Photo / Frank V.
The photon-based neural network processor is 100 times more efficient than the electric processor.
Researchers have made progress in developing artificial intelligence using light instead of electricity to perform calculations. The new approach significantly improves both the speed and efficiency of machine learning neural networks – the purpose of artificial intelligence is to perform repetitive activity, such as the function performed by the human brain to perform a task without supervision.
Current processors used for machine learning have the power to process data, limiting the amount of work that can be done during complex operations. As that task becomes more intelligent and the data becomes more complex, more power is required. The slow transmission of electronic data between processors and memory limits the performance of such networks.
Researchers at George Washington University in the United States have discovered that the use of photons in natural neural network (stress) processing units (TPUs) creates a more powerful artificial intelligence that transcends these limits.
A newspaper describing the research published in reviews by Today’s journal of Applied Physics reveals that their photon – based TPU was able to react between 2-3 orders of magnitude higher than the electronic TPU.
“We’ve found that integrated photonic platforms that combine efficient visual memory can have the same operation as a stress processing unit, but they consume less power and yield higher results,” said Mario Misguglio, the newspaper’s editor.
“Opportunistic training uses light to operate (within the system).” Potential commercial applications for the innovative processor include 5G and 6G networks, as well as a large number of data processing centers.
Dr. Miskuglio said that “photonic processors can save a lot of energy, improve response time and reduce data center congestion.”
Do you have a story you would like published on this website?