Modern smartphones have high processing power. Most of them have separate processors for graphics. This makes them more powerful than desktop computers were 10 years ago. However, experts at the Massachusetts Institute of Technology have succeeded in developing a microchip that will make our pocket "supercomputers" even faster.
The new chip, called Eyeriss, is a lot like today's GPUs. Following their example, Eyeriss has a whole group of processor cores that allows it to successfully cope with parallel computing. In other words, the chip will be able to handle multiple data streams at once, unlike a conventional CPU.
But why do you need Eyeriss if you already have a GPU? The answer is self-learning systems like neural networks. They allow artificial intelligence to learn and solve complex problems that are so easy for humans and overwhelming for a computer. This includes object recognition, voice decoding, predictive text input, and other features.
Eyeriss will allow your smartphone to look through the camera at the world around and recognize it. The new chip will do this in real time and without the huge waste of battery power. Thanks to such opportunities, completely new services related to pattern recognition will appear.
Chipmakers Nvidia and Qualcomm are already working on similar technology.