Today, the State Internet Information Office released the announcement of the fourth batch of in-depth synthesis service algorithm filing information. It can be seen from the published list that a total of 266 algorithms have been listed, involving image generation, video generation, digital people, intelligent dialogue and other categories. Their applications cover education, games, e-commerce and other industries. Internet enterprises such as Alibaba, Tencent, Baidu, NetEase, mobile phone manufacturers such as vivo, Xiaomi, OPPO, and home appliance manufacturers such as Midea and TCL are all among them.
Article 19 of the Provisions on the Administration of Deep Integration of Internet Information Services clearly stipulates that deep integration service providers with the attribute of public opinion or the ability of social mobilization shall carry out the procedures of filing, changing and canceling the filing in accordance with the Provisions on the Administration of Recommendation of Algorithms for Internet Information Services. Supporters of deep synthesis service technology should refer to the procedures for filing, changing, and canceling records.
The Science and Technology Innovation Board Daily found that among well-known Internet enterprises, Alibaba, JD, Baidu and others have algorithms listed. The clothing upper body synthesis algorithm of Ali Damo Academy can integrate the clothing in the clothing image into the model's body based on the user's uploaded clothing reference image and model image, and generate corresponding advertising images for clothing products.
The image intelligent erasure algorithm of Alibaba's quark app is applied to image editing scenarios, which erases and fills the background of images uploaded by users, achieving the function of erasing handwritten, watermarked, or specified areas in the image.
JD.com has the Wenshengtu model algorithm and the Navigator intelligent resource generation algorithm in this announcement. The Navigator Intelligent Resource Bit Generation Algorithm generates marketing resource bit images that meet user needs based on text and image data input.
Baidu's facial fusion and attribute editing algorithm can generate image content that meets user needs based on user input image or video data, achieving functions such as facial fusion and attribute editing.
NetEase involves the NetEase Fuxi 3D scene reconstruction algorithm, NetEase foreign trade generation and synthesis algorithm, NetEase Youdao document question and answer algorithm, and NetEase Yuanqi generation and synthesis algorithm. Among them, NetEase Fuxi's 3D scene reconstruction algorithm generates corresponding 3D mesh models based on user input images or videos. The remaining three algorithms are mainly used to generate text content that meets user needs.
Tencent has a list of 5 algorithms, including QQ Intelligent Image Modeling Algorithm, Tencent Meeting Intelligent Dialogue Algorithm, Tiantian P-image Face Transformation Algorithm, Tencent Intelligent Shadow Painting Generation Algorithm, Tencent Digital Human Video Synthesis Algorithm, Tencent Cloud Intelligent Painting Image Background Replacement Algorithm, and Application Treasure Image Fusion Algorithm.
AI algorithm also covers Internet enterprises in vertical fields such as Tongcheng Travel, Vipshop, Tiger Teeth, Soul, Babytree and What's Worth Buying. Vipshop's virtual fitting algorithm generates two-dimensional model dressing renderings based on user input images or two-dimensional model data, combined with clothing images selected by users. It provides users with the function of selecting, changing, and viewing fitting effects in real-time online. Baobao Tree's maternal and child family domain large model generation and synthesis algorithm is applied in text generation scenarios. Based on the user's proposed pregnancy and child care questions, it generates relevant knowledge content such as pregnancy and child care, and outputs text content replies.
The education enterprise TAL Education, which founded the Xueersi brand, involves MathGPT mathematical big model algorithm, MathGPT big model learning assistant algorithm, and Study Companion learning assistant algorithm, mainly used to solve various mathematical problems.
Huawei Cloud has three digital human algorithms for filing and public announcement, namely MetaStudio Digital Human Visual Driver Algorithm, MetaStudio Clone Digital Human Driver Algorithm, and MetaStudio Digital Human Voice Driver Algorithm, which are used to generate digital human videos.
In addition, Shangtang's instant drawing content generation algorithm, AI code generation algorithm, and cloud driven algorithm are also among them. The publicly available list also includes large model algorithms of listed companies such as Starring Technology and Wanxing Technology.
The reporter from the Science and Technology Innovation Board Daily noticed that the multiple large model algorithms of home appliance companies such as TCL, Midea, and Haier are mainly used for text replies in human-computer conversations. For example, the TCL intelligent assistant algorithm is applied to intelligent dialogue scenarios, generating text replies that meet the user's needs based on the text information input, providing home appliance product descriptions and functional usage guidance. The Midea Meiyan big model algorithm is used for home appliance human-computer interaction services, which can generate text content that meets user needs based on the text data input by users.
Mobile phone companies include Xiaomi's big language model algorithm, vivo AI painting algorithm, and vivo Blue Heart big model algorithm, covering scenes of text and images.
Among the three major operators, China Mobile's Nine Day Natural Language Interaction Algorithm, China Telecom's TeleChat Semantic Model Generation Algorithm, Visual Generation Algorithm, and Speech Generation Algorithm are included in this batch of lists. The TeleChat semantic model generation algorithm of China Telecom is applied to dialogue generation scenarios, generating answer texts based on user input question descriptions, and achieving functions such as knowledge Q&A, article writing, and English translation.