AI chip is only small as salt grains

Optical cable has a difficult problem to solve. Although it is possible to transport encrypted data at light rate, convert coding data into easy -to -understand information often requires a slow operation device and more energy consumption. However, based on previous achievements in the passive Neuron network, the Shanghai University of Science and Technology University (USST) developed a micro solution. They make a new AI chip using optical physics to analyze data with extremely small energy and each chip is only small as salt grains, Popular Science On February 8 reported.

In a study published in Nature Photonics, the author said they relied on a form of Neuron network developed for the first time by researchers at the University of California, Los Angeles in 2018. Optical diffraction worms “, this method uses passive 3D printed parts that are stacked accurately. The system is then trained to complete the complex calculation, using light photons.

The recent research team in USST used this idea as a starting to create a passive Neuron network trained to control the light to make calculation analysis. However, all of that encrypted data is transmitted through the fiber optic cable that is widened as a human hair, so they need AI chip no less to read each photon.

Researchers rely on the method of engraving two -dimensional two -dimensional photons to build each micro chip, using ultra -thin polymer layers. After that, they attach a chip at the end of each fiber optic cable, where the data processed when it transmitted through the cable at the light rate. To test the invention, the research team encoded images of numbers into photons, then transmitted through the fiber optic cable. AI chip then successfully read data and regenerate each image with extremely small blur. This type of image identification is now a rudimentary function in many AI systems. AI chip performs this function in a trillion -second part, with energy equal to a few thousand energy that the image recognition technology is based on AI today.

However, the system is not complete because the minimum recommendation of the chip can cause the whole system to degrade over time. Each chip must be customized separately depending on the task to use. However, inventors believe that technology can provide unprecedented functions such as laparoscopic photography and quantum computer.