What is it about?

This survey explores the growing field of on-device artificial intelligence (AI), which refers to AI models that can process data locally on devices like smartphones, smart home systems, and medical devices, rather than relying on cloud computing. The paper discusses the benefits of on-device AI, such as faster response times, improved data privacy, and reduced reliance on internet connectivity. It also highlights the challenges faced in deploying these models, including limited computational resources and the need for optimization techniques. By examining current applications and future trends, this survey aims to provide insights for researchers and industry professionals looking to implement effective on-device AI solutions in various real-world scenarios.

Featured Image

Why is it important?

This work is timely and significant as it addresses the increasing demand for efficient AI solutions that can operate directly on devices, especially in the context of the Internet of Things (IoT). By identifying key challenges and optimization strategies, this survey not only contributes to the academic understanding of on-device AI but also serves as a practical guide for developers and engineers. The insights provided can help accelerate the adoption of AI technologies in everyday applications, ultimately enhancing user experiences and promoting data privacy.

Perspectives

Writing this survey was an enriching experience, as it allowed me to synthesize a wide range of research on on-device AI and its implications for the future. Collaborating with co-authors who share a passion for advancing AI technology made the process even more rewarding. I believe that this work will not only inform academic discussions but also inspire practical innovations in the industry, paving the way for smarter, more efficient devices that can improve our daily lives.

Xubin Wang

Read the Original

This page is a summary of: Empowering Edge Intelligence: A Comprehensive Survey on On-Device AI Models, ACM Computing Surveys, March 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3724420.
You can read the full text:

Read

Resources

Contributors

The following have contributed to this page