China aims to produce a chip that can add AI to any device

But China still has to work hard in the technology market being changed by AI.

Sitting on a table at an office located in Tsinghua University, Beijing, China is a chip processing data collected from a nearby camera, the camera is recording the faces of each passing person. It brings about a huge data system. A few seconds later, this chip called Thinker is performing the task given to it in a Chinese accent.

Designed to support neural networks, the chip is a "giant" electrical system consisting of 8 AA batteries. More surprisingly, these eight batteries can power Thinker within a year.

Thinker Chip.

Thinker chips can adjust their computing power and memory to meet the requirements of the software. This is an important attribute because many of the current applications - most of them identify objects in images or understand the voice of the commander - need many different neural networks and information classes different logic.

In December 2017, a scientific report describing the design of Thinker was published in the Journal of Solid-state Circuits - the leading magazine on computer hardware design. With the chip research and development community in China, this is an extremely valuable achievement.

The chip is an example of an important trend that is sweeping through every aspect of the Chinese technology industry. Semiconductor device manufacturers see an opportunity to assert themselves, an opportunity to create hardware that can use AI.Computer chips are the key to the door of AI success, so China must develop a hardware industry worthy of the AI's potential, to make the whole industry a technological force of world.

Computer chips are the key to the success of AI.

"Compared to the times when China reacted to previous information technology revolutions, we can see that the speed of China following the trend of artificial intelligence is the fastest in the world," said Shouyi Yin. Deputy director of Tsinghua University's Microelectronics Institute said.

Although China is the leading producer of sheets and smartphones, their semiconductor industry is still behind the US a lot. Between January and September 2017, China spent 182.8 billion USD to import electric circuits - an increase of 13.5% over last year. Major US technology companies - like Intel or Google and other companies and small startups - are working on developing chips for artificial intelligence applications.

China must be present in this race, so the government has released a three-year plan to achieve mass production of neural network chips by 2020.

Current chips can still run AI software but these chip designs are both expensive and cannot be integrated with small electronic devices. That's why the Chinese research team develops Thinker.

Current chips can still run AI software but the chip design is very expensive.

This Thinker chip can be attached to various devices such as smartphone power, smart watches, robots that handle housework or attach to measurement stations located at dangerous locations. Expected, the first products running Thinker will be released in March.

The team at Tsinghua is not the only one who thinks so. Here in China, similar projects still appear. At the end of January, the research team at the Chinese Academy of Science's Computer Science Institute produced a series of robot chips called Dadu . This chip has two cores, one to operate the neural network and one to control the robot's motion.

The history of China's chip manufacturing industry has risen and also has bass. However, one can clearly see that the speed of development of this industry in China is extremely fast. In the Chinese government's policy in 2015, they made it clear that chip design and manufacturing was a key industry that needed a breakthrough.

"In the future, only chip-producing companies will gradually fade away," said Fengxiang Ma, director of ASIC circuit design at Horizon Robotics, a startup from Beijing focusing on applying expertise. AI on self-healing technology and camera scanning images. Last December, they released two chips, one could identify pedestrians (for cars) and one for analyzing the shopping habits of supermarket goers.

Currently, Chinese chip researchers have a lot of work to do: how do they commercialize their chip design, how to scale up production, how to operate in a machine market properties are being changed by artificial intelligence. Once these problems are solved, they will rise stronger than ever.