Microsoft open sources its complex artificial intelligence program. The company started offering the AI structure it utilizes to help speech identification in the Cortana VA or Skype Translate programs. This signifies that people from all over the globe are now allowed to view, change and apply Microsoft’s source code into their own application.
The platform, known as CNTK, is relying on a segment of AI named deep learning, that looks to help devices do simple things like identifying pics or vids and understanding human conversations by resembling the structure or features of our own mind.
Tech leaders like Microsoft, Facebook or Google have spent a lot of effort in studying deep learning for several years, even hiring many specialists who initially developed this complex domain. Now, similar to scientists who post their analysis to be critiqued or developed by other scientists, these organizations are launching their own deep learning programs following the same path.
In 2015, Google shared its AI engine, called TensorFlow for open sourcing, which the search giant uses for a large part of its proprietary apps, such as voice recognition for Android or even its leading search engine.
Right after it, Facebook shared the models for particularized hardware developed to work with the newest AI programs, while Baidu, China’s most important search engine, open sourced the AI training system.
Microsoft actually launched CNTK, a system that is similar to the Google TensorFlow, several weeks before the search engine shared its own structure, but the source code was limited to non-commercial utilization.
Now, the company is allowing anyone, including organizations, to use its CNTK system for anything they want. Microsoft wants this move to be used, not only for educational purposes, but for also professional artificial intelligence or deep learning organizations, say the company’s researchers.
CNTK offers a big benefit in comparison to TensorFlow for individuals outside of academic institutions, since it offers the processing power of numerous web servers at one time. That is important because it is unusual that just one computer is good enough to deal with a real AI program, like speech identification on an application used by many customers.
On the inside, the search engine likely utilizes TensorFlow on a large number of web servers simultaneously. But the edition Google launched to the community, its experts say, cannot be used like this.
Actually, various deep learning systems besides CNTK operate across several web servers, even if it is possible to work with other free programs, like Torch that is utilized by Google and Facebook.
Image source: Paulgreenberg