(Original title: When large models meet art JD Cloud demonstrates the value of the AIGC large-scale model industry)
Many viewers were pleasantly surprised at the “Cézanne·Four Seasons: Large-scale Immersive Art Exhibition” currently being exhibited at Today Art Museum. Through the technical capabilities of JD Cloud, the application of visual content generation large models allows each visitor to generate an exclusive “Cézanne·Four Seasons” oil painting, AIGC’s cross-border art, brings audiences a deeper interactive experience.
On May 11, the “JD Cloud AIGC Technology Exchange Media Private Meeting” was held at Today Art Museum. Dr. He Xiaodong, Vice President of JD.com and head of JD Technology‘s Intelligent Service and Product Department, and Liu Shuhui, Deputy Director of Today Art Museum, discussed the hot topic at the moment. Further exchanges and discussions on the cross-border integration of large-scale model technology and art.
Dr. He Xiaodong and Deputy Director Liu Shuhui communicated on the spot
Paul Cézanne is known as the “Father of Modern Art”. This Cezanne exhibition has set up 3 interactive areas, presenting 17 dynamic works, creating an interactive and immersive exhibition space, bringing Cezanne’s works to life to the Chinese audience. At the end of the exhibition, each audience can generate a unique Cézanne landscape painting through JD Cloud’s AIGC capability. As long as the audience enters their favorite season, mood and understanding of Cezanne’s painting style, AI can imitate Cézanne. Shang’s painting style customizes a painting for the audience.
The media interacted in the JD Cloud AIGC experience area
Regarding the technical capabilities applied in this AIGC cooperation, Dr. He Xiaodong said: “The visual content generation large-scale model technology used in this exhibition is derived from the multi-modal large-scale model base developed by JD. As a leading technology and service provider in China, JD.com The cloud research and development team has been focusing on the research and development of multi-modal large models for a long time, covering the fields of copywriting generation, AI poetry writing, digital virtual human, etc. Before that, it also cooperated with Professor Qiu Zhijie of the Central Academy of Fine Arts to explore the application of AI drawing and AI writing calligraphy. .”
The R&D team of JD Cloud has also developed a set of special scene adaptation technology, which enables the multi-modal large model to be quickly and accurately applied to different scenes, achieving a high degree of flexibility and controllability. The cooperation between JD Cloud and Cézanne·Four Seasons large-scale immersive art exhibition not only demonstrates an innovative implementation of JD’s visual content generation model in the field of art, but also demonstrates the value of this technology in the field of marketing. Utilizing AIGC’s ability to imitate the language, calligraphy, and painting styles of exhibiting artists to generate customized artworks for the audience not only enhances the fun and sense of participation in the exhibition, but also the content generated by users spreads through social media, making the exhibition viral. dissemination and promotion.
User Generated Artwork
With the upsurge of AIGC and large-scale model technology such as ChatGPT sweeping the world, major manufacturers have deployed one after another, and the application of specific industries and industries has attracted more attention. Liu Shuhui said that Today Art Museum is the first private non-profit public welfare art museum in China , focuses on promoting domestic and foreign art exchanges, and is committed to providing an open communication platform for artists, the public and interdisciplinary institutions. The cooperation between JD Cloud and Today Art Museum is an innovative attempt.
Deputy curator Liu Shuhui quoted Flaubert’s famous saying: “Art and science parted at the foot of the mountain, and will meet again at the top of the mountain.” Advanced technological capabilities such as advanced technologies allow more members of the public to get closer to art, which is also consistent with the original intention of Today Art Museum.
Regarding the integration of AIGC and art, Dr. He Xiaodong made a metaphor: “Although humans have invented the camera and photography technology is constantly iterating, the creativity and ideas of the photographer are still inseparable from the exquisite photography works. Today’s AIGC The same is true, technology does not replace people’s ideas, it helps people express their ideas better.” Liu Shuhui, deputy curator of this point of view, also agrees, “In the field of contemporary art, it is still difficult for AI to do so. Because contemporary art pays more attention to creativity, it does not need to be related to design. Design is output according to people’s needs, and AI can play a very good auxiliary role.” At the same time, she expressed her appreciation for the “blocking” generated by JD Cloud’s AIGC capabilities. Still painting” also expressed full recognition.
Not limited to images and art, in the field of large models, JD Cloud’s Yanxi artificial intelligence platform has relied on its own accumulation of more than 10 years of experience in intelligent dialogue, coupled with years of practice and experience in JD’s retail, logistics, finance, health and other businesses. Tens of millions of intelligent interactions are performed every day, focusing on the layout and implementation of AIGC in multiple aspects such as text, voice and digital human generation.
JD Yunyanxi digital human broadcasts live on JD APP
JD.com’s product copy generation technology has been found in the JD.com app’s Haohuo channel, Jingxiaozhi marketing version, Yanxi virtual live broadcast, live version, and JD.com APP-Matching Shopping, and other scenarios. So far, more than 3 billion words of copywriting have been generated; the digital human field , currently possesses full-stack self-developed 2D twin, 3D realistic and 3D cartoon digital human synthesis technologies. Among them, the 2D twin digital human pose synthesis technology under simple shooting conditions is becoming more and more mature, and the training data relied on is compressed from 30 minutes to 5 minutes or so.
In February this year, JD Cloud’s Yanxi AI application platform integrated past industrial practices and technology accumulation, launched the industrial version of ChatGPT, and announced its application roadmap “125” plan, including one platform, two fields, and five applications .
Yanxi large-scale pre-trained language model “125” roadmap
JD Cloud, as the core platform for JD to provide external technology and services, has been committed to integrating technical capabilities and industries. Not limited to art, in the future, JD Cloud will also make more explorations in the fields of cultural creation and cultural tourism, how to use AIGC capabilities to better promote culture and inherit art.