For a number of years, artificial intelligence has been at the top of the technology must-watch list for evolving trends and applications. With our ability to build smart machines that simulate human intelligence, the implications for technological advancement across numerous sectors are endless. So, what could be better than artificial intelligence? Embedded artificial intelligence.
Embedded artificial intelligence (AI) is the application of machine and deep learning in software at the device level. Software can be programmed to provide both predictive and reactive intelligence, based on the data that is collected and analyzed.
Over the past several years, an important shift has occurred from cloud-level to device-level processing of artificial intelligence tasks, data and results. Embedded AI is the direct result of this important shift. Traditionally, complex AI computations, such as producing search engine results, were performed at a data center in the cloud. With the implementation of AI models on graphics processing units (GPUs), session border controllers (SBCs), and systems on chips (SoCs), there is less of a dependence on the cloud for AI data processing.
With embedded AI, devices have the ability to run AI models at the device level and then directly use the results to perform an appropriate task or action. The cloud is still helpful from a data storage perspective, as data can be stored temporarily at the device level and eventually sent to a cloud server for safekeeping.
While the uses and application of embedded AI are vast, here’s a short list of industries where this technology is automating processes, providing advanced analytics and business insights, and improving customer service, among numerous other benefits.
- Agriculture
- Aviation
- Field Service Management
- Finance
- Healthcare
- Manufacturing
- Retail
- Shipping
- Supply chain
As the technology behind embedded AI continues to evolve, two evolving applications to keep a beat on include embedding AI capabilities onto custom SoCs and Internet-connected devices or the Internet of Things (IoT). Embedding AI models on SoCs optimizes the chip’s architecture thereby reducing instruction counts, power consumption, and calculation time. While embedding AI in Internet-connected devices sounds like both a slippery slope and the next logical step, a number of companies, such as Google, Siemens and HPE, have already dipped their toes in this space.
In manufacturing and industrial settings, pairing embedded AI and the Internet of Things can result in predictive maintenance for equipment, increased operational efficiency, improved products and services, and enhanced risk management. Of course, the applications for embedded AI and IoT devices in other settings abound, with implications in security and monitoring systems, smart homes and cities, scientific research, and healthcare to name a few.
Sources
G2
Embedded AI: Embedded Systems Trends for 2019
By Rob Light, January 18, 2018
https://learn.g2.com/trends/embedded-ai
NWES
The Future of Embedded AI and Custom SoCs
By Zachariah Peterson, March 31, 2020
https://www.nwengineeringllc.com/article/the-future-of-embedded-ai-and-custom-socs.php
Towards Data Science
The power of combining AI and IoT
By Shanika Perera, October 2, 2019
https://towardsdatascience.com/the-power-of-combining-ai-and-iot-4db98ac9f252