AI computing has become very fast due to AI chip technology enhancement over the years. But now, It’s time to say lightning fast. The clever Engineers from “The University of Pennsylvania School of Engineering and Applied Science” might have found a big change to the AI chip that could make technology way better.
AI Chip Uses Light
The chips we use today are pretty much the same as the ones from the 1960s when computers first got big. But with light, we might be able to make computers even better. The clever Engineers at Penn have come up with a new chip that doesn’t use regular electricity for crunching those complex AI math problems. The chip uses light waves instead and has the potential to make computers way faster while also saving a lot of energy. It’s like a win-win for everyone!
The design of the silicon-photonic (SiPh) chip represents a significant milestone as it integrates the pioneering research of Benjamin Franklin Medal Laureate and H. Nedwill Ramsey Professor Nader Engheta. Professor Engheta’s work focuses on manipulating materials at the nanoscale to conduct mathematical computations using light, known for its unparalleled speed in communication. This innovative chip merges Engheta’s research with the SiPh platform, which utilizes silicon, a cost-effective and abundant element widely employed in the mass production of computer chips.
We Wanted To
Their objective was to establish a platform capable of executing vector-matrix multiplication, a core mathematical operation crucial for the development and operation of neural networks— the backbone of modern AI tools. Rather than employing a uniform silicon wafer, they engineered specific regions of the silicon to be thinner, approximately 150 nanometers in thickness. These deliberate variations in thickness enable precise control over light propagation through the chip. By strategically scattering light in predetermined patterns, the chip achieves rapid mathematical computations at the speed of light.
No Memory, No Hack
Along with being faster and using less energy, the chip offers privacy benefits. Since multiple computations can be processed at the same time, there’s no need to store sensitive information in a computer’s memory. This makes a future computer using this technology nearly impossible to hack. Simply your information stays safe. Why? if there’s no memory, how they can hack into it?
AI Chip Ready For Commercial?
Based on the constraints imposed by the commercial foundry that made the chips, the developers explain that this design is already set for commercial use. It could even be adjusted for use in graphics processing units (GPUs), which are in high demand due to the growing interest in creating new AI systems. Additionally, they can incorporate the Silicon Photonics platform as an extra add-on, boosting the speed of training and classification tasks.
