Google reveal slices of artificial intelligence for machine learning

This theme Google reveal slices of artificial intelligence for machine learning appeared on Engadget.

Two years ago, the company revealed Google from processing units Tensor or segments of TPU, which is slices of specialized work in data centers of the company and lead light work of the tasks of artificial intelligence. Now, the company is transferring its experience in the field of artificial intelligence to the cloud, where it detects the sliced Edge TPU New, which is about speeding artificial intelligence small implements machine learning tasks on devices the Internet of things.

It is designed to Edge TPU to do what is known as “inference”, it is considered part of machine learning where the algorithm performs the tasks that were trained to actually doing it, such as recognition of objects in images. And improved devices TPU-based servers of Google training on this process, while the fillet Edge TPU New will had to be imputed.

The allocation of this new slide to be used in projects, business, not for smartphones, they are you say quality control in the factories automatically for example. The value of this type of functions on the devices is considered better than use of devices that send data across the Internet to communicate. Learning is the automated on the devices more secure in general, as it provides results faster.

I don’t think Google is the first from the manufacture of strips of artificial intelligence at all, has each of Qualcomm and arm وMediaTek industry accelerators, the artificial intelligence of their own. While the graphics processing units created by Nvidia dominate a particular market algorithms training.

However, what belongs to the Google company from its competitors don’t talk in the group of artificial intelligence to the full. The client can store its data on Google Cloud and the training algorithms using segments of TPU, and then perform inference on the device by using the fillet Edge TPU the new. It is likely that is a program for automated using TensorFlow, which is under the Encrypt is created and run by Google.

If this type of vertical integration has clear benefits, it can buy Google to make sure that all of these different parts are talking to each other efficiently and smoothly as possible, making it easier for workers to stay in the ecosystem of the company.

The head of Google Cloud of the Internet of things Injong Rhee as a new slide as “slices ASIC designed for a specific purpose such as running models TensorFlow Lite ML on the edge”.

It is interesting that Google is also working on providing a sliced Edge TPU package development tools, which will allow the customer to choose the potential of the hardware and its suitability for their products. This includes the package system (SOM) which contains the slices Edge TPU, CPU NXP, and Microchip safe and Wi-Fi. Where will be able to connect to the computer or server via USB or PCI Express. Don’t expect development tools this only with the demo version, it will be for some clients to apply for it.


This theme Google reveal slices of artificial intelligence for machine learning appeared on Engadget.

Leave a Reply

Your email address will not be published. Required fields are marked *