What is a smartphone AI processor?

Artificial Intelligence processors seem to be the biggest mobile breakthrough in 2017. Especially, in flagship phones starting with Huawei’s Kirin 970 AI processor announcement, which will be seen in the Huawei Mate 10 (launching next week). Apple also added in an Neural engine in their new A11 Bionic chip which ships with their new iPhone lineup. So how is the AI processor different from what most of us have in our smartphones? Putting it simply, AI processors have an additional inbuilt Neural Processing Unit (NPU) that is capable of lots of parallel processing, uses low power and is capable of cognitive tasks like prediction & classification. More like how the human brain works.

So far, most of the AI processing that you see on mobile apps and assistants take place primarily in the cloud. Mostly on parallel GPU systems that are more efficient that CPUs in AI related processing like prediction, detection, pattern matching and classification tasks. Now with AI processors inbuilt in the phone, these AI tasks can be carried out quickly with complete privacy and no connectivity required.

Heterogeneous computing delivers efficiency by using multiple types of processors for different taks

Huawei Kirin 970

Announced at IFA 2017, the Huawei Kirin 970 is the world’s first mobile processor with an inbuilt Neural Processing Unit. For AI related processing, NPUs have 6x more performance and efficiency than a GPU. That leads to lower power consumption and faster/instant processing. Compared to a CPU a NPU has 25x performance being half the size. It also has a Kirin AI API layer that can be accessed by other apps.

Apart from the AI NPU, the processor has a 12 core GPU, 8 core CPU and a new 4.5G LTE modem that supports dual 4G sims, 1.2 GBPS download speeds and more 4G band coverage.

The AI computing platform in the Kirin processor will enable faster AI experiences with better privacy for the user. We can expect better image recognition and processing with this new AI processor. Looking forward to the in-built applications using the AI processor which will be unveiled at the Huawei Mate 10 launch on October 16th.Β In the near future, we can expect it to enable a better Augmented Reality experience and real-time computer vision.

The Competition, Apple A11 Bionic Chip

Apple added a neural engine in their new A11 Bionic Chip. It basically is two cores that are dedicated to handling specific Machine Learning algorithms. The Neural Engine supports tasks like the CoreML library used by Machine Learning apps and facial recognition in the upcoming iPhone X. It goes with the theme of Apple’s AI on device strategy to improve user privacy.

We can expect Qualcomm, Samsung and other chip manufacturers to follow an addition of the NPU to their processors with similar AI capabilities. With billions of dollars poured in AI research by tech companies, AI processors can provide new opportunities for consumers to experience Artificial Intelligence & Machine Learning in apps on their smartphones. The applications that these AI processors enable can be really valuable for the future of mobile.

 

Shashwat Pradhan

Founder of Emberify, a technology startup behind the world's leading lifelogging and Quantified Self Platform- Instant. To learn more visit emberify.com

  • http://www.aboveandroid.com/ Abhishek Vohera

    Computing going to next level. Next generation NPU will make smartphone more smarter. Nice piece of info πŸ‘πŸ»

%d bloggers like this: