AI at the edge: Bringing intelligence closer to machines

• Author:
• Author name

  Quantumrun Foresight
• July 29, 2022

Insight summary

In the age of artificial intelligence (AI) and big data, edge AI is emerging as a crucial technology. It enables faster decision-making by processing data without relying on an internet connection. Edge AI has seen significant growth, addressing issues of connectivity, privacy, and security, benefiting industries like healthcare, retail, and the Internet of Things (IoT).

AI at the edge context

In the age of artificial intelligence (AI) and big data, ever more processing is being done in the cloud. This paradigm makes sense for certain types of data, like small snippets of text, but breaks down when it comes to larger data sets—this is where edge AI comes in. Edge AI refers to a class of machine learning (ML) architecture in which AI algorithms are operated locally on devices (at the edge of the network). A device utilizing edge AI does not have to be connected online to function correctly and can process data and make judgments without a network connection. This capability is becoming increasingly essential in today’s AI applications. 

For example, in a scenario where a child walks into the path of a self-driving vehicle, traditional computing would see the vehicle transmit the situation to a central cloud server and wait for the cloud mainframe to return the instruction to stop or swerve to avoid the child. This transmission may take longer than the reaction time needed to protect the child. However, if the vehicle could process the situation using an onboard computer, its reaction time would be significantly faster, improving safety outcomes for all involved.

The edge AI era being is driven by the increasing need to process higher volumes of data. Consultancy firm Deloitte estimated that more than 750 million edge AI chips that execute or speed up machine learning operations on-device rather than in a remote data center were sold in 2020, generating USD $2.6 billion in revenue. Tech research firm Gartner predicts that over 50 percent of data created and processed in businesses will happen outside the data center and cloud by 2022. Moreover, edge AI heavily improves on cloud computing AI by eliminating the need for middleman tech. Still, challenges remain, such as data privacy compliance issues caused by storing data in one centralized location (namely, the device).

Disruptive impact

The benefits of edge AI are varied. For one, edge AI can help to overcome poor network connectivity. It can also improve privacy and security by keeping data local, and it can help reduce costs by avoiding the need to transfer large amounts of data over the Internet. Edge AI is also becoming increasingly important for industrial applications. For example, energy company General Electric (GE) has been using edge AI to improve the efficiency of its wind turbines. The company has developed an AI system that can detect turbine faults and predict when they will need maintenance. This application has led to a significant reduction in turbine downtime.

Another common use for AI at the edge is facial recognition. By installing cameras with AI capabilities at the network’s edge, businesses can scan crowds for people of interest or control access to a facility by only allowing authorized personnel. Smart retail is another common application for AI/ML in edge computing. By using AI to analyze customer service conversations, retailers can recognize patterns that lead to successful outcomes and suggest products that elevate the customer experience. Additionally, AI can recommend related items or services to customers based on their personal attributes.

Healthcare is another industry that is benefiting from edge AI. Doctors can now use AI for predictive diagnoses based on patient history, and AI can also analyze images to check for anomalies like tumors. Finally, the Internet of Things (IoT) benefits the most from edge AI, particularly for manufacturing companies that need real-time updates to correct production chain errors and lapses. 

Implications of edge AI 

Wider implications of edge AI may include: 

• Rapid developments in natural language processing (NLP) ML, resulting in better customer responses for call centers, more intuitive security (AI being able to detect broken glass and gunshots), and legal assistants that can review and connect multiple documents.


• Companies using edge AI to provide real-time information on products even without packaging, such as cosmetics, nutritional facts, expiration dates, etc. Consumers can scan the product themselves (without QR codes), and all the details will be provided.


• Federated learning being used to train edge devices by using local data, ensuring that personal information never leaves the device, resulting in better data privacy protection.


• Smartphones and other personal devices potentially having longer battery lives and faster performance.


• New legislation governing how and what data can and cannot be stored on local devices using edge AI.


• A growing consumer expectation that all the products they buy must become “smart” in some way. Future generations may view items without any computational element as “broken.”

Questions to consider

• Have you interacted with AI at the edge technology in your workplace?


• How else might devices that can operate without online connections better serve customers?