Three things you should know about neuromorphic engineering
By Charlene Chin
(It mimics a brain’s processes to run machines. )
Imagine a mouse. It scurries in an alleyway, sniffs around some garbage and decides, finally, to feast on some diner scraps. This happens in a moment, and you can’t even see the rapid fire neurons in the rodent’s brain.
TrueNorth is a neuromorphic chip that has been described as the equivalent to the brainpower of a rodent. It mimics natural processes to make decisions, rather than using binary code.
Why is breakthrough in the tech so important? GovInsider has set out three things that the public sector should know about neuromorphic engineering.
1. What is it?
Neuromorphic engineering is computing that is “biologically inspired” to mimic a brain’s neuro-sensing, which consists of thousands of interacting networks.
“We try to mimic the human brain to understand how it is processing things, such as touch and vision” to design artificial neural systems, Sunil Kukreja - Head of Neuromorphic Engineering & Robotics at the National University of Singapore - tells GovInsider.
Neuromorphic chips are then “able to process visual data in very much the same way as the human retina”, he says. The tech can also convey touch experiences such as texture and weight to computers, doing this at a lower power consumption than current systems.
2. Why is it important?
Since neuromorphic architecture is built to mimic the brain’s biological structure and behaviour, it makes systems “fault-tolerant,” Kukreja says.
He compares this to how the human brain works: if someone has a stroke and suffers paralysis of the arm, the brain can reroute these functions, and through physiotherapy, regain functionality.
In the digital world, this means that if a few components fail, the computer system can function by getting around the faulty component.
Some chips can already sustain defect rates as high as 25%. It can also perform “vast computing at very low power”, he says, which is a contrast to ‘Von Neumann’ architecture - which conventional computers are built on.
For such systems, power consumption has increased “as the clock speed and number of transistors have gone up”, he says. Neuromorphic architecture, on the other hand, does not suffer from the same problem, and can use far less power.
This leads to promising use in small and nimble applications like prosthetics and robots, he says.
3. How could it affect government?
There is vast potential for this in the public sector. In healthcare, the tech could allow prosthetics to perform a dynamic range of motions.
“Prosthetic limbs are usually very small in size, so you can’t put very large power sources on these.” But “if you put a small battery that’s able to power visual and tactile perception, then you can perform advanced functions that give amputees a lot of capabilities that they currently do not have”, Kukreja says.
Electronic skin placed on the prosthetic hand, for example, can allow Kukreje’s team to do computing work that can “help understand what object is being handled - the weight and texture”, for instance.
This also means the patient will have better “flip control”, knowing when to exert the appropriate force to ensure that an object doesn’t break, or slip out of hold, he adds.
He believes that smaller robots with such architecture hold promise in reconnaissance work, search and rescue, and mining operations - those that require scouting and sensing the environment.
“Vision requires a tremendous amount of computation”, Kukreje says, so “when there’s something moving in the scene, that’s what we’re really interested in”.
The infrastructure processes the bare minimal “so it gives tremendously less computations, which in turn allow us to compute things at very low power”.
In the future, perhaps prosthetics can function almost like our real limbs, and covert operations will be run by robots. We have the rats to thank.