AI in Baidu Developers Conference held today, Intel vice president and general manager of Artificial Intelligence Products Division Naveen Rao announced CONTROL ENGINEERING China Copyright , Intel is working with Intel to develop Baidu Nervana neural network training process is (NNP-T). This cooperation includes new customization accelerator CONTROL ENGINEERING China Copyright , in order to achieve the purpose of speed training depth learning model.
Naveen Rao said: “The next few years, the complexity and the need for large-scale computing depth study of AI model of the explosive growth of Intel and Baidu will continue more than a decade of cooperation and focus on the joint design and development of new. the hardware and supporting software, so as to continuously move to ‘AI 2.0’ the new frontier. “AI is not a single workload, but a broad, powerful capabilities can enhance the performance of all applications, whether those applications are running on the phone or within large-scale data centers. However, mobile phones, data centers and all facilities between the two have different requirements for performance and power consumption, and therefore a single AI hardware can not meet all needs. Intel provided in artificial intelligence superior hardware selection, and release to maximize the performance of hardware through software to help customers no matter how complex or located where can freely run AI application data. The NNP-T is a class of highly efficient deep learning system newly developed hardware, training can accelerate large-scale dispersion. Work closely with Intel to ensure that Baidu’s development department always keep up with customer demand for the latest hardware training. , Intel has been optimized Baidu will fly paddle (PaddlePaddle *) deep learning framework for the Intel Xeon processor can be extended from 2016. Today, by fly paddle as Baidu optimization NNP-T, the two sides can provide more hardware options for data scientists. At the same time, Intel also to further enhance the performance of these solutions through more AI technology. For example, CONTROL ENGINEERING China Copyright , with Intel Aoteng data center-class higher memory lasting memory providedPerformance CONTROL ENGINEERING China Copyright , Baidu can * (information flow) provides services through its Feed Stream to millions of users personalized mobile content, and a more efficient customer experience through Baidu AI recommendation engine. In addition, in view of the security of user data is extremely important for Intel and Baidu also committed to build together * Intel-based MesaTEE software protection extension (SGX) technology – Memory Security lets a Service (FaaS) computing framework. More: Intel AI | outside the CPU or GPU: enterprise-scale artificial intelligence Why do we need a more comprehensive approach (Naveen Rao signed article) | Intel AI Artificial Intelligence on Baidu Developers Conference: AI camera, FPGA-based workload and speed up services, as well as optimized Intel Xeon depth study of scalable processor