Neurological Hardware: IIT researchers develop neurological hardware with built-in memory

Science News: Not all rural and remote areas of the country have internet access; even with scant or limited access, various problems like slow internet speed, power outages, etc., lead to frequent data loss. This, in particular, can seriously hamper students’ online or self-studies using the internet in rural and remote areas. They need a device that can aid their studies in the face of no internet, slow internet, or power outages.

Also, these days, interactive voice assistants like Alexa and Siri are very popular. They have made it very easy to obtain a response to a query or get some tasks executed online. However, they are based on von-Neumann architecture, where the computer processing of the voice data needs support from cloud memory. In other words, there is a separation of processing and memory in these devices. A neuromorphic computing can change this scenario, and while preventing data loss, it can also reduce power consumption.

A research team led by Prof. Pranaba Kishor Mriduli of IIT Delhi and Prof. Devanjan Bhowmick of IIT Bombay has developed a neurological hardware with built-in memory, using magnetic materials. The device is capable of storing data even when the power is off. It functions like a brain-inspired computing system. Its function, similar to synapses in the neurological system, is mainly based on the mechanism of the brain. The researchers have published an article on this in ACS Applied Electronic Materials.

The experimental work, leading to the creation of the device, has mainly been carried out by Ram Singh Yadav, a Ph.D. student at the department of physics,IIT Delhi. Prof. Mriduli and Prof. Bhowmick have jointly supervised the work.

The device consists of an ultrathin cobalt layer, which was fabricated using state-of-the-art facilities at IIT Delhi. The thickness of the coating is in the manometer range, which is 80,000-1.00.000 times thinner compared to human hair. The fabrication needed enormous precision. Hence, it was carried out in an ultra-high vacuum chamber containing almost no air molecules.

The work has been supported by the “Scheme for Transformational and Advanced Research in Sciences (STARS)”, funded by the Ministry of Education, Government of India.

Neuromorphic computing is inspired by the workings of the brain. It uses artificial neural networks that work the way the human brain works. The neuromorphic hardware developed by IIT Delhi and IIT Bombay researchers has built-in memory. Although it functions using Artificial Intelligence (AI), algorithms, and data analysis, it is based on non-von Neumann architecture. Unlike Alexa or Siri, the device does not go to the cloud memory to give response to a query as it has a collocated processing and memory. This means there is no separation of processing and memory in this (neuromorphic) device.

The device is, however, in the initial stages of development and only its prototype has been tested. It will need more time before finally getting released to the open market. The device will “contribute significantly to both the India Semiconductor mission and the Make-in-India mission of government of India,” hopes Prof. Mriduli.