Futuristic artificial intelligence (AI) applications in electronic business, urban surveillance, and medicine require more teamwork among researchers and practitioners. But with the rising number of autonomous AI applications in different domains, including gaming, natural language processing (NLP), and weather forecasting, which platforms and algorithms are essential for the development and deployment of AI products via collective intelligence? Wang et al. propose ACE, an “application-centric, edge-cloud, collaborative intelligence,” for collaborative, universal AI advancement and implementation. The integrated ACE platform is intended for creating extensible, trustworthy, stout, and resourcefully cohesive platforms for high-performance edge-cloud collaborative intelligence (ECCI) products.

The authors concisely review the deficiencies of current ECCI applications. Indeed, future AI applications ought to promote migration into generic edge-cloud collaborative (ECC) infrastructures, to foster cooperative traffic optimization and address the bandwidth overhead required by complex AI workloads in ECCIs. Consequently, they highlight what’s currently missing:

(1) A generic abstraction for ECCI application representations, to encourage collaborative data processing, machine and deep learning, and inferential models;

(2) An integrated, scalable, steadfast, and robust platform for effectively creating and installing ECCI applications, which includes design principles for effectively managing diverse infrastructures, clear user services, and confined performance optimizations in edge-cloud collaborations; and

(3) A platform that specifically focuses on ECCI applications: “the general ECCI application development and deployment procedure based on ACE ... comprises three major phases: user registration, application development, and application deployment.”

The article clearly illustrates how the layers and components of ACE work. The experimental results from the development and distribution of an ACE-based video query are reliable. However, researchers and practitioners will find that there are still unanswered questions. There is currently no security component in the ACE model. The current implementation of ACE is static rather than dynamic, that is, it does not accommodate any resource competition between ECCI applications.

Nevertheless, AI researchers and practitioners should consider the provocative questions discussed in this article. Personally, I wish I had some graduate students and financial resources in order to pursue some of the exciting, timely challenges raised.