Distributed intelligence is an architecture that breaks computation workloads hosted in a centralized system into a combination of a central and end node. Edge Artificial Intelligence (AI) gateways and edge AI servers are the enablers of distributed intelligence, and the steep and steady rise of these gateways and servers takes off in 2021.
ABI Research forecasts the annual shipments of edge AI gateways and AI servers will grow from 8.55 million and 69,000 in 2020 to over 10 million and 105,000 in 2021 respectively. This significant 2021 growth will usher in the age of distributed intelligence over the next decade with annual shipments of edge AI gateways reaching 59.5 million and AI server shipments hitting 1.5 million in 2030.
Stuart Carlaw, chief research officer at ABI Research, warns that success in 2021 requires an early understanding of fundamental trends. “Take a view on those trends that are buoyed by hyperbole and those that are sure to be uncomfortable realities. Now is the time to double down on the right technology investment,” he suggested.
Distributed Computing will become significant
ABI Research says distributed intelligence has greatly benefited from the design and implementation of various systems. Examples include cloud computing clusters, warehouse robots, and smart home systems. These systems are often limited by geographical factors, connectivity options, and the processing capabilities in end nodes.
The emergence of 5G and AI is set to change these. “Combining the high throughput, low latency, and massive IoT connectivity of 5G, with on-device inferencing capabilities of AI, 5G-enabled edge AI devices will have the flexibility to centralize all their workloads in the cloud or perform time-, latency-, and security-sensitive workloads at the edge,” explained Lian Jye Su, AI & Machine Learning principal analyst at ABI Research.
This removes data privacy, safety, and security concerns, while allowing the overall system to update and optimize itself. “Expect major webscalers and chipset suppliers to align their products and solutions in 2021 to address demand for more distributed intelligence across both the consumer and the enterprise markets,” Su added.
Zero-Code onboarding for Edge AI
Edge AI deployment has been a challenge as there is a diverse range of edge AI chipsets, frameworks, and toolkits. Some players in the market are coming up with a zero- or low-code deployment platform.
“These platforms support zero-code web user interface-based deployment, cloud-based device monitoring, orchestration and management, alert management, and ML model performance monitoring and retraining. Expect more startups and system integrators to focus on new offerings targeting zero-code onboarding,” Su concluded.