On January 29, Wu Tian, vice president of Baidu group, visited the “dialogue” program of China Central Television financial channel to talk about the technical driving force behind the outbreak of virtual digital human popularity and the increasingly broad application scenarios. Wu Tian said that the overall improvement and cross modal integration innovation of AI technologies such as computer vision, voice, natural language processing and knowledge atlas enable the virtual digital human generated and driven by AI to truly play its value and further expand the application scenarios. Technological innovation and industrial demand promote each other, forming a resultant force of spiral upward development.
Virtual human is not a new thing. In the past few years, driven by technology and demand, two-dimensional style and super realistic virtual human have emerged one after another. On the one hand, the emerging media represented by short video and live broadcast have provided more application scenarios and commercialization paths; On the other hand, virtual people are used in finance, operators, digital exhibition halls and other fields, emerging service-oriented roles such as bank digital employees, virtual idols and virtual anchors, and showing commercial value.
But at present, the development of virtual human is still in a relatively early stage, especially the production cost is very high. To create a high-precision and high fidelity non-specific virtual human, it often needs millions or even tens of millions of capital investment, while the specific virtual human needs to spend more and the production cycle is relatively long.
Cost reduction has become one of the keys to the development of virtual digital human. Wu Tian said that Baidu hopes to shorten the production cycle of virtual human to the hour level, reduce production costs and drive the generation of content through the leading full stack AI capability. To this end, Baidu has released an intelligent digital human platform integrating digital human production, content creation and business configuration services, Baidu Intelligent Cloud Xi Ling, which provides a convenient and simple operation platform based on four technical engines, namely, portrait driven, intelligent dialogue, voice interaction and intelligent recommendation, to help enterprises quickly customize 2D cartoons, real images and 3D cartoons at low cost. The super realistic digital portrait realizes the whole process business configuration of dialogue service and content creation, which greatly reduces the difficulty of digital human development.
The Shanghai Pudong Development Bank Co.Ltd(600000) financial officer digital person “Xiaopu”, China United Network Communications Limited(600050) digital person customer service “xiaou”, baidu app spokesperson “Gong Jun digital person”, baidu group digital person “higaga”, Mars rover digital person “zhurong”, AI sign language anchor released at create2021 Baidu AI developer conference, etc. are all based on “Xiling”.
In order to provide services for human beings and communicate naturally with human beings, virtual digital human needs the support of powerful knowledge system and dialogue ability behind it. The 550 billion knowledge accumulated by Baidu knowledge map helps digital people have “their own” experience and understanding to think. Based on the 10 billion parameter dialogue model Plato released by Baidu last year, virtual people can communicate with users with information, interest and emotion.
As the representative of new technology products, virtual digital human has not only ushered in the explosion of market heat and demand, but also triggered more humanistic thinking and discussion. Wu Tian believes that the underlying values of artificial intelligence system are very important. While technology develops, it needs to be standardized, and technology should be applied to places with temperature and better. For example, Baidu and China Central Television, based on a small amount of corpus, used AI technology to reproduce the voice of the dead old platoon leader, which comforted the 91 year old veterans’ thoughts of the platoon leader.
In the future, artificial intelligence and human beings will enter a stage of mutual cooperation and common development. Wu Tian believes that artificial intelligence can provide assistance and help to liberate human beings from boring and repetitive work and do more creative work.
Baidu adheres to long-term high-intensity R & D investment, long-term technology accumulation and innovation, and has developed into a leading AI company. In the future, Baidu will continue to adhere to the investment in “two bases and one bottom”, make hard efforts in basic research, basic technology and bottom innovation, and actively strive for the goal of training 5 million AI talents in the next five years, promote artificial intelligence to create more value and help China achieve high-level scientific and technological self-reliance and self-improvement.
