Seoul National University Pioneers Ultra-Low Power Neuromorphic Hardware for Advanced AI

A probe squad from the College of Engineering astatine Seoul National University (SNU), headed by Professor Ho Won Jang, has achieved a breakthrough successful ultra-low powerfulness neuromorphic hardware designed for artificial quality (AI) computation. The team's innovative hardware, which importantly reduces powerfulness consumption, is acceptable to beforehand AI exertion crossed assorted industries.

This probe addresses halfway issues successful semiconductor materials and neuromorphic devices, showing imaginable for large-scale, array-level exertion applications. The findings were precocious published successful Nature Nanotechnology, a starring diary successful the field, marking a large accomplishment for the probe community.

As AI-powered applications similar the Internet of Things (IoT), generative AI, and autonomous vehicles progressively trust connected monolithic information processing, existent silicon-based computing faces challenges specified arsenic precocious vigor demands and processing limitations. This makes the improvement of next-generation neuromorphic hardware—a exertion that emulates the quality brain’s neural processing—a captious measurement forward. Unlike accepted computing, neuromorphic systems usage synapse-like components to execute analyzable computations with vigor ratio and accuracy.

The SNU squad focused connected halide perovskite materials, antecedently notable for their relation successful star cells and LEDs, to make neuromorphic devices with unprecedentedly azygous ion distribution. These precocious materials, designed with hybrid organic-inorganic structures, enabled the squad to execute ultra-linear and symmetric synaptic value control, enhancing computation accuracy and efficiency.

Performance investigating revealed that the instrumentality could grip ample datasets with an mistake borderline beneath 0.08%, achieving precocious precision connected tasks ranging from basal representation designation to analyzable AI inference. Collaborative studies with the University of Southern California further demonstrated that the exertion could run connected ultra-low power, some astatine the idiosyncratic instrumentality level and successful arrays, reinforcing its scalability and imaginable impact.

This innovative neuromorphic hardware offers a promising solution to the escalating vigor demands of AI computation. The exertion is anticipated to person wide applications successful fields similar autonomous driving, aesculapian diagnostics, and AI-driven industries, portion fostering advancements successful AI hardware and semiconductor innovations.

The probe builds connected a erstwhile survey by Dr. Seung Ju Kim and Prof. Ho Won Jang, published successful Materials Today, with patent applications underway successful South Korea and the U.S. Dr. Kim, a distinguished researcher and Seoul National University alumnus, is presently advancing this exertion astatine the University of Southern California, moving alongside American probe labs to make intelligent semiconductors for aerospace applications.

Seoul National University, established successful 1946, is South Korea’s premier nationalist assemblage and a cardinal subordinate successful planetary engineering advancements, with the College of Engineering driving innovation done cutting-edge probe and planetary collaborations.