When Huawei launched the latest version of its flagship smartphone, the Mate 10, perhaps the biggest selling point was the fact that it has an embedded AI chip. Huawei is firmly convinced that AI chips will be a major differentiator for smartphones and other connected devices, and it went out of its way to tell everyone that the Mate 10 was leading that charge.
Time will tell how much of a difference it makes to consumers (as well as Huawei’s bottom line), but there’s little doubt that the semiconductor sector sees AI as the next big step forward for electronics of any persuasion. And Chinese chip companies in particular see AI as a grand opportunity to not only catch up with US chipmakers, but surpass them to the point of enabling AI chips to be installed in all kinds of gadgets, according to this article from MIT Technology Review.
China has already stated its intentions to become the global leader in overall AI development by 2025. Even half of its new cars will purportedly be AI-powered by 2020. That includes the chipset front, a sector where Chinese companies have made great strides in recent years. But according to PwC, while China accounted for 58% of worldwide growth in the integrated-circuit market between 2000 to 2016, its global share of semiconductor fabrication capacity was still less than 15%.
But thanks to increasing demand for AI – the success of which will depend on AI-optimized hardware as much as the software bit – Chinese companies are keen to capitalize. So is China’s Ministry of Industry and Information Technology, which last month set a goal of developing the capability to mass-produce neural-network processing chips by 2020.
That may sound like a tall order – and a familiar one, given China’s previous attempts to make a name for itself in the chipset space – but as TR reports, this time market conditions are arguably in their favor:
… Chinese chip startups find themselves in an environment that’s vastly different from the one that gave birth to Intel or Nvidia. Businesses have taken to cloud computing in droves, meaning there may be less of a market for off-the-shelf hardware, says Dongrui Fan, president of SmarCo, a Beijing-based startup that designs an AI chip for data centers that process video footage.
But China’s AI companies are increasingly also developing their own hardware.
The article also covers numerous AI-chip innovations emerging from Chinese research, the most interesting of which is ‘Thinker’, a chip for neural networks designed by researchers at Tsinghua University in Beijing:
Thinker can dynamically tailor its computing and memory requirements to meet the needs of the software being run. This is important since many real-world AI applications — recognizing objects in images or understanding human speech — require a combination of different kinds of neural networks with different numbers of layers.
Even more interesting is the fact that it requires relatively little power, especially compared to graphics chips and FPGAs that could technically handle the demands of the AI software, but not cheaply. Plus, their energy requirements prohibit them from being used in small devices because it’s a drain on the batteries. Thinker is designed to change that, which would make it possible to stick AI chips in … well, just about anything you can stick a chip in.
All the details are here.