Based on the study published in the Science Advanced Journal, a team of researchers has finally developed a new optical computer for AI and ML. Not only it is much more energy-efficient and faster, but it also addresses the ‘noise’ that is predominantly present in optical computing. 

This AI computing system is not only able to mitigate the noise problem, but it also uses some of the interferences as input while enhancing the creative output of the artificial neural network currently present within the system. 

According to the lead author of the report, Mr. Changming Wu, “this optical computer is many folds faster than any conventional digital computer out there in the market. Taking random input from the generated noise, the optical computer is also able to create various outputs.”

Read: What is Artificial Intelligence (AI)?

Addressing Noises in Optical Computing 

The noises that we often come across in optical computing come from stray light particles. These photons originate from the laser operations within the computer along with the thermal radiation occurring in the background. 

In order to mitigate the noise, researchers have actually connected the core of the optical computer to a Machine Learning network known as the Generative Adversarial Network. In order to test out, several random noises were fed right into the GAN. 

By collaborating with Duke University, GAN can now mitigate the downsides of noise created from the optical computer using its robust and error-free algorithm. Further including the network can make the best use of the noise as inputs while generating output instances. 

Future Scope of GAN

An experiment was tried on GAN with its ability to write distinct numbers. Along the way, not only it could write numbers from 0 to 10 using computer simulation, but at the same time developed its own writing style

The next step towards its success is building the device at large using the current semiconductor manufacturing technology. So, the team plans of using wafer-scale technology instead of devising it in a lab. 

Not to mention, it can further improve its performance as well as allow the research team to study more complex tasks, including creative artwork and even videos. 

The Bottom Line

With the application of AI growing at a faster speed, now is the best time than ever to look into their energy consumption. Implementing this technology can significantly reduce energy consumption while making AI and ML more eco-friendly. 


Please enter your comment!
Please enter your name here