Professor Hocheon Yoo's Team at Hanyang University

Develops Next-Generation Probabilistic Bit Device

A new semiconductor device technology capable of dramatically improving both processing reliability and power efficiency in "probabilistic computing"—a core hardware for next-generation artificial intelligence (AI) and optimization operations—has been developed by a team of Korean researchers.


On May 29, Hanyang University announced that the research team led by Professor Hocheon Yoo from the Department of Convergence Electronics Engineering at the College of Engineering, in collaboration with a team from Seoul National University, has developed a "light-induced bias-tunable probabilistic bit (LBP-bit)" device. This new device generates random signals using light and allows for the precise control of their probability with minute voltage adjustments.


At Hanyang University College of Engineering, researchers conducting next-generation probabilistic bit device research include Dr. Joohyeong Seo (left), Dr. Taehyun Park, and Professor Hocheon Yoo of the Department of Convergence Electronics Engineering. Hanyang University

At Hanyang University College of Engineering, researchers conducting next-generation probabilistic bit device research include Dr. Joohyeong Seo (left), Dr. Taehyun Park, and Professor Hocheon Yoo of the Department of Convergence Electronics Engineering. Hanyang University

View original image

Recently, in the semiconductor industry, probabilistic computing—which utilizes "probabilistic bits (p-bits)" that probabilistically fluctuate between 0 and 1, as opposed to traditional digital bits fixed at either value—is emerging as a future technology. This is because it enables much faster and more efficient solutions through probabilistic search methods in complex optimization problems or machine learning tasks where there are too many candidate answers for conventional approaches to handle efficiently.


However, previously developed probabilistic bit devices have relied entirely on "electrical signals" for both generating random signals and controlling their probability. As a result, the two processes interfered with each other, limiting the ability to precisely control the probability or maintain a stable state.


To address this issue, the research team devised the "LBP-bit" structure, which completely separates the roles of light and voltage at the physical level. The device is designed so that when light enters its interior, photo-currents in different directions are probabilistically generated depending on the wavelength and intensity of the light, thus obtaining random signals from "light" and adjusting the probability distribution using "voltage." This device demonstrated ultra-precise control, enabling the output probability to be continuously tuned from 0% to 100% even with extremely low voltages at the sub-millivolt (mV) level. In particular, even when voltage is applied, the inherent properties of the random signals remain undisturbed and stably maintained, significantly enhancing reliability.


Professor Yoo emphasized, "This research is a case in which light is utilized not merely as a medium for information delivery, but as a key resource for computation, establishing a hardware foundation for low-power probabilistic computing." He added, "This technology is expected to have broad applications in advanced industries such as AI inference and next-generation semiconductor design."



These research findings were published on May 15 in the international journal Science Advances.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing