In a groundbreaking development, the scientific team from the Department of Informatics at the Aristotle University of Thessaloniki (AUTH) has unveiled one of the world’s fastest AI processors. Leveraging photonic neurons that operate with light instead of electricity, this cutting-edge processor marks a significant milestone in the quest for alternative computing systems.
Headed by postdoctoral researcher Miltiadis Moralis-Peios, the research team’s revolutionary processor boasts a novel photonic neuron architecture, enabling the execution of algebraic operations at unprecedented speeds. The recent experimental demonstration showcased the processor’s remarkable capability, achieving operational speeds of up to 50 GHz—a staggering 25 times faster than contemporary GPU processors. What’s more, this groundbreaking technology offers the simultaneous advantage of 10 times lower power consumption, positioning it as a potential game-changer in the world of computing.
Speaking about the project, Moralis emphasized the necessity of exploring computing alternatives due to the physical limitations posed by traditional microelectronics technologies. The utilization of light, being an electromagnetic wave, presents a viable solution, given its superior speed and the minimal disruption it causes to integrated optical circuits compared to electricity. The team’s pioneering use of light in implementing multiplication and addition operations has catapulted processing speeds to levels far surpassing those of contemporary computing giants such as NVIDIA’s GPU and Google’s TPU.
The integration of photonic technology in information processing, particularly in the domain of neural networks, offers a novel approach to achieving faster and more efficient data processing. By leveraging data in the form of light and advanced photonic integration technologies, the team aims to simulate the intricate operations of the human brain, revolutionizing the capabilities of computing systems for enhanced information processing.
At the recent Optical Fiber Communication Conference (OFC) in San Diego, USA, the Department of Informatics’ research findings garnered significant attention. Led by the dedicated efforts of PhD candidates Apostolos Tsakiridis, Georgios Giamougiannis, and Christos Pappas, under the guidance of Professor Nikolaos Pleros, the team’s exceptional contributions were recognized as they secured a spot as finalists for the esteemed international OFC Corning Student Award.
This recent stride by the Aristotle University of Thessaloniki not only underscores the university’s commitment to pioneering research but also signifies a potential paradigm shift in the landscape of AI processors and their applications, promising a future where computing capabilities are no longer confined by traditional constraints.