New computer chip enables AI computing at the speed of light

Engineers from the University of Pennsylvania have introduced a novel computer chip that could redefine the landscape of artificial intelligence (AI) development.

This innovative silicon-photonic (SiPh) chip leverages light waves to perform complex mathematical computations, a critical process in training AI, with unprecedented speed and efficiency.

This development not only promises to drastically boost computer processing speeds but also significantly reduce energy consumption.

Bridging light and silicon

The genesis of this chip lies at the intersection of Benjamin Franklin Medal Laureate and H. Nedwill Ramsey Professor Nader Engheta‘s trailblazing research on nano-material manipulation and the SiPh platform’s capacity to use silicon for mass chip production.

Silicon, a widely available and affordable element, has been a staple in manufacturing computer chips. This collaboration marks the first time these two pioneering fields have converged to exploit light — the fastest communication medium — for computational purposes.

This approach aims to transcend the limitations of contemporary chips, which, despite decades of technological advancements, still operate under principles dating back to the 1960s.

The integration of light and matter opens new avenues for surpassing the current capabilities of computer chips, heralding a new era in computing technology.

The collaborative effort showcases the synergy between Engheta’s team and that of Firooz Aflatouni, Associate Professor in Electrical and Systems Engineering.

Aflatouni, known for his work on nanoscale silicon devices, emphasized the strategic partnership, stating, “We decided to join forces,” to harness their combined expertise in advancing this technology.

Revolutionizing neural networks

Central to their innovation is the chip’s ability to perform vector-matrix multiplication, a fundamental operation in neural network development.

Neural networks underpin the AI tools that are increasingly integral to various sectors. By ingeniously adjusting the thickness of the silicon in targeted areas — a method described by Engheta as varying the silicon to as thin as 150 nanometers — this design manipulates light’s propagation.

These modifications allow light to scatter in precise patterns, enabling the chip to execute calculations at the speed of light without the need for additional materials.

Aflatouni highlighted the chip’s readiness for commercial applications, a direct outcome of the design’s compatibility with existing manufacturing processes.

This adaptability is particularly relevant for graphics processing units (GPUs), which have seen a surge in demand due to the AI development boom.

“They can adopt the Silicon Photonics platform as an add-on,” Aflatouni explained, suggesting a significant potential to expedite AI training and classification processes.

AI computer chip and the unhackable future

Beyond its impressive speed and energy efficiency, the chip offers substantial privacy benefits. The capacity for simultaneous computations eliminates the need for storing sensitive data in a computer’s working memory.

This feature, as Aflatouni points out, could make future computers virtually impervious to hacking: “No one can hack into a non-existing memory to access your information.”

The development of the SiPh chip by Engheta, Aflatouni, and their teams represents a significant leap forward in computing technology.

By harnessing the power of light for computational purposes, they are paving the way for a new generation of AI development, characterized by unparalleled speed, efficiency, and security.

This fascinating innovation holds promise for enhancing the capabilities of existing technologies while exploring new realms of artificial intelligence, potentially transforming our approach to data processing and machine learning.

The full study was published in the journal Nature Photonics.

—–

Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates.

—–

Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

—–