Chinese scientists invented a revolutionary artificial intelligence that feeds on light instead of electricity

< img src = "/uploads/blogs/d2/75/ib-Fr6im3utn_f7b27152.jpg" Alt = "Chinese scientists invented a revolutionary artificial intelligence that feeds on light instead of electricity"/> ~ < p > scientists from Nanchin University (China) have presented a breakthrough development & mdash; A photo neural network that works in light instead of electricity. This discovery can revolutionize data processing for systems such as distributed acoustic sensor (DAS) used to monitor infrastructure, detect earthquakes and protect underwater cables.

~ ~ > 62 > < p >DAS systems generate huge data streams that are difficult for traditional electronic computing devices in real time. It is here that the photograph gives the advantage of & mdash; Replacing electronic calculations for optical processes allows you to speed up work and significantly reduce power consumption.

< p > Scientists have developed the so-called & laquo; a photo neural network accelerator with time-frequency cartoon & raquo; (Twm-Pnna). In this system, the role of mathematical filters of the neural network serve rays of light of different wavelengths, and signal processing occurs with the help of lasers and special optical modules.

< p > The team solved complex technical tasks & mdash; In particular, the problem of signal distortion due to frequency displacement and providing full -fledged optical connections within the network. The results of the experiments showed that the developed system reaches the accuracy of recognition more than 90%, approaching standard electronic systems.

< P > Twm-Pnna strikes up to 1.6 trillion operations per second with energy efficiency of 0.87 Tops per Wat. Theoretically, the system can reach 81 trillion operations per second of 21 Tops on OJSC & mdash; This is at times exceeding the capabilities of modern graphic processors.

< P > According to researchers, such technology can be the basis for the latest monitoring systems of critical infrastructure, seismic control and safety of transport, where the processing of giant data in real time is required.

~ ~ ~ ~