This year, both Apple and Huawei have opted for processors with artificial intelligence capabilities.
The Apple Bionic A11 processor features Face ID and Animoji face recognition, while Huawei’s Kirin 970 processor automatically adjusts camera parameters and improves the performance of the Microsoft translation app.
It is quite likely that upcoming Qualcomm and Samsung star processors will also have artificial intelligence features, and the challenge will be for manufacturers and developers to take advantage of this capability to deliver new features and applications.
According to Baofeng Zhang, Huawei’s Vice President for Software Engineering, most apps using cloud services in AI can easily be redesigned to take advantage of NPU(Neural Processing Unit) of new chips by using the API Kirin or Open Source as TensorFlow and Caffe2.
This will allow apps to improve their performance, be less dependent on having connectivity and increase privacy as user data will not have to leave the device.
Zhang has explained that the first batch of third-party apps will probably focus on voice and image recognition, and then come services that will leverage natural language processing to search for content stored on your phone.
This could lead, for example, to better video editing tools. We could imagine a scenario in which you record a complete football game for your son and then you can ask the app to highlight the best moments with a command like “Show me fragments of this video in which Luis is playing football. “
This could be a jump from the apps that select the best moments depending on the sound of the recording and other sensor data. An AI chip could help make smarter decisions about which parts of the video should be highlighted.
Microsoft is already working with Huawei to improve the translation capabilities of its offline app thanks to the NPU chip. The app comes installed on the Huawei Mate 10 but still does not allow voice translation without an Internet connection.
Regarding the next generation of AI chips, Zhang has explained that they will have more processing power, but the evolution of frameworks like TensorFlow Lite, which developers can use to create their apps, could affect the development of these processors.