A few days ago, Intel CEO Brian Krzanich spoke at the Wall Street Journal’s D. Live conference about their venture into the world of cognitive and artificial intelligence (AI) technology. He described these two areas of computing to be “transformative to the industry and world,” that is – a market projected to reach $46 billion by 2020.

In their effort to bring these two technologies closer to human understanding, Krzanich announced that Intel will soon come up with its first silicon specifically designed for neural network processing, the Intel Nervana Neural Network Processor (NNP).

The Intel CEO revealed that they plan to ship NNP before the year ends.

Revolutionizing AI computing

Intel has also been in close collaboration with Facebook in the development of Nervana. According to Naveen Rao, former CEO and co-founder of Nervana and now Intel’s VP and GM of the AI Products Group, Facebook’s expertise in the field of Artificial Intelligence and machine learning is instrumental in the development of these new technologies. Rao suggested that they expect Facebook to be one of the first companies to adopt the new Intel hardware.

Social media is just one of the industries Intel hopes to revolutionize in terms of AI computing using its Nervana NNP. But they have bigger hopes to achieve: important research related to health care, accelerated learning on autonomous driving, and understanding and predicting weather to name a few.

Krzanich also revealed that they have more generations of Intel Nervana NNP coming up in the future in their hopes of delivering high performance and reaching new heights for AI models. The Intel CEO also shared their breakthroughs in technology research on neuromorphic and quantum computing.

Research breakthroughs

Back in the 1980s, American scientist and engineer Carver Mead came up with the term “neuromorphic processors,” describing how a computer chip could closely resemble that of human brains.

His studies became the foundation of artificial intelligence understanding how the human brain works.

In connection with this, Intel has recently unveiled its first self-learning neuromorphic test chip. As Krzanich shared, this chip will gather the data and will the ability for self-learning, it can understand and recognize the situation, becoming more intelligent in the process.

Potentials of this technology are limitless: from preventing cyber-hacking to understanding human anatomy.

Meanwhile, quantum computing enables computers to confront problems regular computers can’t handle. Problems which can be solved by quantum computing technology can range from weather simulation, understanding chemistry and other fields of science and possibly discovering new medicines.

From Lake Crest to NNP

According to Rao, the Intel Nervana NNP was formerly known as Lake Crest, which was teased early this year. Like Nervana, Lake Crest was aimed at deep neural network designed to deliver an unprecedented amount of compute density and raw power than modern GPUs, as posted by tech website WCCFTech.com.

Based on initial details about the hardware specification of Lake Crest, it was designed to have a total of 32GB of HBM2 memory in the form of four (4) 8-Hi stacks. This is said to deliver a total of 1 TB/s memory speeds at 2GHz clock speed and memory access speed of 8Tb/s. Intel has yet to release the hardware specifications of Nervana NNP.
© ALL RIGHTS RESERVED