Engineers at the University of Pennsylvania have revealed a chip that uses light waves instead of electricity to execute complex mathematical operations critical for AI training
This innovation could enhance processing speeds and significantly cut energy consumption.
Silicon-photonic chip revolutionising AI
Nader Engheta and Firooz Aflatouni have created a silicon-photonic (SiPh) chip. Engheta’s revolutionary research in nanoscale material manipulation and Aflatouni’s expertise in nanoscale silicon devices could revolutionise computing.
The chip can perform vector-matrix multiplication, a fundamental mathematical operation key in the development and functionality of neural networks, the backbone of AI systems.
What makes the SiPh chip so special?
The design of the SiPh chip includes localised variations in silicon thickness, as thin as 150 nanometers, strategically positioned to control the propagation of light.
These variations cause specific scattering patterns, enabling the chip to execute mathematical calculations at the speed of light. This approach avoids the limitations inherent in conventional chips, dating back to the 1960s, directing in a new era of computational efficiency.
AI innovation
Aflatouni emphasises the chip’s immediate relevance for commercial use, highlighting its compatibility with existing graphics processing units (GPUs), which have witnessed growing demand with interest in AI innovation.
By integrating the Silicon Photonics platform as an augmentation, the chip uses training and classification processes, unlocking unusual levels of computational performance.
Beyond its remarkable speed and energy efficiency, the Engheta-Aflatouni chip offers inherent privacy advantages. The nature of calculations eliminates the necessity to store sensitive data in a computer’s working memory, generating future systems reinforced with this technology virtually immune to hacking attempts.
The future of the SiPh chip
Supported in part by grants from the U.S. Air Force Office of Scientific Research’s Multidisciplinary University Research Initiative (MURI) and the U.S. Office of Naval Research, the collaborative effort at the University of Pennsylvania emphasises the transformative potential of interdisciplinary research in driving technological advancement.
Co-authors Vahid Nikkhah, Ali Pirmoradi, Farshid Ashtiani, and Brian Edwards contributed to developing this groundbreaking chip.
The unveiling of the SiPh chip indicates a shift in computing capabilities. With its remarkable speed, energy efficiency, and inherent security features, this breakthrough will catalyse advancements across many industries.
