Overview

AI (Artificial Intelligence) has long been viewed as a foundation stone for automated and IoT (Internet of Things) technologies to grow and develop in innovative ways. In the last decade, the progress made by embedded computing systems has been truly impressive, with the AI chips market alone being poised to grow at a CAGR of more than 42% between 2020 and 2024. Within this tech blossoming, Embedded AI systems are now readier than ever to leverage the advantages of this new digital landscape.

EAI (Embedded AI), as defined by Gartner, refers to the use of Artificial Intelligence (AI)/Machine Language (ML) techniques within embedded systems to enable the analysis of locally captured data. When combined with edge computing (computing inside the device) and FOG computing (computing in Local Area Network), EAI significantly reduces latency, enables deeper predictive analytics, and reduces the time of response in severe failure cases.

What EAI Means For You

Figure 1: Edge – Fog – Cloud - Compute

Figure 2: Typical latencies

Given the rapid increase in the rate of data generation and acquisition, it has become increasingly important to have intelligent edge devices run AI models that are capable of adaptive and incremental learning for specific tasks.

As the volume of data increases, it becomes increasingly compute intensive for AI model development.

In typical embedded AI-EAI applications, the computational and memory-intensive parts (for instance, training) are often offloaded onto the powerful GPUs/computations in the cloud, and only certain light inference models are deployed on edge devices such as the IoT or mobile devices. EAI can greatly reduce turnaround time by providing precise insights and trends from the captured data while maintaining high levels of privacy and security.