Microsoft Democratizing AI with Cognitive Toolkit Release

Rich OrdIT Management, ITProNews

Share this Post

Microsoft is changing folks, it's no longer the hated technology company that hoards power and technology. Today, Microsoft released to developers an updated version of the Microsoft Cognitive Toolkit that uses deep learning so that computers, using huge data sets, can learn on their own.

For instance, developers could feed CPUs and NVIDIA® GPUs millions of images of vegetables and it would learn over time which are cucumbers, no matter how distorted and different they appear. It will match was is similar and over time become very good add it. This matching and learning technology is applicable to an infinite number of software solutions.

It's free, easy-to-use and open-sourced, that Microsoft says trains deep learning algorithms to learn like the human brain. In fact, it's helping to change the world while simultaneously changing Microsoft.

"This is an example of democratizing AI using Microsoft Cognitive Toolkit," says Xuedong Huang, who is Microsoft's Chief Speech Scientist.

Microsoft originally created the Toolkit for internal use. "We’ve taken it from a research tool to something that works in a production setting," noted Frank Seide, a principal researcher at Microsoft Artificial Intelligence and Research and a key architect of Microsoft Cognitive Toolkit.

The current version of the toolkit can be downloaded via GitHub with an open source license. It includes new functionality letting developers use Python or C++ programming languages and allows researchers to do a type of artificial intelligence work called reinforcement learning.

The latest version is also much faster when adding big datasets from multiple computers, which is absolutely necessary in implementing deep learning across multiple GPUs. This allows developers to create smart AI enabled consumer products and enables manufacturers to connect more smart devices, empowering the IoT revolution.

Deep learning, according to Microsoft, is an AI technique where large quantities of data, known as training sets, literally teach computer systems to recognize patters from huge quanities of images, sounds or other data.


Just last week, Microsoft announced an historic voice recognition breakthrough, reaching virtual parity with human speech. Microsoft's AI team credited Microsoft Cognitive Toolkit speed improvement in allowing them to reach this level of performance so soon.

Rich Ord