A research team from the College of Engineering at Seoul National University (SNU), led by Professor Ho Won Jang, has made a significant advancement in ultra-low power neuromorphic hardware for artificial intelligence (AI) computing. This innovative hardware greatly reduces energy use, promising to enhance AI technology in many industries.
This research tackles key challenges in semiconductor materials and neuromorphic devices, indicating its potential for large-scale technology applications. The results were recently published in Nature Nanotechnology, a top journal in the field, marking a significant accomplishment for the research team.
As AI applications like the Internet of Things (IoT), generative AI, and self-driving cars increasingly depend on heavy data processing, traditional silicon-based computing faces issues like high energy demands and processing limits.
Developing next-generation neuromorphic hardware technology that mimics how the human brain processes information is crucial. Unlike conventional computing, neuromorphic systems use components similar to synapses to perform complex tasks more efficiently and accurately.
The SNU team explored halide perovskite materials, which are known for their use in solar cells and LEDs, to create neuromorphic devices with an outstandingly even distribution of ions. These advanced materials, featuring hybrid organic-inorganic structures, allowed the researchers to achieve precise synaptic weight control, improving computation accuracy and efficiency.
Testing showed that the device could manage large datasets with an error rate below 0.08%, achieving high accuracy in tasks ranging from simple image recognition to complex AI reasoning.
Collaborations with the University of Southern California also showed that the technology can work on ultra-low power at both individual device levels and in larger arrays, highlighting its scalability and potential impact.
This cutting-edge neuromorphic hardware presents a promising solution to the growing energy needs of AI computing. It is expected to have wide-ranging applications in areas such as self-driving technology, medical diagnostics, and AI-driven industries while encouraging improvements in AI hardware and semiconductor technology.
The research builds on previous work by Dr. Seung Ju Kim and Professor Ho Won Jang, published in Materials Today, with patent applications currently in progress in South Korea and the U.S. Dr. Kim, a prominent researcher and SNU alumnus, is advancing this technology at the University of Southern California, collaborating with American labs to develop smart semiconductors for aerospace applications.
Seoul National University, founded in 1946, is South Korea’s top national university and a leader in global engineering advancements, with the College of Engineering promoting innovation through advanced research and international partnerships.
