Machine learning is rapidly integrating into the consumer end-user experience. Netflix routinely recommends shows I may enjoy based on my viewing history. Snapchat's filters are driven by complex facial recognition and evaluation algorithms. Facebook can identify which of my friends are in a photo with uncanny accuracy.
The potential impact of machine learning is not bound by industry or trade. Health care, agriculture, media / entertainment, household appliances, aerospace, and defense are all being explored and enhanced by machine learning.
The most common architectures found in today's machine learning implementations are cloud-based. Voice assistants such as Siri and Alexa require connection to powerful servers, and use massive amounts of computational power and data streaming. However, a new trend is emerging to optimize machine learning algorithms and move the computation onto the mobile device.
The shift towards performing machine learning tasks on mobile devices has been heralded by Apple's CoreML library for iOS 11, as well as Google's release of TensorFlow Lite (supports both iOS and Android). Qualcomm is also making waves on this trend. Their Snapdragon processors are designed from the ground up to harness the power of machine learning on smartphones and tablets. They also offer an SDK for supported Android devices. Developers can now take advantage of these libraries and integrate AI into their mobile applications without needing to connect to the cloud.
With these and other recent advances, we are witnessing the beginning of an exciting new era of smartphone app intelligence. Machine learning is becoming more efficient and ubiquitous on mobile devices.
Mobile applications that perform machine learning on device are more reliable. When internet connection becomes intermittent or non-existent, these apps can still provide value to users and perform sophisticated tasks without relying on the cloud. This makes the applications much more dependable.
Many people have raised privacy concerns in this modern era of technological advancement with our newfound dependence on the internet. People today are concerned about issues like identity theft and mass surveillance. By performing machine learning tasks locally within mobile apps, data security stands to be greatly improved. All of the input data for a machine learning model can now be accessed and analyzed on the mobile device itself. Instead of the data being routed through various network switches and external servers, it can now remain safely on the device which greatly improves security.
As non-technical end users become aware of the potential for AI, consumer expectations are rapidly outpacing what the current AI / machine learning platforms are actually capable of delivering. For example, the release of the Apple HomePod was met with widespread criticism. Rather than praising its ability to analyze listening habits and provide suggestions for new music, many users criticised it for lacking the ability to call an Uber or differentiate between voices of different household members. Developers, UI/UX designers, project managers, and business leaders alike should all be cognizant of the rising user expectations, and strive to improve products and tools by exploring the possibilities of machine learning.