AI technology is set to penetrate at least 30-percent of the smartphone space within the fourth quarter of this year, according to research by DigiTimes. The firm also noted that hardware-based solutions that use separate processors for AI functions are slowly becoming more popular than the somewhat more limited software solutions still seen in many smartphones. The research points to specialized neural processing units made for neural networks as seeing a rise in popularity in the near future. This is reportedly thanks to high demand for AI-based services like image recognition, voice processing and facial recognition.
The reason for the rise of neural processing units is that, quite simply, they’re the best at what they do in their field. One example of that can be seen in a comparison between two 2017 flagship chipsets, Qualcomm’s ever-popular Snapdragon 835, which found its way into phones like the Lenovo Moto Z2 Force and the Samsung Galaxy S8, and Huawei’s HiSilicon Kirin 970 chip, which graced its top-of-the-line Honor offerings. The HiSilicon chip uses a separate neural processing unit made by Cambrian, while the Snapdragon 835 uses Qualcomm’s own Hexagon digital signal processor, repurposed with new code and a few hardware tweaks. The dedicated silicon left the Hexagon in the dust in a number of image recognition and AI processing tests. Some vendors, such as Apple, are still using software-side solutions that don’t have any dedicated silicon inside the chipset but instead pull resources for AI processing from the main system stack.
Native AI technology in smartphones is very quickly making the shift from novelty to necessity as big players lead the way with useful AI features that users can take advantage of in their everyday lives. Apple, for example, introduced a new feature for select handsets in iOS 12 that allows users to measure things around them, and even in pictures that have already been taken in some cases, without any additional apps. Naturally, this trendsetting move will almost certainly be emulated by Android OEMs in the near future. Meanwhile, Qualcomm and its contemporaries have been using onboard machine learning and other AI tools in smartphones to power small optimizations to the system and to the user experience, along with AI-based applications in the area of image processing and natural language processing.