Researchers from the University of Texas (UT) are developing energy-efficient technology for artificial intelligence in collaboration with the Taiwan Semiconductor Manufacturing Company (TSMC). Their focus is on spin-orbit torque magnetic random access memory (SOT-MRAM), which consumes less energy than traditional memory and operates without needing data centers.
The project aims to reduce the resource demands of AI without sacrificing model accuracy, according to project lead Jean Anne Incorvia. She highlights the need for more efficient methods in AI development, considering the sustainability of energy and water consumption by data centers.
The collaboration with TSMC is expected to continue even after the initial findings were published. Ph.D. student Vivian Rogers, who is involved in the study, expressed enthusiasm about working with a leading semiconductor company and the opportunities provided by UT to explore advanced material sciences and computer engineering.
The UT administration has shown support for enhancing resources in materials science, creating a conducive environment for innovative research.
