A research team from The Korea Advanced Institute of Science and Technology created a next-generation neuromorphic semiconductor-based ultra-compact computing chip capable of self-learning and error correction. The study was published in the international academic journal Nature Electronics.
The research team of the School of Electrical Engineering posed by the newly developed processor. (From center to the right) Professor Young-Gyu Yoon, Integrated Master's and Doctoral Program Students Seungjae Han and Hakcheon Jeong and Professor Shinhyun Choi Image Credit: The Korea Advanced Institute of Science and Technology (KAIST)
Computer systems are ineffective at processing complex data, such as artificial intelligence, because they have distinct data processing and storage components. A KAIST research team created an integrated system based on memristors that mimic how the brain processes information. It can now be used in various devices, such as smart security cameras, which enable them to identify suspicious activity instantly without depending on distant cloud servers, and medical devices, which enable real-time health data analysis.
The research was carried out by the joint research team of Professor Shinhyun Choi and Professor Young-Gyu Yoon of the School of Electrical Engineering at KAIST (President Kwang Hyung Lee).
This computing chip is unique because it can learn and fix mistakes caused by less-than-ideal features that were challenging to address in previous neuromorphic devices. For instance, when processing a video stream, the chip learns to automatically distinguish moving objects from the background and improves over time.
Real-time image processing has demonstrated this self-learning capability by attaining accuracy on par with optimal computer simulations. Beyond the creation of brain-like components, the research team's primary accomplishment is completing a dependable and useful system.
The research team has created the first memristor-based integrated system in history that can instantly adjust to environmental changes. The team has also offered a creative solution that gets around the drawbacks of current technology.
The core of this innovation is a memristor, a next-generation semiconductor device. This device's variable resistance properties can replace synapses in neural networks, allowing data storage and computation to occur simultaneously, just like in human brain cells.
The research team created a highly dependable memristor that can accurately regulate resistance changes and an effective self-learning system to eliminate intricate compensation processes. The study is noteworthy because it experimentally confirmed that a next-generation neuromorphic semiconductor-based integrated system that facilitates real-time learning and inference could be commercialized.
With the ability to process AI tasks locally rather than relying on distant cloud servers, this technology will completely transform how artificial intelligence is used in commonplace devices. This will make AI tasks faster, more energy-efficient, and more private.
“This system is like a smart workspace where everything is within arm’s reach instead of having to go back and forth between desks and file cabinets. This is similar to the way our brain processes information, where everything is processed efficiently at once at one spot,” elaborated KAIST researchers Hakcheon Jeong and Seungjae Han, who led the development of this technology.
Jeong and Han have enrolled in the Integrated Master's and Doctoral Program.
The National Research Foundation of Korea's Excellent New Researcher Project, PIM AI Semiconductor Core Technology Development Project, Next-Generation Intelligent Semiconductor Technology Development Project, and Electronics and Telecommunications Research Institute Research and Development Support Project all funded the study.
Journal Reference:
Jeong, H., et al. (2025) Self-supervised video processing with self-calibration on an analog computing platform based on a selector-less memristor array. Nature Electronics. doi.org/10.1038/s41928-024-01318-6