Chinese search-and-everything-else web giant Baidu has joined Google and Facebook in open-sourcing its artificial intelligence (AI) code in a bid to become a standard in an increasingly important market.
The CTC part stands for "connectionist temporal classification." This combines different neural network designs to process data that is not perfectly aligned. In other words, making sense of complex patterns. The approach has proved invaluable in speech recognition.
Baidu Research built its system on top of CTC to improve its speech recognition products. Speech recognition is a critical component of business in China, thanks to its tonal language and the fact that its written form is hard to work with digitally.
Baidu claims that its Warp version requires far less memory and can be tens to hundreds of times faster than just CTC.
As to why it is open sourcing the software, it says that it wants to "make end-to-end deep learning easier and faster so researchers can make more rapid progress" and that "previous code for training end-to-end networks for sequences has been too slow." It says it wants to "start contributing to the machine learning community by sharing an important piece of code that we created."
The reality of course is that due to Google's decision to open source its TensorFlow software in November, and Facebook's decision to open source its Big Sur hardware specs in December, Baidu is worried that its approach will be left behind as developers start to learn and understand easier-to-access technologies.
Artificial intelligence is a critical component of online businesses and can provide a huge business advantage. AI is complex however, and success often rests in the optimal combination of software and hardware. Increasingly, if your code isn't out there and readily available, you risk becoming obsolete. ®