News & Updates

Scientists at the Indian Institute of Science (IISc), Bengaluru, have reported a significant breakthrough in neuromorphic computing, which could transform the landscape of artificial intelligence (AI) by enabling advanced processing on personal devices rather than relying on large data centers. This innovation, detailed in the journal *Nature*, involves the development of a Memristor using a metal-organic film, allowing it to mimic the brain's neural processing more efficiently than traditional silicon-based technologies.

Led by Prof. Sreetosh Goswami, the team integrated this Memristor with a conventional silicon processor to enhance energy efficiency and processing speed dramatically. Their device can store and process data in over 16,500 states simultaneously, significantly reducing the computational steps required for tasks like vector-matrix multiplication, which is crucial for AI algorithms. For example, they demonstrated processing capabilities that are 220 times more efficient than a standard NVIDIA GPU, achieving 4.1 tera-operations per second per watt.

The research addresses critical limitations of current digital computing, particularly the inefficiencies of the conventional Von Neumann architecture, which separates processing and memory, leading to energy and speed bottlenecks. The IISc team's approach not only overcomes these challenges but also positions India to compete in the global AI arena by potentially allowing complex AI tasks to be performed on everyday devices like laptops and smartphones.

Future plans include scaling the technology to larger arrays and developing a System-on-Chip solution, with aspirations to commercialize the innovation within three years, marking a significant step forward in AI technology and its accessibility

Source :