Prior to this, Google has allowed users to use TPU resources on a paid cloud infrastructure. But by the 'Edge TPU' project announced on Wednesday, the company will allow programmers to install TPUs on their own robots.
Injong Rhe, vice president of Google Cloud Internet of Things, said in a blog post:
Users can get a lot of benefits from these networked smart, real-time decision-making devices. Your sensors will be more than just data collectors, but more local, real-time, and intelligent decisions.
The so-called artificial intelligence (AI), often referred to as machine learning (ML) or deep learning (DL). It uses a technology called Neural Network, which has a profound impact on the computing field.
It allows people to use real-world data to train computers so they can find the appropriate patterns themselves.
Take the self-driving car as an example. This technology will make the driving computer aware of whether there is a pedestrian in front. In the image field, it can find suitable exposure parameters for the photo.
It is reported that Google will provide a hardware module for the TPU to facilitate users to access the PCIe slot on the motherboard, or to connect to a USB peripheral port. Interested friends can apply online (portal).
It should be noted that although Google's TPU chip is running out of the data center, there is still a long way to go before it plays a role in mobile devices such as mobile phones.