A team of international researchers has developed a neuromorphic hardware platform that can boost the performance of AI-driven computing applications. It also has the potential to give rise to better and more efficient energy systems.
Neuromorphic devices are electronic systems designed to mimic the architecture and functionality of the human brain and nervous system. Such devices are part of neuromorphic computing , an advanced field focusing on the development of brain-like computers.
The neuromorphic device developed by the researchers is quite special because it can be used to study, control, and manipulate molecules that make up a material. It would make it possible to introduce changes in materials at a fundamental level.
“This outside-the-box solution could have huge benefits for all computing applications, from energy-hungry data centers to memory-intensive digital maps and online gaming,” Damien Thompson, one of the researchers and a professor of molecular modeling at the University of Limerick in Ireland (UL), said .
A device that offers high computing resolution
The platform proposed by the researchers in their study is an analog molecular memristor. It is a type of programmable neuromorphic device that works like the human brain ’s memory system. However, it’s not exactly a brain-like computer but a component required to make such computing systems.
The memristor is made up of molecules that can change their electrical properties based on the amount of charge that passes through them.
The device’s design “Draws inspiration from the human brain, using the natural wiggling and jiggling of atoms to process and store information. As the molecules pivot and bounce around their crystal lattice, they create a multitude of individual memory states,” Thompson stated.
However, this isn’t the first neuromorphic platform. Scientists have developed such devices in the past as well, but those designs had low computing resolution, and therefore, they were only capable of performing “low-accuracy operations.”
“The team’s reconceptualization of the underlying computing architecture achieves the required high resolution, performing resource-intensive workloads with unprecedented energy efficiency of 4.1 tera-operations per second per watt (TOPS/W),” the researchers note.
So unlike previously proposed neuromorphic devices, the new memristor can perform high-level tasks such as neural network training, natural language processing, and signal processing.
“We created the most accurate, 14-bit, fully functional neuromorphic accelerator integrated into a circuit board that can handle signal processing, AI, and machine learning workloads such as artificial neural networks, auto-encoders, and generative adversarial networks,” said Sreetosh Goswami, one of the study authors and an expert in brain-inspired computing at the Indian Institute of Science (IISC).
This memristor can trace and manipulate molecules
The study authors claim that using the proposed neuromorphic platform, they can track the movement of molecules inside a device or a material and match each moment to a specific electrical state.
Once they identify the electrical state connected to a molecule, they can introduce desired changes in the molecule simply by passing a different voltage. This will allow scientists to manipulate materials and integrate them with electrical systems.
For instance, it can lead to the development of computing systems integrated into packaging, textiles, and other materials. Imagine an AI-chip-powered T-shirt whose color you can change as per your mood.
This may sound like science fiction for now, but the proposed neuromorphic device has the potential to turn such things into reality. It can give rise to an entirely new generation of computing applications.
“The ultimate aim is to replace what we now think of as computers with high-performance ‘everyware’ based on energy efficient and eco-friendly materials,” Thompson said.
The study is published in the journal Nature .
Local News
Comments / 0