Researchers from the United States have presented a new method for teaching AI that will speed up the process by 2-3 times. In the future, such processors can be used in 5G networks.
Researchers have made a breakthrough in artificial intelligence (AI) training using light instead of electricity. The new method dramatically improves both the speed and efficiency of neural networks, a form of AI that aims to reproduce the functions performed by the human brain. This is how the machine teaches itself to a certain task without observing how the person copes with it.
Current machine learning methods are limited in performing complex operations – they require enormous power. Moreover, the more difficult the task, the more data and, therefore, power consumption. Such networks are also limited by the slow data transfer in devices.
Researchers at George Washington University in the United States have found that using photons in neural processors will help overcome these limitations and create more powerful and energy-efficient AI.
A research paper published in the scientific journal Applied Physics Reviews describes their photon-based processor capable of performing complex tasks 2-3 times faster.
Scientists are confident that in the future they will be able to teach AI very quickly. Potential commercial applications for the innovative processor include 5G and 6G networks and data centers that handle massive amounts of data.
In their opinion, photonic processors can save huge amounts of energy, improve response times, and reduce traffic in data centers.