20-F
BAIDU, INC. filed this Form 20-F on 03/31/2017
Entire Document
 


Table of Contents

Targetizement Technology

Our targetizement technology matches our customers’ promotional links with their targeted internet users. Our automatic algorithm can analyze a user’s audience attributes based on his or her past search experience and display promotional links that the user may be interested in viewing.

Large-Scale Systems and Technologies

Large Size Cluster Management. In order to provide highly efficient and stable search services, we have developed an automated management platform for large size clusters. The platform enables us to intelligently manage and allocate resources and automatically debug and relocate services, thereby allowing tens of thousands of different source requests on the Baidu search engine and other non-search business to function stably across multiple internet data centers and thousands of servers.

Storage. We have developed an efficient, distributed and structured storage system to support our search services. Our storage system supports PB-level holistic, sequential data storage, and ten thousand times of real-time processing per second per device. Our storage system also has dynamic data attribute addition and subtraction function and historical data management capability.

Distributed Computing System. We have developed our proxy computing system, a comprehensive set of ultra-large scale distributed computer system, to increase the utility rate of idle resources, providing a strong base support for our core operations. Our proxy computing system has realized various distributed computing software stacks, such as resource isolation, resource distribution, computing modeling and application framework, and supports commonly used computing modules such as MapReduce, Spark, Stream and WebService.

Indexing Technology. Our indexing technology supports billions of daily search requests on over tens of thousands of servers located across multiple internet data centers of different network operators. Through our indexing technology, we have been able to index over one hundred billion of web pages without utilizing additional resources and have improved the freshness of indexed information.

Artificial Intelligence (AI)

We have been investing in AI for many years and have been recognized as one of the leading AI innovators globally. AI enables computers to simulate the working mechanisms of human brains, and to learn and be trained with extremely complicated models. The core AI technology is AI supercomputer, which is a super-speed heterogeneous computing cluster designed for AI-based applications and has integrated heterogeneous computing server, GPU Box and FPGA, among others developed by us in house. Currently, our AI mainly comprises three types of technologies, namely, parameter, sample and feature training, computing capabilities (servers and GPU clusters) and big data (webpages, search data, image and video data and locating data). Integrating these three types of AI technologies, we are able to apply AI into areas such as natural language processing, speech recognition, image recognition processing, user portrait and other capabilities. Our Conversational Interfaces has been recognized by MIT Technology Review as one of the ten breakthrough technologies in 2016. Our AI has now been applied to Baidu Search, Duer, Autonomous Driving Car, Baidu Cloud and other products.

From September 2016, we open-sourced our AI platform “PaddlePaddle” to the developer community, providing access to Baidu technology in areas of voice and image recognition, natural language processing, and machine learning. PaddlePaddle is accessible to developers from all over the world and thus promotes the popular use of AI.

 

57