Microsoft Open Sources Artificial Intelligence Toolkit

• By Angela Moscaritolo 
•  
• Microsoft is following Google's lead and making its deep learning tools a whole lot more accessible.
  The software giant has open sourced its Computational Network Toolkit, or CNTK, which the company says is "more efficient" than four other popular solutions used to create deep learning models for things like speech and image recognition — including Google's recently open-sourced TensorFlow. Now, the toolkit is available via GitHub for anyone who wants to use it, from deep learning start-ups to more established companies processing huge amounts of data in real time.
  "The CNTK toolkit is just insanely more efficient than anything we have ever seen," Microsoft's Chief Speech Scientist, Xuedong Huang, said in a statement, adding that it has the power to "drive artificial intelligence breakthroughs."
  
  
  Huang and his team developed the toolkit out of necessity: They wanted to improve how well computers can understand speech, but all the tools they had were slowing them down. So they built their own.
  
  
  
  Internally, Microsoft is using its CNTK on a set of computers that use graphics processing units (GPUs). They were designed for computer graphics but are also "ideal" for processing the algorithms that are leading to "major advances in technology that can speak, hear and understand speech, and recognize images and movements," the company said.
  The toolkit can scale across more GPU-based machines than other publicly available solutions, making it useful for those with the resources to create their own large cluster of GPU-based computers for major experiments and calculations, as well as researchers on more limited budgets, Microsoft added.
  Meanwhile, the field of deep learning has "exploded" in recent years as more and more researchers have started running machine learning algorithms using deep neural networks, which are designed to mimic the way the human brain works, Microsoft said. Redmond's researchers have used this approach to create systems that can do everything from translate conversations to identify the objects in a photograph or video, and answer questions about images.