This post is also available in:
עברית (Hebrew)
Neuromorphic devices are electronic systems that mimic the architecture and functionality of the human brain. Now, a team of researchers has unveiled a groundbreaking neuromorphic hardware platform designed to enhance AI-driven computing applications, representing a significant advancement in the field of neuromorphic computing.
The newly developed neuromorphic device is particularly noteworthy as it enables the study and modification of molecules within materials, paving the way for fundamental changes at the molecular level. At the core of this innovation is an analog molecular memristor, a programmable neuromorphic device that emulates the brain’s memory system. While not a fully realized brain-like computer, the memristor serves as a vital component for future neuromorphic systems. It comprises molecules that alter their electrical properties based on the charge passing through them. This design draws inspiration from the human brain, using the natural movements of atoms to process and store information, explained Damien Thompson, a researcher and professor of molecular modeling at the University of Limerick in Ireland. “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,” he said.
Previous neuromorphic platforms suffered from low computing resolution, limiting their capabilities to low-accuracy operations, according to Interesting Engineering. However, this new design achieves high resolution, enabling workloads that require intensive resources with remarkable energy efficiency of 4.1 tera-operations per second per watt (TOPS/W). The new memristor can perform advanced tasks such as neural network training, natural language processing, and signal processing. “By precisely controlling the vast array of available molecular kinetic states, 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 Professor Sreetosh Goswami, a study author and an expert in brain-inspired computing at the Indian Institute of Science (IISC).
The platform allows for tracking the movement of molecules and linking each movement to specific electrical states. By identifying these states, researchers can introduce changes in the molecules simply by adjusting voltage, and facilitating material manipulation integrated with electrical systems.
This technology could lead to innovative applications, such as AI-enabled textiles that change color based on mood. While this may seem futuristic, the neuromorphic device holds the potential to revolutionize computing applications.
The study is published in the journal Nature.
