ChatGPT’s influence is not limited to the realm of education, and it is making notable transformations in other domains. The AI language model is acknowledged for its versatility in carrying out diverse functions such as generating written content, translating languages, coding, and more, all facilitated through interactive exchanges based on queries and responses.
The functioning of AI systems relies on deep learning, which necessitates extensive training to minimize errors. This, in turn, leads to frequent data transfers between memory and processors. However, conventional digital computer systems, built on the von Neumann architecture, have a separation of storage and computation of data, resulting in increased power consumption and significant delays in AI computations. To address this challenge, researchers have come up with semiconductor technologies tailored to AI applications.
Recently, a team of researchers at POSTECH, headed by Professor Yoonyoung Chung (Department of Electrical Engineering and Semiconductor Engineering), Professor Seyoung Kim (Department of Materials Science and Engineering, Department of Semiconductor Engineering), and Ph.D. candidate Seongmin Park (Department of Electrical Engineering), have developed a high-performing AI semiconductor device that uses indium gallium zinc oxide (IGZO), an oxide semiconductor commonly used in OLED displays. The new device has proven to be excellent in terms of performance and power efficiency.
For AI operations to be efficient, it is essential that computations take place within the memory that stores the data. However, previous semiconductor technologies designed for AI computations have not fulfilled all the necessary requirements to improve accuracy, such as uniformity, linearity, and symmetry.
To address these limitations, the research team decided to focus on IGZO as a key material for mass-produced AI computations. IGZO has a fixed ratio of four atoms, namely indium, gallium, zinc, and oxygen, which provides uniformity, durability, and computing accuracy. Its electron mobility and leakage current properties make it an ideal backplane for OLED displays.
By utilizing IGZO, the researchers created a novel synapse device that consists of two transistors linked through a storage node. The precise control of this node’s charging and discharging speed enables the AI semiconductor to meet the diverse performance metrics required for high-level performance.
To apply synaptic devices in a large-scale AI system, it is necessary to minimize the output current of these devices. To achieve this, the researchers investigated the possibility of utilizing ultra-thin film insulators inside transistors to control the current. This approach was found to be effective, making it suitable for large-scale AI systems. The newly developed synaptic device was tested by the researchers in training and classifying handwritten data. The results showed a remarkable accuracy of over 98%, confirming its potential for application in high-accuracy AI systems in the future.
Professor Chung explained, “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. To do this, we utilized materials already in mass production. Furthermore, Linear and symmetrical programming characteristics were obtained through a new structure using two transistors as one synaptic device. Thus, our successful development and application of this new AI semiconductor technology show great potential to improve the efficiency and accuracy of AI.”
You might also be interested in reading, Researchers Develop Colorful Plant-Based Film That Cools Buildings Without Electricity
