A team of researchers at the University of Pennsylvania has made a significant advance by developing a silicon-photonic (SiPh) computer chip that runs on light rather than electricity, according to Interesting Engineering.
This breakthrough has the potential to improve data transfer speed while also reducing electricity usage, which will be very useful for developing artificial intelligence (AI).
Led by Professor Nader Enghata, the team created the SiPh device to do mathematical computations using light, taking advantage of its unprecedented speed for data transmission.
Scalability is a significant advantage when using readily available silicon. The chip concentrates on vector-matrix multiplications, which are crucial mathematical computations in the development and operation of neural networks required for modern AI models.
Crucially, the researchers addressed data privacy problems related to concurrent computing.
Because the SiPh device can perform many computations at once, there is no need to retain information in working memory during the operation, offering an extra layer of security against potential intrusions.
The SiPh device is ready for immediate use, with potential uses including replacing graphics processing units (GPUs) in training and classifying AI models.
Firooz Aflatouni, an associate professor at the University, believes that the SiPh platform might effortlessly integrate into existing AI infrastructure, resulting in faster computations and lower electricity consumption, representing a substantial advancement in computing technology.
