KAIST has developed an integrated system based on memristors that processes information in a manner similar to the human brain. This is significant because it addresses the inefficiency of traditional computer systems where data processing units and storage units are separated, making it difficult to handle complex data.
The developed system is expected to be applied in various fields, from security cameras that can instantly recognize suspicious activities without relying on remote cloud servers to medical devices that can analyze health data in real time.
(From left) Professor Choi Sin-hyun, Professor Yoon Young-kyu, Integrated MS-PhD student Jung Hak-cheon, Integrated MS-PhD student Han Seung-jae, Department of Electrical Engineering. Provided by KAIST
On the 17th, KAIST announced that a joint research team led by Professors Choi Sae-Hyun and Yoon Young-Kyu from the Department of Electrical Engineering and Computer Science developed a next-generation neuromorphic semiconductor-based ultra-compact computing chip capable of self-learning and even correcting errors during the learning process.
This computing chip has the advantage of being able to self-learn and correct errors caused by non-ideal characteristics that were difficult to resolve in existing neuromorphic devices.
The self-learning capability of the computing chip developed by the joint research team achieved and demonstrated accuracy comparable to ideal computer simulations in real-time image processing.
Through this research, the joint research team developed the world’s first memristor-based integrated system that can immediately adapt to environmental changes, presenting a solution that overcomes the limitations of existing technologies.
A memristor, a portmanteau of memory and resistor, is a next-generation electrical device whose resistance value is determined by the amount and direction of charge that has previously flowed between its two terminals.
At the core of this solution is the memristor, a next-generation semiconductor device. Its variable resistance characteristic can replace the synaptic role in neural networks, enabling simultaneous data storage and computation similar to brain cells when utilized.
Based on this, the joint research team designed a memristor with high reliability capable of precisely controlling resistance changes and improved system efficiency by eliminating complex calibration processes through self-learning.
The joint research team emphasized that this study is significant in experimentally verifying the commercialization potential of next-generation neuromorphic semiconductor-based integrated systems that support real-time learning and inference.
In the future, this technology is expected to revolutionize the way artificial intelligence (AI) is used in everyday devices, allowing AI tasks to be processed locally without relying on remote cloud servers.
Researchers Jung Hak-Cheon and Han Seung-Jae from KAIST explained, “The system developed by the joint research team acts as a ‘smart workspace’ where everything needed is within reach, instead of moving between desks and filing cabinets. This is similar to the brain’s highly efficient information processing method where everything is handled in one place.”
Meanwhile, this research was conducted with support from the National Research Foundation of Korea’s Next-Generation Intelligent Semiconductor Technology Development Project, Excellent Young Researchers Project, PIM AI Semiconductor Core Technology Development Project, and the Korea Electronics and Telecommunications Research Institute R&D Support Project of the Institute for Information & Communications Technology Planning & Evaluation.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
