During the last few months, we have seen a vast number of projects in which artificial intelligence has become the protagonist. Software in this sense is advancing in great strides. And hardware must accompany movement always to offer what machines need to continue learning.
The Wall Street Journal now reports that Intel is working with Facebook and other companies on a new processor called Nervana. That is a neural network processor designed to accelerate the process of learning artificial intelligence. That will help computers recognize objects, patterns, speech and complex concepts.
This family of processors takes more than three years in manufacturing. And now it begins to come to light. Facebook has already started to analyze the processor more thoroughly and realize its capabilities. That is something fundamental to be able to recommend content to people based on their tastes, for example.
Intel Nervana, They comment in the official note:
The Intel Nervana NNP is an architecture designed for deep learning. The goal of this new structure is to provide the flexibility to support all deep learning primitives while making essential hardware components as efficient as possible. Improved memory management enables the chip to achieve high levels of utilization of the massive amount of calculations in each die. This translates into faster training time for deep learning models.
More tools for Artificial Intelligence from Microsoft
Nervana is able to take shortcuts to reduce the amount of processing power required. And the goal is to put it to work in tasks as complex as the mathematics and autonomous cars. Intel, in fact, also partnered with Waymo (of Google) with the aim of perfecting that last point.
The goal of Intel is to achieve 100 times greater performance in Artificial Intelligence processors by 2020. They want to remove the limitations imposed by existing hardware. That is not specific for AI, and they want to do it from scratch.
Source