Scientists create world's first photonic neural network

In the latest bid to push the boundaries of Artificial Intelligence (AI), a team of scientists from Princeton University has developed an artificial neural network that uses light (or photons) instead of the traditional electrons. 

For those who aren't familiar with the concept, a neural network works similar to a human brain in solving problems. Equipped with a large number of neural units, each is connected to a larger collection of other units, making the flow of information far more efficient in dealing with parallel or multitasking challenges as compared to the more linear way conventional computer chips process data.

The Neuromorphic chip is the first implementation of a fragment of silicon with photonic technology--light can carry packets of data much faster than electricity, meaning that the new technology could pave the way for devices capable of ultra-fast data transmission. The system consists of 49 nodes, each of which transmits at a specific wavelength of light, enabling it to concurrently process different streams of information.

What makes neural networks interesting is that it gives machines the ability to do things like identify different languages, differentiate between objects and machine translation--activities that are typically associated with human cognition.

The researchers claim that the new system is capable of mathematical equations 1,960 times faster than an ordinary electron-based central processing unit. With such an increase in processing ability, the possibilities are endless--from technology that can predict the stock markets to a supercomputer that can map out the entire universe and all its cosmic dust.