According to Computerworld website, Intel is worried about the next generation of its accelerators computing Xeon Phi , it wants to make suitable for use in the training system.
When the corresponding optimization will be carried out is not specified, but the next generation of accelerators computing Xeon Phi will be released in 2018, so Intel has time . The processor giant is hoping that its x86-based computing accelerators in some problems of "machine learning" will be able to beat graphics processors competitor. Support OmniPath interface for high-speed transmission of information on this issue will play an important role.
In addition to Google TensorFlow platform, Intel will be able to provide support solutions Caffe open platform. This will attract the attention of the maximum number of software developers to learning systems. The latter can be used for pattern recognition, or analytical applications, making it easier to work exchange investors or medical specialists.