infoTECH Feature

January 27, 2016

Microsoft Has Open Sourced its Computational Network Toolkit

The quest to have a technology listen and talk like human being has been a long one, but we’re getting close. Artificial Intelligence or AI has evolved into a standard offering on the devices we use every day. The most recognizable example is the virtual/personal assistants, such as Apple’s Siri and Microsoft’s (News - Alert) Cortana, currently deployed in smartphones and computers. Now, the need for more development and innovation has led Microsoft to open source its Computational Network Toolkit, or CNTK.

The open source and API culture in IT has brought in millions of developers around the world to start creating solutions that are unique, inventive and in many instances far from what the original creators intended.

The CNTK itself was created by Microsoft's Chief Speech Scientist, Xuedong Huang out of necessity for improving computers ability to understand speech. The available solutions didn’t quiet meet their needs, so they came up with CNTK.

After several tests, the team concluded it was more efficient than other computational toolkits that were being used by developers to create learning models. The communication capabilities of CNTK had better image and speech recognition.

As Huang said, “The CNTK toolkit is just insanely more efficient than anything we have ever seen.” And in the field of deep learning, which is very important for making AI better, efficiency and performance are critical.

Microsoft is using the tool on computers that use graphics processing units or GPUs, which have become the ideal technology to process complex algorithms used in speech recognition, listening, understanding and recognizing images and movements.

The improvements Huang’s team have made has resulted in systems that can recognize and translate more accurately, as well as some that were able to recognize images and answer questions about the same images.

In resolving the problems it was facing, Huang also acknowledged there were many researchers in the same field that could benefit from this technology. Initially they decided to make the tool available to academic researchers, via Codeplex and under restricted open-source license, that was in April of 2015. However, today they have extended the availability of CNTK to everyone.

If you are a developer or researcher looking to be part of the next AI evolution, and you need a tool with proven capabilities, you can download the CNTK toolkit from GitHub.

“With CNTK, they can actually join us to drive artificial intelligence breakthroughs,” Huang said.




Edited by Kyle Piscioniere
FOLLOW US

Subscribe to InfoTECH Spotlight eNews

InfoTECH Spotlight eNews delivers the latest news impacting technology in the IT industry each week. Sign up to receive FREE breaking news today!
FREE eNewsletter

infoTECH Whitepapers