AI News | January 2018

  • Overview
  • Resources
  • Transcript

WELCOME TO 2018! In this episode you'll see the world’s first neural network processor, get a look at a new deep learning edge solution, and we explore Unity Technologies’ ML-Agents exclusively on Intel Architecture!

Hi, I'm David Shaw. And welcome to AI News. A very Happy New Year to all, and welcome to 2018. 

In this episode, we discover the world's first Neural Network Processor. Take a look at a new Deep Learning Edge Solution, and explore Unity Technologies ML Agents exclusively on Intel Architecture. Don't forget to check out the links provided to learn more. 

Intel is launching its first family of processors designed from the ground up for AI, the Intel Nervana Neural Network Processor. It's a purpose built architecture for deep learning. The goal is to provide the needed flexibility to support all deep learning primitives while making core hardware components as efficient as possible. 

This processor also has a new memory architecture, and is designed to maximize the utilization of silicon computation. In addition, it can achieve a new level scalability for AI models. Discover a deep learning edge solution that pairs the Raspberry Pi 3 single board computer with the Intel Movidius Neural Compute Stick. 

This article will show you how to increase the Pi 3 performance for computing inference using Google's mobile nets. Learn how to install and run Unity Technologies Machine Learning Agents in CPU only environments from this article. 

It will teach you how to train ANS to run the ML Agents Balance Balls example on Windows, perform a Tensorflow CMake Build on Windows optimized for Intel AVX 2, and create a simple Amazon Web Services Ubuntu Machine Image Environment from scratch. 

Additionally, it will show you how to build a Headless Version of Balance Balls for Linux. 

Thanks for watching. Don't forget to like this video and subscribe to the Intel Software YouTube Channel. 

Follow the links provided to learn more.