I. From Artificial Intelligence to Edge Artificial Intelligence
Therefore, edge intelligence was born. Edge intelligence refers to edge intelligence. It is an open platform that integrates core functions of networking, computing, storage, and applications, and provides edge intelligence services to meet key requirements such as agile connectivity, real-time business, data optimization, application intelligence, security, and privacy protection. Deploying intelligence on edge devices brings intelligence closer to users and enables them to receive intelligent services faster and better.
With the development of network technology and the widespread use of mobile devices, edge intelligence technology has attracted great attention from governments, academia, and industry both domestically and internationally since its inception. However, edge intelligence is still in the early stages of development and faces significant challenges.
II. Edge Artificial Intelligence and the Internet of Things
Edge AI can be combined with other digital technologies, such as 5G and the Internet of Things (IoT). IoT generates data for edge AI systems to use, while 5G technology is crucial for the continued development of both edge AI and IoT.
The Internet of Things (IoT) refers to a variety of smart devices interconnected via the public internet. All these devices generate data that can be fed into edge AI devices, which can also serve as temporary storage units for the data. Data processing methods offer greater flexibility.
5G technology is crucial for the development of edge artificial intelligence and the Internet of Things (IoT). 5G can transmit data at speeds up to 20Gbps, while 4G can only transmit data at speeds of 1Gbps. 5G also supports more concurrent connections and lower latency than 4G. These advantages are significant compared to 4G because as the IoT develops, data volumes will increase, impacting transmission speeds. 5G enables more interaction between more devices, many of which can utilize edge AI technology.
III. What are the obstacles to the development of edge AI?
(I) Accuracy and Real-time Performance
The latency of edge intelligence includes computational latency and communication latency. The former depends on the capacity of the edge nodes and the scale of the computational model, while the latter is affected by the amount of data transmitted and network bandwidth. Due to the development of deep learning, most current intelligent models use deep neural network algorithms. Large neural network models improve the accuracy of computational results, but also increase the computation time of edge nodes. Ensuring that the accuracy of computational results meets the real-time requirements of edge intelligence computing is a significant challenge in edge intelligence research.
(II) Accuracy and Energy Consumption
When training intelligent models in a distributed manner, both computation and communication processes consume significant amounts of energy. However, these are energy-constrained for most end devices. Energy efficiency is primarily influenced by the size of the target training model and the resources available to the edge devices. Generally, the higher the accuracy of the computation and the larger the model, the greater the energy consumption of the edge nodes. A key challenge in edge intelligence is striking a balance between the accuracy of the computational model and the energy consumption of the edge nodes.
(III) Service Quality and Privacy Protection
Regarding user data security, since edge nodes are located close to the user terminal devices that generate the data, using edge nodes to store data can, to some extent, prevent data leakage and protect user privacy. However, as data-driven algorithms, artificial intelligence algorithms typically require a large amount of data to support their implementation. Insufficient data prevents intelligent algorithms from performing perfect training, reduces their accuracy, and ultimately affects service quality. Therefore, it is urgent to study how to protect data privacy and security in edge intelligence scenarios without compromising service quality.
