The edge AI market is forecast to grow from $1.4 million in 2021 to $8 million in 2027, at a CAGR of 29.8%. Much of this growth will come from factors such as artificial intelligence for the Internet of Things, wearable consumer devices, and the need for faster computing in 5G networks. These bring opportunity and retention, as real-time data from AI at the edge is vulnerable to cyberattacks.
Let’s take a look at five trends that are likely to shape the AI landscape at the edge over the next year.
Decoupling AI from the cloud
One of the big changes these days is the ability to run AI processing without a cloud connection. Two new chip designs, for example, recently announced, could push the processing power of IoT devices to the limit, skipping remote servers or cloud computing. Their current Cortex-M processors can handle object recognition, while other features like gesture or voice recognition come into play with the addition of ARM’s Ethos-U55. Google’s Coral, a toolkit for building products using native AI, also promises to handle a lot of AI “offline”.
machine learning in action
Best practices for machine learning operations will prove that edge AI is a valuable business process. It requires a new lifecycle for production — or, at least, that’s speculation during MLOps development. MLOps can help enterprise data flow and push it to the edge. As more businesses discover what works best for them when it comes to AI at the edge, a continuous refresh cycle may prove effective.
dedicated chip
To do more processing at the edge, companies need custom chips to provide enough power. For example, an AI accelerator chip is paired with a software suite that essentially converts an AI model into a computational graph. IBM released their first accelerator hardware in 2021 aimed at fighting fraud.
New Use Cases and Capabilities for Computer Vision
Computer vision remains one of the main uses of edge AI. A major development in this field is multimodal artificial intelligence, which extracts data from multiple data sources, goes beyond natural language understanding, analyzes poses, and performs inspections and visualizations. This could come in handy for AI that interacts seamlessly with humans, such as shopping assistants.
Higher-order vision algorithms can now classify objects by using more fine-grained features. Rather than identifying the car, it goes deeper into determining the make and model.
It is difficult to train a model to recognize granular features specific to each object. However, approaches such as feature representation using fine-grained information, segmentation to extract specific features, algorithms to normalize object poses, and multi-layer convolutional neural networks are all current approaches to achieve this goal.
Early enterprise use cases include quality control, real-time supply chain tracking, using snapshots to identify internal locations and detecting deepfakes.
AI growth accelerates on 5G
5G and more advanced technologies are coming. Satellite networks and 6G are awaiting the arrival of telecom providers. For the rest of us, it will take some time to transition between 4G core networks that are compatible with some 5G services before fully moving into next-generation networks.
What does this have to do with AI at the edge? AI on 5G can bring better performance and security to AI applications. It can provide some of the low-latency benefits required by artificial intelligence and open up new applications such as factory automation, toll and vehicle telemetry, and smart supply chain projects.
There are more emerging trends in edge AI than we can list. In particular, its development may require some changes on the human side. Edge AI management will become the job of the IT department, and using IT resources instead of letting lines of business manage edge solutions can optimize costs.
GIPHY App Key not set. Please check settings