Google is rolling out a couple of new products aimed at helping customers build and deploy intelligent Internet of Things (IoT) devices at scale. One of those products is the Edge TPU (tensor processing unit), a tiny hardware chip designed to muscle through machine learning tasks in IoT gadgets, and the other is Cloud IoT Edge, a software stack that extends Google's Cloud AI to gateways and connected devices.
Used together, Google's two new products enable customers to build and train machine learning models in the cloud, then run those models on the Cloud IoT Edge device through the power of the Edge TPU hardware accelerator.
"By running on-device machine learning models, Cloud IoT Edge with Edge TPU provides significantly faster predictions for critical IoT applications than general-purpose IoT gateways—all while ensuring data privacy and confidentiality. Plus, Cloud IoT Edge and Edge TPU have been extensively tested to natively run open source reference models like MobileNet and Inception V3," Google explains.
Edge TPUs are not the same as what Google uses to inject intelligence into services like Photos and Search—those are much bigger and burlier. In comparison, Edge TPUs are much smaller, the fraction of the size of a US penny. These purpose-built ASIC chips are optimized for performance per watt and per dollar, enabling IoT devices to make local, real-time, intelligent decisions, rather than be limited to mere data collection.
Despite the size and power differences, there are some similarities. It's all about analyzing data on the edge, then using that information to perform related tasks. This is not something that is likely to be implemented into consumer devices, but for enterprise customers, Google's multi-pronged solution can help with automated tasks in factories, such as quality control checks.
Along those lines, LG told CNBC that it is testing Google's Edge TPUs in a system that detects manufacturing defects in glass for displays. It's one of many possibilities for an IoT device that can quickly and intelligently process real-time data.