Researchers from George Washington University have built a machine learning light-dependent computer system. The photonic platform will be faster and more efficient than the current electric dependent processors that are slow in transmitting data to and from the memory.
Researchers have successfully modeled a machine learning processor that can uses light photons instead of electricity to compute data. The development has been hailed a breakthrough in the field of artificial intelligence by the researchers.
For decades, machine learning has mesmerized researchers, but the field has been held back by the speed and efficiency of computations. For simple tasks, the power required to perform the tasks can be easily met through electricity, however, complex tax requires more power and higher computational power that current processors have failed to adapt to.
Slow transmission of data of electronic data from processors to memory has also contributed greatly to the slow growth of artificial intelligence. To be able to replicate the human brain, train and operate without supervision requires speed that the current systems have not been able to offer.
Machine learning research
Researchers from George Washington University are, however, looking to change the machine learning field through their recent innovation that will allow usage of photons within neural networks (tensor) processing units (TPUs), which they believe will be able to address this problem. TPUs will be able to overcome the current machine learning limitations and create more powerful and efficient artificial intelligence systems.
The research was published on Applied Physics Reviews and showed that TPUs were able to perform between 2-3 orders of magnitude higher than the current electric processors.
The research is expected to have applications in fields such as 5G and 6G networks and the processing of vast amounts of data in a short time. It will also help to advance the field of machine learning.
Featured image by Pixabay
