Artificial intelligence stands at a crossroads. On the one hand, AI use is growing faster than ever, with the market expanding at a compound annual growth rate of 57.2 percent, on track to reach a value of nearly $59 billion by 2025, Grand View Research projects. But as artificial intelligence usage becomes more widespread, the nature of AI is changing. One major change involves a shift from cloud-based to on-device AI applications. While the cloud has provided key infrastructure support for AI applications in recent years, users are increasingly turning toward applications that run directly on mobile devices. Both efficiency and security are helping promote this shift. Here’s a look at some of the pros and cons of cloud-based vs. on-device AI and when it makes sense to use one versus the other.
Cloud-based AI
Last year, tech giants introduced cloud-based artificial intelligence applications on a large scale. AI applications can be costly for small and medium-sized companies due to the amount of computing resources required, opening up an opportunity for cloud providers to offer services which meet the demand for affordable AI. Amazon has led the way with its AWS Amazon Cloud Service, rolling out an integrated cloud AI development environment along with AI tools that can turn audio speech into text and translate videos into seven languages. Microsoft is also heavily invested in cloud-based AI services.
While the cloud does provide the computing resources needed to make artificial intelligence affordable, it also suffers from some major limitations. One is the amount of time it takes to transfer big data from local resources to the cloud and back, which can make running AI applications in the cloud a slow process. In addition, the popularity of the cloud has made it a target for hackers, resulting in data breaches such as last year’s compromise of 14 million Verizon customers, the result of a misconfigured cloud-based file repository. Cloud services are particularly vulnerable to web application attacks.
The Emergence of On-device AI
To overcome the limitations of the cloud and make AI more efficient and secure, mobile device manufacturers have begun rolling out smartphones that can run AI applications directly on-device without tapping into the cloud. For example, Qualcomm has taken a lead in the field with its artificial intelligence platform, built into its latest generation of smartphone components. Qualcomm’s artificial intelligence platform uses a processor fast enough to allow smartphones to run resource-intensive AI applications such as smart photography focusing, facial recognition, speech recognition and virtual reality.
Cloud vs. On-device AI: Pros and Cons
On-device AI has some distinct advantages over cloud AI, which make it preferable for certain applications. Because on-device AI doesn’t need to rely on remote resources, it can process applications much faster than cloud AI. This makes it superior for applications that require real-time data processing. For instance, autonomous vehicles that rely on AI for safe navigation decisions in real-time will depend on on-device AI. Additionally, on-device AI represents a security advantage. Local devices don’t have to face the remote risks opened by cloud exposure, and AI-equipped devices can feature real-time automatic intrusion detection powered by artificial intelligence.
On the other hand, cloud AI also retains its value for certain applications. Applications involving analysis of large databases often require the scalable computing power offered by the cloud. For example, astronomers are using AI machine learning to scan telescopic data in order to identify inhabitable planets. Similar big data analysis applications in fields ranging from medicine to stock market forecasting likewise require computing resources that outstrip the limits of smartphones.
While the cloud is currently the most popular way to deliver AI services, cloud AI suffers from efficiency and security limitations. On-device AI overcomes these limitations by allowing AI to be run securely on smartphones without waiting for remote resources. As on-device AI becomes ubiquitous over the next five years, on-device AI looks poised to become the dominant delivery method for AI applications.
