Continuing its open-sourcing spree, Microsoft made its artificial intelligence available for all developers. Anyone in the world is now free to view, use, and modify Microsoft’s code to their heart’s content.
Open-sourcing is one of the initiatives Microsoft has shown great interest during recent years, and now it was the company’s AI framework turn to become available to the public. Microsoft uses CNTK – standing for Computational Network Toolkit – to power speech recognition in its Skype Translate app and Cortana digital assistant.
At the foundation of Microsoft’s framework stands deep learning, one of the most interesting branches of artificial intelligence. It’s deep learning that allows machines to do things like understand human speech or recognize photos and videos, as it learns to mimic the functions of the human brain.
In recent years, deep learning research has become the focus of many tech giants to the likes of Facebook, Google, and Microsoft, all of whom invested heavily in the field. Some went as far as hiring the academics who pioneered this innovative field.
In a manner similar to scientists publishing their work to be critiqued and added to by fellow researchers, these companies found they advance more if they release their deep learning software.
Google, for example, open-sourced TensorFlow, the AI engine the company uses for voice recognition in Android and even for its renowned search engine. Last year, Facebook followed suit by open-sourcing its designs for custom hardware that run the social networks’ latest AI algorithms.
Microsoft’s CNTK, which resembles the functions of Google’s TensorFlow, was actually released back in April, long before Google thought of releasing its own framework; Microsoft, however, restricted the code to non-commercial use.
But Microsoft has decided to invite anyone – corporations and individual developers alike – to use CNTK for whatever they want. Xuedong Huang, Microsoft’s chief speech scientist, stated the company wants deep learning and commercial AI companies to also benefit.
CNTK is actually a lot more desirable than TensorFlow for people outside of academia, as it has the ability to power more servers at the same time. This aspect is important because it’s impossible for a single computer unit to handle a real-world AI application.
While CNTK might prove more efficient than other deep learning tools, it also has a downside: the framework only supports its own custom language and C++, which can create some difficulties for developers who use other languages.
Image Source: Popsci