Bloomberg reported a while back that Apple (AAPL 1.13%) is testing a dedicated artificial intelligence chip that's internally known as the Apple Neural Engine.
This is a reportedly a dedicated piece of chip technology that, per Bloomberg, will "improve the way the company's devices handle tasks that would otherwise require human intelligence -- such as facial recognition and speech recognition."
Image source: Apple.
Ultimately, it seems that Apple is building the technology to help improve the power efficiency of its devices, as doing these intensive computations on specialized hardware will surely be more efficient than running those same computations on the CPU and/or GPU in the Apple A-series processors.
It hasn't been clear what this so-called Apple Neural Engine would be -- until now.
Huawei implemented something similar
Apple rival Huawei, which, like Apple, builds customized chips for its flagship smartphones, recently announced a next-generation processor known as the Kirin 970.
One of the highlights of the new chip is the inclusion of a dedicated "Neural-network Processing Unit," or NPU for short. Huawei reportedly says that the NPU leads to a dramatic increase in both the efficiency and performance of artificial intelligence tasks (think facial recognition) over the CPU or the GPU.
Huawei claims a 25 times improvement in performance and a 50 times improvement in efficiency when such tasks are run on the NPU instead of the CPU.
I think Huawei's announcement (which I suspect was done ahead of the new iPhone launch to steal Apple's thunder) shows us what we should expect from the Apple Neural Engine -- big speedups in artificial intelligence tasks that can make use of the dedicated processor.
The actual specifics and performance characteristics of the Apple Neural Engine compared to Huawei's obviously won't be known until Apple announces, but those specifics aren't really that important.
The future of mobile chips
At the end of the day, mobile chips are all about stuffing a lot of computational capabilities into small form factors at low power consumption.
There will continue to be value in improving the performance of the general-purpose CPU, and there will of course be value in continuing to boost the capabilities of other traditional mobile chip blocks (e.g., GPU, image signal processor, and so on), but for maximum efficiency and capabilities, additional specialized hardware dedicated to certain tasks will continue to emerge.
In such a paradigm, Apple has a big advantage over everyone else. Since Apple tightly controls the hardware design, operating system, and development environment, it can build hardware and expose it to developers relatively easily.
Furthermore, since new Apple technologies quickly proliferate within the Apple iPhone/iPad installed base, it only takes a couple of generations before developers can count on a certain feature being present across a wide range of iOS devices -- incentivizing them to invest in making use of those technologies in their apps.
The future of mobile computing is Apple's to lose.
