Google continues to expand its range of artificial intelligence products and services and has launched three new hardware devices for the development community.
They will be launched under the new brand of Google Coral, including a $149.99 development board, a $74.99 USB accelerator, and a $24.99 5 megapixel camera.
Both the development board and accelerator use Google’s Edge TPU chip, which is an ASIC processor smaller than a nail and designed specifically for running AI models. Meanwhile, this camera is an additional component on the development board.
These products are not suitable for training machine learning algorithms. This is still a heavy task that requires GPU or cloud processing capabilities. On the contrary, they advocate for the “reasoning” part of artificial intelligence, which is generated after you train your model and use it on-site. For this reason, both accelerators and development boards support TensorFlow Lite, a lightweight AI framework designed by the company for mobile and IoT devices. Of course, you can also
Run AI models on any old CPU, but they will be slow and unstable. Specially designed chips, such as Google’s TPU processor, can perform this task more efficiently without the need for special internet connections or bulky power supplies.
The USB accelerator can enhance the inference ability on any Linux machine, and the pin and port array of the development board makes it very suitable for prototype hardware and other experimental applications.
Google is not the first company to provide developers with this type of product, as Intel released its own USB accelerator – a movie driven neural computing stick – a few years ago. Such products can help companies reach out to the developer community and promote the adoption of other artificial intelligence services.