, New York, says.

But if you make a device independent of any linking, it could become more resilient, the researchers say. Plus, the more processing work one can do on the machine the more energy you’ll save because you won’t have to come up with power to communicate.

“You could put 5G and 6G everywhere and assume that you have a reliable internet connection all the time, or you could address the problem with hardware processing, which is what we’re doing,” Louis Piper, associate professor of physics and director of materials science and engineering at the university, says in the article.

The Binghamton researchers, along with researchers at Georgia Tech, are working on developing a kind of neuristor circuit that will allow all device processing to take place at the chip level, meaning there’ll be no network load or indeed any requirement to communicate using a network at all. Neuristor circuits are brain-copying computer chips.

“The idea is we want to have these chips that can do all the functioning in the chip, rather than messages back and forth with some sort of large server,” Piper says.

By doing that, power is saved, but also the machine becomes powerful enough to react to its environment without having to query a larger set of machines somewhere else. That’s not only faster and maybe more reliable, but it also saves energy. One could look at it as edge networks gone extreme.

Neuristor circuits work like neurons in the brain

The man-made neuristor circuits in development try to copy actual biological neurons in the brain. Neurons are the electrically responding nerve cells and fibers in the brain that process information. They send signals to cause muscle contractions, and so on—communicating with the spinal cord and nerves with, importantly, very little energy consumed.

Neuristor, brain-replicating electronics circuits were first theorized in 1962, and by 2013 were being tested using a material called niobium dioxide (NbO2), the academics write in their paper, published by Nature. The problem with the circuits, however, has been that they require a large voltage and complicated, related fabrication called electroforming to create the only theoretically efficient switching developed thus far. Creating the reality actually defeats the object.

“Like with Frankenstein’s monster, you basically pulse a large amount of electricity through the material, and suddenly it becomes an active element,” Piper says. “That’s not very reliable for an engineering step with fabrication.”

Scalability is a problem, for example.

But it’s in this area that the scientists say they have been making breakthroughs, and the team says it’s now going to be able to perform the switching functions without the unwieldly bolt of electricity.

“You can build a neuristor out of this, and because you don’t need the electroforming, it’s more reliable — and what you can build an industry on,” Piper says.

The group’s research could lead to more inexpensive, energy-efficient, and high-density neuristor circuits and give us more energy-efficient and adaptable computing sooner.


Patrick Nelson was editor and publisher of the music industry trade publication Producer Report and has written for a number of technology blogs. Nelson wrote the cult-classic novel Sprawlism.

The opinions expressed in this blog are those of Patrick Nelson and do not necessarily represent those of IDG Communications, Inc., its parent, subsidiary or affiliated companies.