Shulou(Shulou.com)11/24 Report--

Recently, China Multimedia Conference (ChinaMM2023) was held in Kunming, Yunnan Province, focusing on the development of technology and applications in the field of multimedia, bringing together professionals in the field of multimedia. Yu Bing, senior vice president and head of R & D line of Kuaishou, was invited to deliver a keynote speech on "Frontier Progress and thinking of Intelligent Media Technology in the AGI era", combining Kuaishou's innovative practice in the field of multimedia technology, sharing Kuaishou's technological frontier progress and related thinking.

Explore business development strategy in stages and continuously promote technology energy efficiency optimization

Since its establishment in 2011, Kuaishou has developed into a national-level short video live digital community, maintaining healthy growth in users, content, commerce and other aspects. The vigorous development of Kuaishou community ecology is inseparable from Kuaishou's long-term huge investment, with large-scale self-research technology to support business development.

Reviewing Kuaishou's past experience, Yu Bing said that business development will go through many stages, such as entry, growth, maturity, and so on. The growth period of innovation pays more attention to continuous investment, exploring new products and new cycles, while in the mature stage, we need to focus on optimizing efficiency and taking into account experience, which ensures not only the competitiveness of product experience, but also the steady growth of business.

Yu Bing takes the fast hand audio and video business as an example. at present, the business has entered a mature stage, and the focus of the R & D strategy has shifted to improving efficiency, optimizing costs, and pursuing the improvement of computing power, storage and network use efficiency per unit cost. Based on a series of core technologies such as self-developed video quality assessment KVQ, video enhancement repair KEP and KRP, and video compression coding algorithm KVC, Kuaishou forms a data-driven closed loop for video processing to compress video at the same time.

In 2022, Kuaishou toB business StreamLake launched the first self-developed intelligent video processing chip SL200, which integrates Kuaishou's leading video codec, image processing, AI reasoning, content adaptive coding and other key technologies, with the characteristics of high quality, high density and intelligence, and is in a leading position in the industry. In the MSU2022 World Encoder Competition held in July this year, SL200 won 16 of the 24 indicators of 4K and 1080p. At this conference, SL200 also won the China Multimedia Enterprise Innovation Technology Award. At present, SL200 chip has been fully used in Kuaishou's live broadcast and short video services, realizing mature technology to empower the industry through StreamLake.

In the field of digital human, Kuaishou has also made many leading achievements. Kuaishou full-link self-developed digital human core technologies such as light field scanning and reconstruction, hyperrealistic portrait modeling, intelligent binding, motion capture and drive, and physical simulation, and launched two solutions: "Kuaishou Virtual Broadcasting Assistant (KVS)" and "Kuaishou Smart broadcast". Combined with the unique advantages of the content platform in the field of brand marketing, and empowering in many dimensions, such as art, technology, operation, marketing, etc., we have created a number of well-known IP such as Guan Xiaofang, the virtual anchor with millions of fans, and Milk, the first virtual employee of Mengniu Group. In the future, combined with intelligent arrangement, image / video generation, LLM and other AI capabilities, Kuaishou digital human technology will continue to be widely used in literature and travel, education, games, live e-commerce and other fields, helping customers to create high-quality and low-cost 3D and 2D digital people.

Deeply ploughing the field of AI large model, expanding the ecology of industry and research to realize two-way empowerment

With the deep combination of multimedia and AI technology, Kuaishou continues to cultivate AI technology, promote the development of product form and user experience, and explore the second curve of short video service. Yu Bing believes that the AI large model, as the most important revolutionary technology, has entered an explosive period and is expected to open the AGI era, and multimodal content generation and understanding is its core competence.

At present, because of its outstanding ability to generate and understand many kinds of content, such as text, code, image and video, the multimodal AI model is expected to open up a new technical perspective for the full link of video production, understanding, distribution and consumption, break through the technical ceiling of traditional audio and video coding, break through the traditional idea of searching and calculating method based on user behavior, and drive video content creation from PGC and UGC into the AIGC era. Create AI model-driven video content creation tools to stimulate creators' creative space and produce high-quality video content with high efficiency and low cost.

While short video and live broadcast are the most typical multimodal media, Kuaishou also grasps the platform gene and invests heavily in the field of AI large model to explore technological breakthroughs in an all-round way. Yu Bing said that at present, the layout system of Kuaishou's AI model is divided into three levels: based on the "big infrastructure" with high performance, high concurrency and high computing power, build Kuaishou's multimodal AI "big model", and then create "big applications" in the fields of Sou Guangtu, content creation, user growth, and R & D efficiency.

For example, in the field of search and push, Kuaishou search and push algorithm has reached the international leading level, and the related achievements have won honors such as CIKM Best Paper and SIGIR Best Paper-Honorable Mention in the field of information retrieval and data mining. at present, Kuaishou breaks through the traditional technical ideas based on user behavior, explores deeper model networks, develops recommendation models, and uses content generation and understanding to explore new paths of deep user interest.

At the same time, with the blessing of the multimodal AI model, AI technology and tools can empower film and television creators in all directions, helping them stimulate creativity, improve efficiency and content quality in the creation, shooting, and post-production stages. The production cycle can also be greatly improved, and blockbusters that have taken several years to shoot are expected to be completed in a few months.

Technological development from germination to maturity depends not only on the self-research of enterprises, but also on the talent empowerment of scientific research institutions in colleges and universities. Prior to this, Kuaishou successively established joint scientific research institutions with Tsinghua University, Beijing Research Institute of the University of Science and Technology of China, and Renmin University of China, and established scientific research cooperation with the world's top universities and laboratories to jointly explore cutting-edge technologies in the fields of audio and video, multimedia and AI, and jointly train scientific research personnel.

"academia has top technology and excellent scientific research talents, while industry has real application scenarios, with the advantages of massive data and computing power. The value of the two will be multiplied by deep cooperation and two-way empowerment." Yu Bing said: on the one hand, technological breakthroughs will be used on a large scale in Internet business, resulting in huge economic and social benefits; on the other hand, the real scene of the Internet, huge amounts of data, and powerful computing power can also help scientific research and technology to continue to iterate. In the future, Kuaishou will continue to promote the ecological improvement of industry-university-research, open scenes, data and computing power to academia, jointly explore new technologies of intelligent media in the AGI era, and empower industry innovation and development with technology.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.