MIT Researchers have come up with a new approach of using the light waves instead of electricity. This approach can vastly improve the speed and efficiency of the system that improves heavy computation like Deep learning.

Deep learning is based on the artificial neural network that tends to mimic the way a human brain works. There has been a number of development in the field of deep learning in the recent times like voice and face recognition system. Deep learning methods include lots of computation and have to deal with the massive amount of data. Yes, the computers have developed a lot in the recent times that can handle heavy computation. But, as we will move forward the need for faster computation will increase. This demands for the change in the basic electrical units of the computer system. Researchers from MIT have just done that.

There were many claims about the optics based computer in past but none of them was efficient or practical. But the MIT researchers claim that their system can be applied to the Deep learning technologies and many more fields.

Traditional computer architecture is not efficient in handling the neural network task. The neural networks involve frequent calculation like the repeated matrix multiplication and traditional computers are not efficient to handle that. But the researchers claim that their system can do the repeated matrix multiplications with zero energy and instantly.

The system uses multiple light waves which are directed in such a way that their waves interact with each other. This interaction produces interference patterns that convey the result of the indeed operation. The resulting device is called as ‘programmable nanophotonic processor’. These processors can carry out the calculation performed in AI algorithms using less than one thousandth as much energy per operation compared to the traditional ones.

The system uses an array of the waveguide that is interconnected and can be modified as per the need. The processor guides light through a series of coupled photonic waveguides. The researchers added that it can be used in many other sectors like signal processing for data transmission. They also said that this system can do the processing directly in the analog domain.

Currently, the researchers are trying to scale the system so that it can be used in self-driving cars and drones.

Image Source: Phys.org