Artificial intelligence (AI) is rapidly becoming a game-changer in the succeeding generations of smartphones. According to one analysis, AI was only present in 3% of smartphones shipped all over the globe in 2017. That figure shot up more than tenfold to 35% three years later. With the market saturated, manufacturers have shifted their focus to improving user experience. (1)
It’s hard to deny AI’s active role in making future smartphone models better across the board, from increasing computing power to borderline reading users’ minds. But how does it work exactly? More specifically, what are the AI-based technologies behind making smartphones of today and tomorrow smarter?
Data labelling
How does a smartphone’s camera know what it’s trying to capture? How does its onboard search function find similar images just by looking at a picture? All these features and more are nothing short of machine learning at work, and central to it is a process known as data labelling.
Simply put, data labelling involves taking raw data—whether pictures or videos—and adding a few labels to help provide the correct context. Think of it as teaching a machine that “This is a car” while showing a picture of a Honda Accord. It also teaches that a Dodge Ram is a “pickup” but still qualifies as a car in terms of purpose.
To accomplish this, developers use a data labelling platform, which utilizes either one or several types of data labelling (not limited to the following):
- Audio processing involves converting audio data, whether speech or natural ambience, into a usable format for machine learning. This usually involves manually transcribing the audio file and then adding the labels.
- Computer vision labels pixels or standout features in an image to generate a dataset for the machine. This approach enables the device to sort pictures by quality or filter them based on their contents.
- Natural language processing analyzes parts of text files to contextualize the author’s sentiment in a specific blurb or section. Not only does this work for actual text files but also for text-heavy images. (2)
Most current data labelling methods require human input, which is resource-intensive and prone to error. However, recent technological advances have allowed better machine learning for less human intervention. For instance, active learning combines algorithms and limited supervision through synthesis and selective sampling. (2)
Computational photography
The Camera & Imaging Products Association revealed that smartphone sales grossly outscored conventional camera sales in 2015, and the gap has only widened since. It’s not that difficult to see why: smartphones are less bulky than cameras and can directly upload photos to the internet without having to organize them on a desktop or laptop first. (3)
Several years ago, it was unthinkable that smartphone cameras could go toe-to-toe with their DSLR counterparts in terms of quality. The vast processing power of smartphones has made it possible for them to make adjustments with just a few taps instead of using dedicated photo editing software. Experts refer to this as computational photography.
Algorithms in smartphone cameras run to improve a photo in virtually every aspect, from colour and contrast to noise reduction. The addition of AI means today’s models can run these systems simultaneously with just a press of the Enhance button. One notable example is the “bokeh” or background blur effect, which smartphones can achieve with the proper setup.
AI-enhanced algorithms also eliminate the need for bulky hardware, keeping smartphones as compact as possible. Future models may sport anamorphic lenses to allow for sensors to work 15% to 20% larger than those in use. Of course, sensor size isn’t everything, but photography experts and enthusiasts agree that it’s proportional to quality. (4)
Adaptive battery
Amid the features manufacturers have put in their smartphones, battery life remains a significant shortfall. While they understand the need to make their products last longer, the science behind batteries and marketing challenges have largely prevented them from doing so. The few solutions available, like making the battery capacity larger, usually come with several negative tradeoffs.
The entry of AI in this aspect of smartphone development gave manufacturers a much-needed helping hand. For instance, Android’s adaptive battery system shuts down apps working in the background after a set period of inactivity. Its developers claimed that it reduces CPU usage by 30%, a significant figure in battery life management. (5)
There’s no reason adaptive batteries will fall out of favour when newer battery compositions like silicon anode become widespread. If anything, such a combination will drastically reduce the need to look for a place to charge when needed.
Conclusion
These days, one can’t talk about smartphone development without talking about the role of AI. The technologies explained here are some of AI’s most important contributions but are by no means the only ones. Whatever the smartphone will look like several decades from now, AI will undoubtedly have a hand in its design.
References:
- “How AI and Machine Learning are Transforming Mobile Technology”, Source: https://www.greenbook.org/mr/market-research-technology/how-ai-is-transforming-mobile-technology/
- “What is data labelling for machine learning?”, Source: https://aws.amazon.com/sagemaker/data-labeling/what-is-data-labeling/
- “Smartphones vs Cameras: Closing the gap on image quality”, Source: https://www.dxomark.com/smartphones-vs-cameras-closing-the-gap-on-image-quality/
- “Glass rethinks the smartphone camera through an old-school cinema lens”, Source: https://techcrunch.com/2022/03/22/glass-rethinks-the-smartphone-camera-through-an-old-school-cinema-lens/
“Android is now using AI to help manage your battery life”, Source: https://www.theverge.com/2018/5/8/17327104/android-battery-management-ai-artificial-intelligence-google-io-2018