Beyond developing components similar to the brain, this system is both reliable and practical,
(Daejeon=News1) Reporter Kim Tae-jin = A Korean research team has developed a neuromorphic semiconductor chip that can learn and self-correct.
The existing computer systems were not efficient for processing complex data like artificial intelligence because their data processing and storage units were separate.
The Korea Advanced Institute of Science and Technology (KAIST) announced on the 17th that a research team led by Professors Choi Shin-hyun and Yoon Young-kyu from the Department of Electrical and Electronic Engineering has developed a next-generation neuromorphic semiconductor-based ultra-small computing chip capable of self-learning and error correction.
This computing chip can learn and fix errors that arise from non-ideal characteristics, which were difficult to resolve in existing neuromorphic devices.
For example, when processing video streams, the chip learns to automatically separate moving objects from the background and performs this task better over time.
This self-learning ability has proven to achieve accuracy comparable to ideal computer simulations in real-time video processing. It is a system that is completed with both reliability and practicality, beyond developing components similar to the brain.
The research team has proposed an innovative solution that overcomes the limitations of existing technologies by developing the world’s first memristor-based integrated system capable of adapting to immediate environmental changes.
At the core of this innovation is a next-generation semiconductor device called a memristor. Its variable resistance properties can serve as a substitute for synapses in neural networks, allowing data storage and computation to occur simultaneously, much like our brain cells.
The research team designed a highly reliable memristor with precisely controllable resistance changes and developed an efficient system that excludes complex calibration processes through self-learning. This experimental verification of the next-generation neuromorphic semiconductor-based integrated system that supports real-time learning and inference holds significant implications for commercialization.
This technology is expected to revolutionize the way artificial intelligence is used in everyday devices by enabling AI tasks to be processed locally, rather than relying on remote cloud servers, thereby increasing speed, privacy protection, and energy efficiency.
The KAIST researchers leading the technology development, Jung Hak-cheon and Han Seung-jae, said, “This system is like a smart workspace where everything is within reach, instead of moving back and forth between a desk and a filing cabinet. It is similar to our brain’s very efficient way of processing information where everything is handled in one place.”
This research, supported by the National Research Foundation of Korea’s Next-Generation Intelligent Semiconductor Technology Development Project, Excellent Young Researcher Project, PIM AI Semiconductor Core Technology Development Project, and the Institute for Information & Communications Technology Promotion’s Korea Electronics and Telecommunications Research Institute R&D Support Project, was published online in the international journal ‘Nature Electronics’ on January 8th.