The power of the IoT achieved without the cloud? Its possible, thanks to edge AI chips
By Max BurkhalterFebruary 18, 2020
Most would believe that the internet of things inherently needs a connection to the internet to function. After all, the IoT is composed of connected devices sharing their data thanks to network access. But, this is not actually the case, as the concept of edge AI shows. British chipmaker ARM has unveiled two new edge AI chips that can conduct advanced machine learning without a connection to the internet.
What is edge AI
In simplest terms, Edge AI is artificial intelligence or machine learning algorithms that are processed locally on a piece of hardware without a connection to the internet, according to MIT's Technology Review. The device uses data from signals or sensor data created on the device to make its decisions. Therefore, decisions can be made and analytics performed with or without an internet connection. This method of AI has several benefits. Using only local data for digital learning means solutions are formulated efficiently and with speed. In addition, by not constantly connecting to the cloud to update and synchronize data, the device can have a much longer battery than IoT-enabled ones. Most importantly of all, the lack of a connection to the outside world means that the information on these devices is far more secure than information on a traditional IoT device.
Edge AI is such a focus right now thanks to the fact that it enables devices to efficiently formulate data, decisions, and actions in a matter of moments, even in areas that have intermittent or no access to the internet, or on devices that aren't equipped to connect to the cloud (whether due to space, cost, or similar issues). And thanks to the fact that all data is processed locally, firms can have the peace of mind that this information is at no risk of falling into the wrong hands. Edge AI is a cost-cutting, space-saving, privacy-bolstering technique that will still allow devices to harness the benefits of the IoT without actually being connected to it.
Edge AI is also very fitting for consumer devices. Thanks to the low latency of local processes, customers can conduct the tasks and retrieve the information they want in no time. If a customer has a poor internet connection where they are, their device can be still be used to its full extent, and relevant data uploaded or downloaded from the cloud once a connection is restored at a later point.
What do these chips offer?
ARM's chips are AI-enabled devices that, by connecting to the cloud, can conduct complicated data processing and decision-making tasks with all the capability of an IoT-connected device. The company believes these products are perfect for any company that makes machine learning devices, as they offer "a transformational amount of compute [sic] capability." Currently the company offers two chips: the ARM Machine Learning Processor and the ARM Object Detection Processor.The ARM Machine Learning Processor is a generalized chip that can be used in many different ways. ARM states that this chip can be used for operations like machine translation and facial recognition. Thanks to its adaptable and versatile nature, this chip is the forerunner product of ARM's line of products. Meanwhile, the ARM Object Detection Processor is designed more with visual tasks in mind - its ideally suited to detecting objects from photo and video and inputs.
Recently, however, the company expanded its product line with two new designs, the ARM Cortex-M55 and the Ethos-U55. The Ethos is "a neural processing unit that is meant to pair with the Cortex-M55 for more demanding use cases," according to TheVerge. These new devices are unique as they are not meant for phones or other consumer devices - instead, the company says they are applicable to a wide variety of IoT devices that don't often have the capabilities that a smartphone would.
Edge AI might not require a connection to the IoT to function optimally, but most modern businesses rely on it anyway. For companies looking to make the most out of the IoT and emerging technologies like machine learning, strong network infrastructure is key. Perle offers high-performance connectivity tools that can help enterprises prepare for the future of business technology. Our industrial-grade Ethernet converters and reliable console servers enable IT admins to focus on what really matters - building a strong framework for IoT integration. Read some of our customer stories to find out how we've helped other companies take full advantage of connected technologies.
