Light-based Computer Chips: Replacing Electrons with Photonsby Adilin Beatrice January 19, 2021
Scientists are developing light-based computer chips that use photons rather than electrons
Researchers have so far worked on developing electronic chips that mimic the way the human brain processes visual information, combining AI algorithms with the hardware necessary for capturing images. But technology has evolved since. These normal advances in the chips come with an enormous price in the computing resource and energy they consume. To address the issue, scientists are developing light-based computer chips that perform many of the required mathematical calculations using photons rather than electrons.
Deep learning is a major source that helps in many processes including computer chip making. However, the limitations of conventional computer hardware are already hindering progress. In the 1980s, researchers hailed optical computing as the next big revolution in information technology. But it turned out to be unsuccessful as the substance has bulky components like fibre optic cables and lenses that didn’t cope for particularly robust and compact computers. Fortunately, beyond the horizon of practicality, researchers are trying to develop a new generation of chips that would control photons as reliably as today’s chips control electrons. They have identified that the best material to use in the formation of an enhanced chip is light. Photonic chips could become the basis for light-based quantum computers that could break codes and solve certain types of problems beyond the capabilities of any electronic computers. Analytics Insight has listed some of the remarkable initiatives taken so far by companies and institutes to make light-based computer chips.
Lightmatter’s light-based chip
Lightmatter, the photonic computer technology start-up has designed a ‘wafer-scale’ programmable chip interconnect fabric that uses light to boost performance while reducing energy usage. The light-based chip named Passage will allow arrays of heterogeneous chips, including processors, graphics cards, memory, and AI accelerators to communicate with each other. It also operates at an unprecedented speed compared with other existing solutions. Besides, Passage also offers a fully reconfigurable connection topology between chips, dramatically reducing the cost and complexity of building heterogeneous systems.
Jelena Vuckovic’s photonic chip made of diamond
Jelena Vuckovic, a Stanford electrical engineer has created a prototype photonic chip made of diamond. This initiative was to educate the world on how to make a light-based chip from a material nearly as hard as diamond, but far less exotic like silicon carbide. Like most chip-making materials, silicon carbide is a crystal, a group of specific atoms arranged in a consistent lattice. The team made efforts to convert this crystal as a photonic chip. They removed atoms at strategic locations throughout the lattice. Each vacancy in the lattice created a subatomic trap that captured a single electron from one of the surrounding carbon atoms. In order to make the light-based chip work, the researchers sent a stream of photons through the lattice. Whenever a photon struck a trapped electron, the collision between those two particles sent a photon spinning off at a particular energy level or a quantum.
MIT’s Photonic chip that uses light instead of electricity
MIT researchers have developed a novel photonic chip that uses light instead of electricity. The chip consumes relatively little power in the process. The photonic chip could be used to process massive neural networks, millions of times more efficiently than today’s classical computers do. MIT researchers describe the new photonic accelerator as a chip that uses more compact optical components and optical signal-processing techniques to drastically reduce both power consumption and chip area. This allows the chip to scale to neural networks several orders of magnitude larger than its counterparts.
Lightintelligence’s AI accelerator chip
Lightintelligence, a Boston-based optical chip maker start-up is developing a radically new kind of AI accelerator chip that uses light to carry out the core mathematical computations needed for machine learning. Accelerator chips that use light rather than electrons to carry out computations promise to supercharge AI model training and inference. In theory, they could process algorithms at the speed of light, faster than today’s speediest logic-gate circuits. The chip is about the size of a printed circuit board, which packs photonic circuits similar to optical fibres that transmit signals. It requires only limited energy and is less susceptible to change in ambient temperature, electromagnetic fields, and other noise.