AI at the edge: 5G and the Internet of Things see fast times ahead

AI algorithms running on edge devices now offer real-time insights and actions that can improve response times.
Written by Joe McKendrick, Contributing Writer
5G in city
AerialPerspective Images/Getty Images

Connected devices linked to the Internet of Things (IoT) -- in association with 5G network technology -- are now everywhere. But just wait until next-generation applications, such as artificial intelligence (AI), start running within these edge devices. Meanwhile, the low latency and higher data speeds of 5G and IoT will add a new real-time dimension to AI.  

Consider an extended reality (XR) headset that not only provides a 3D view of the inside of an aircraft engine, but which also has on-board intelligence to point you to problem areas or to information on anomalies in that engine, which are immediately and automatically recognized and adjusted. 

Also: Ahead of AI, this other technology wave is sweeping in fast

Chipmakers are already developing powerful yet energy-efficient processors -- or "systems on a chip" -- that can deliver AI processing within a small footprint device. For instance, Qualcomm just announced AI-capable Snapdragon chips that run on smartphones and PCs. Also on the horizon are a generation of NeuRRAM chips, developed at the University of California San Diego, which are capable of running sizeable AI algorithms on smaller devices. 

Overall, the global number of connected IoT devices is projected to surpass 29 billion by 2027, which is more than 16.7 billion at the present time, a recent analysis from zScaler shows. "Consumer devices are smart and most common, but business process-oriented IoT generated the most transactions," the report's authors point out. "Manufacturing and retail devices accounted for 50%-plus of transactions, highlighting their widespread adoption and business-critical function in these sectors. Enterprise, home automation, and entertainment devices are generating the highest counts of plaintext transactions."

Now, 5G and IoT technologies are opening new doors to innovation within AI -- and vice versa. AI "will be more effective when enabled with a local-level decision-making framework and with near real-time data," says Arun Santhanam, vice president and head of telecommunications at Capgemini Americas. "5G low latency innovation will be key for enabling the outcome of real-time data coming from relatively inexpensive IoT solutions."

Also: If AI is the future of your business, should the CIO be the one in control?

Most viable edge and AI use cases have been in the enterprise and IoT space, within industries such as healthcare and manufacturing, says Haifa El Ashkar, director of strategy of the telecommunications market and solutions at CSG. These companies "need to offer faster data transmission and real-time communication," she says. "5G's lower latency and faster processing capabilities, coupled with edge architectures, have proven crucial for applications that require quick decision making and responsiveness."

In healthcare, for example, "there are now AI-edge-supported medical devices such as laparoscopes, allowing surgeons to leverage real-time insights and make faster decisions on life-saving measures such as identifying anomalies that might otherwise have been missed or detecting bleeding in real time," says El Ashkar. "Without 5G, these industries would be unable to tap into edge networks and offer the services needed to meet the needs of powerful critical IoT uses cases such as these."  

The proliferation of AI-enabled applications and services is also amplifying the power of 5G edge applications, El Ashkar continues. "When you combine the low latency of 5G networks and AI capabilities at the edge, enterprises can access real-time decision-making," she says. "With less time needed for data to travel back and forth between devices and data centers, AI algorithms running on edge devices are now offering real-time insights and actions that can improve response and increase the amount of valuable data available to the enterprise."

AI also improves connectivity, as it "can have a dramatic impact on the reliability and efficiency of wireless networks and enable new ways of staying connected," says Milind Kulkarni, vice president and head of InterDigital's Wireless Lab. "For example, the combination of 5G, cloud, and edge computing is crucial for empowering immersive experiences on new devices in more places and the development of connected ecosystems such as the metaverse. Innovations in 5G and computing capabilities help make these experiences a reality."

Also: Businesses need a new operating model to compete in an AI-powered economy

While more centralized environments -- cloud and data centers -- may provide computing power for immersive experiences, "they may be too far from where low-latency resources are located," says Kulkarni. "So, to take advantage of the ultra-low latency that is one of the key benefits of 5G, edge computing plays a vital role by offering smaller amounts of storage and computation much closer to the device where it's needed. In addition, edge computing can be customized to support specific use cases such as storing content for delivery of video on demand or running AI algorithms for fast decision making on incoming data."

XR is an area where the capabilities of 5G are being pushed to the limit. "Currently there is a large amount of ongoing work within 3GPP that is focused on enhancing current networks to be more aware of and better support XR traffic," says Kulkarni. "XR pushes the limits of 5G in terms of latency at very high data rates, efficient video coding and network architecture, for example by taking advantage of edge computing's benefits." 

5G's high speeds and low latency "will be required for industries to transition into the next stage of digital transformation," says El Ashkar. "This is critical to industries such as supply chain, healthcare and manufacturing, where increasingly more AI-infused and connected devices are becoming vital to daily operations."

Editorial standards