Oct. 10, 2021
The traditional way of doing AI has been to offload the computation to remote servers in the cloud, rather than on the local device where the data was collected. This is for a very good reason: many AI tasks are extremely resource-intensive, requiring large amounts of processing power in order to finish in a reasonable time.
Up until now, local devices simply didn’t have the capacity to perform these computations themselves. For example, if you have an iPhone, you may have noticed that the Siri virtual assistant won’t work unless your phone is connected to the internet. This is because the tasks that Siri performs, such as speech recognition and natural language processing, are simply too computationally intensive to run on your iPhone’s hardware.
However, all of this has changed with technological advancements that have enabled edge AI. Rather than having to upload data to the cloud for processing and analysis (and then download the results), edge AI allows users to perform computations on so-called “edge devices” that are in close physical proximity to the point of data collection. This paradigm shift has several advantages: less lag and latency, greater reliability and availability, lower costs for third-party cloud services, etc.
Edge AI is certainly shaking up the way that businesses of all sizes and industries approach the question of artificial intelligence. According to the market research firm Valuates Reports, the global edge AI market will grow from $713 million in 2019 to $3.4 billion in 2026, with a ferocious annual growth rate of 25 percent.
In the near future, we can expect to see edge AI chips installed in a wider and wider variety of consumer devices: mobile devices like smartphones and tablets, wearable technologies, “smart” electronics such as televisions and virtual assistants, and even the cameras and sensors themselves that collect data. These edge AI chips will allow for exciting new capabilities and use cases such as facial recognition, speech recognition, and augmented and virtual reality—all of it faster and cheaper, without the need for an internet connection.