A research team from South Korea has developed a high-performance AI chip that is suitable for AI computations that will address challenges like higher power consumption and significant delays in calculating results.
(Photo : Chris McGrath/Getty Images)
ISTANBUL, TURKEY - MAY 06: A tablet is seen in the center of a high tech art installation at Salt Galata on May 6, 2017 in Istanbul, Turkey. The "Archive Dreaming" installation by artist Refik Anadol uses artificial intelligence to visualize nearly 2 million historical Ottoman documents and photographs from the SALT Research Archive.
High-Performance AI Chip
As artificial intelligence rises and was recognized through its ability to perform several tasks, systems of traditional digital computer separates the storage and computation of information, which leads to much-increased power usage and lags in computations.
Due to this, researchers from Pohang University of Science & Technology, also known as POSTECH, developed a solution that will address these challenges. Interesting Engineering reported that through chip technologies, researchers were able to produce a high-performance device that offers high performance and power efficiency.
The material used to build this device is indium gallium zinc oxide (IGZO), which is an oxide chip that is commonly used in OLED displays. POSTECH researchers stated that previous systems were only limited to meeting all AI requirements like linear and symmetric programming and uniformity to improve accuracy.
Researchers then used IGZO as the key material for these calculations, which provided uniformity, durability, and computing accuracy. This consists of four atoms in a fixed ratio of indium, gallium, zinc, and oxygen, with excellent electron mobility and leakage current properties.
Also Read: AI Vision is Less Superior than Human Vision New Study Finds
They were able to develop a novel synapse device that comes with two transistors connected through a storage node. The AI chip was also able to meet the performance metrics required for a much-improved performance by controlling the node's charging and discharging rate.
As per the team, "Possibility of utilizing the ultra-thin film insulators inside the transistors to control the current, making them suitable for large-scale AI." The output current of synaptic devices will be minimized as technologies were included in a large-scale AI system.
Testing the Device
EurekAlert reported that the researchers used the newly developed device to train and classify handwritten data and achieved a high accuracy of over 98%. With this percentage, this technology may potentially be used in the future in high-accuracy systems related to AI.
This team was led by Professor Yoonyoung Chung, Professor Seyoung Kim and Ph.D. candidate Seongmin Park from the Department of Electrical Engineering. Chung stated, "The significance of my research team's achievement is that we overcame the limitations of conventional AI semiconductor technologies that focused solely on material development."
The study was then published on the inside back cover of Advanced Electronic Materials, with support from the Next-Generation Intelligent Semiconductor Technology Development Program through the National Research Foundation. This foundation was funded by the Ministry of Science and ICT of South Korea.
Related Article: AI-Generated Works May Now Qualify for Copyright Protection
