New Chip Opens Door to AI Computing at Light Speed
A team of engineers from the University of Pennsylvania has developed a novel computer chip that uses light waves instead of electricity to perform complex mathematical operations. The chip could drastically improve the speed and efficiency of artificial intelligence (AI) models, as well as enhance their privacy and security.
The chip is based on the silicon-photonic (SiPh) platform, which uses silicon, the cheap and abundant element that is widely used to make computer chips. However, unlike conventional chips that rely on electric currents to transfer data, the SiPh chip manipulates light waves to perform computations at the speed of light.
The design of the chip is inspired by the pioneering research of Professor Nader Engheta, who has shown how materials can be engineered at the nanoscale to perform mathematical functions using light. By varying the height of the silicon in specific regions, the researchers were able to control how light propagates inside the chip, creating patterns that correspond to vector-matrix multiplications, a core operation in the development and function of neural networks, the computer architecture that powers today’s AI tools.
The SiPh chip has several advantages over conventional chips. First, it can process large amounts of data much faster, as light travels faster than electricity. Second, it consumes much less energy, as light does not generate heat as electricity does. Third, it offers more privacy and security, as many computations can happen simultaneously, without the need to store sensitive information in the computer’s memory, making it virtually unhackable.
The researchers collaborated with a commercial foundry to fabricate their SiPh chips, which means that the technology is ready for deployment and can be easily integrated with existing computing systems. The researchers hope that their chip will enable new applications of AI, such as faster and more accurate image recognition, natural language processing, and self-driving cars.
Leave a Reply