Microsoft is making the tools behind applications like Cortana available on code repository GitHub, opening the doors to new open source machine learning projects based on its software programmes.
Redmond originally created the toolkit, known as Computational Network Toolkit (CNTK), out of necessity to help its developers make faster improvements to how well computers can understand speech.
But in a blog post, Microsoft explained that CNTK has proved to be “more efficient” than the four notable computational toolkits – namely Theano, TensorFlow, Torch7, and Caffe – that developers were using to create deep learning applications.
“The CNTK toolkit is just insanely more efficient than anything we have ever seen,” said Xuedong Huang, chief speech scientist at Microsoft.
This has allowed Microsoft’s researchers to create systems that can accurately recognise and translate conversations, as well as ones that can recognise images and even answer questions about them.
Huang said the swift gains his developers got from CNTK convinced Microsoft to make it available to everyone, adding that it wants to provide the same resources to other researchers who are making similar advances in deep learning.
CNTK is flexible enough that it can be used by anyone, from a researcher on a limited budget with a single computer, to teams with the budget to create their own large cluster of GPU-based computers.
It will be available via an open-source license on GitHub from Monday.
Machine learning applications have also been a major talking point in cloud and mobile technology in recent years.
Microsoft’s own personal assistant, Cortana, is getting cleverer at recognising users’ routines and communications where meetings are agreed.
Google has heavily invested in a number of machine learning projects, including Google Now, Google Translate and its much-discussed self-driving car project.
In December, Facebook also opened some of its own deep learning tools for use by developers.
