Jan 26, 2024

Exclusive: Another 14 AI Large Models in Third Batch Registered With Regulator 全文12128字 According to TMTPost, China has approved over 40 AI large-scale model products, and an additional 240 large-scale models and AI technology projects have been included in the Domestic Deep Synthesis Service Algorithm Filing List. BEIJING, January 26 (TMTPOST) -- A total of 14 large models from 13 domestic AI companies have just received a new round of registration approval, following the first batch approved in August 2023 and the second batch in November 2023. They include Xiaomi's "XiaoAi Assistant", 4Paradigm's Large Language Model (LLM) "Shi Shuo," and "01.AI Large Models,”BOSS Zhipin's LLM "Nanbei Pavilion" and LLMs from Ainergy, ThreatBook, Frontis, Xin Yi Tech, Effyic as well as, Zhi Tech Group's "AI Q&A Robot", Maimai's new feature of "Intelligent Q&A", Zhaopin's "AI Resume Revision", and IReader Technology's WeChat mini-program. TMTPost App learned that some of the large model products have been available to users  from Friday. On Tuesday, 4Paradigm announced the registration of its LLM "Shi Shuo".On Friday afternoon, Frontis said on its official account that both "Pinshang" and "Moxiao Xian" LLMs have been approved. TMTPost have asked for further details from several involved companies, including Xiaomi Group and BOSS Zhipin. However, Xiaomi Group said they were unaware, and other companies have not yet responded. The "Interim Measures for the Management ofGenerative Artificial Intelligence Services," which officially took effect on August 15, 2023, stipulate that providers of generative artificial intelligence services with the capacity to guide public opinion or mobilize society must conduct security assessments in accordance with relevant laws and regulations, and fulfill the algorithm registration, modification, and cancellation procedures in accordance with the "Internet Information Service Algorithmic Recommendation Management Provisions." On August 31 last year, the first batch of more than eight AI large models were publicly announced after passing the registration process, including Baidu's "Ernie Bot," Baichuan Intelligent's "Baichuan Large Model," SenseTime's SenseChat, TikTok’s Lark, ZhipuAI’s GLM, Chinese Academy of Sciences’s Zidong Taichu, MiniMax’s ABAB from Shanghai, and Shanghai Artificial Intelligence Lab’s Shusheng General, etc. On September 20, Qihoo 360 Technology's "360 Zhinao" also announced that it would be open to the public from that day forward. On November 4 last year, the second round of 11 AI large model products were approved, including NetEase Youdao's "Zi Yue" Large Model, Mianbi Intelligence's "Mianbi Luca," Mobvoi's "Sequence Monkey," Kunlun Wanwei's "Tiāngōng" Large Model, Meituan's self-developed Large Model, Zhihu's "Zhihai Tu AI" Model, Moonshot by The Dark Side of the Moon, Kingsoft Office's WPS AI, TAL Education Group's MathGPT Large Model, Qihoo 360's Qiyuan Large Model, and Ant Group's Bailin Large Model. By the end of 2023, products such as Shenyan Technology's "Yu Jing Large Model," Lanzhou Technology's "Mengzi GPT" Large Model, Alibaba's "Quark Large Model," and Zhongke Wenge's "Ya Yi Large Model" all announced their record-filing status. This new round of 14 AI large models represents the largest number approved in history and also marks the first time that a single company has had two large models approved concurrently. Among them, "Xinyi Video" large model is the first video type model product to be approved; Xianyuan Technology's "Pinshang" and "Moxiaoxian" two large models were approved at the same time; In addition, 5 large models were approved based on new features of products, including the new features of Xiaomi's "XiaoAi AI Assistant," What's Worth Buying's "AI Q&A Robot," Maimai's "Intelligent Q&A," Zhaopin's "AI Resume Revision," and iReader's "Yue'ai Chat" mini program. Xiaomi Group (1810.HK)'s new feature of "XiaoAi AI Assistant" was developed by Xiaomi Corporation, based on Xiaomi's hardware ecosystem and massive data, to provide users with world-leading AI technology services including voice recognition and NLP. In December 2022, Xiaomi officially announced the 6.0 version of XiaoAi, bringing features such as scanning, object recognition, business card scanning, real-time translation, call translation, screen translation, and intelligent reminders; In August 2023, the large model-enhanced "XiaoAi" began internal testing, with the built-in XiaoAi interaction model, which can achieve more powerful AI functions and capabilities such as answering questions, inspiring creativity, and creating art. 4Paradigm (HKG: 6682)'s "Shishuo" large model was released in February 2023, and after continuous iterations, is committed to transforming enterprise software with generative AI. It has carried out large model cooperation with customers in finance, retail, manufacturing, logistics, education, medical and other industries. 4Paradigm indicated that in the future, Shishuo will continue to deepen joint exploration with enterprise customers on the large model side, to provide more reliable services. The "01.AI" large model, released in last Novemer, is an open-source AI large model product developed by a Beijing-headquartered startup founded by Kai-Fu Lee. It has versions with 6 billion and 34 billion parameters. Lee, Chairman and CEO of Innovation Works and CEO of 01.AI, has said that the company has already initiated the training of 100 billion parameter model, which is expected to be released in the coming months. It is also expected to launch super applications based on AI 2.0 and other technological products. The "Magic Cube" large model developed by CreateView Technology, a subsidiary of the IoT listed company CSOT, is an AI large model product that claims to cover large models from edge computing, language, multimodality, to robotics, closely integrated with business scenarios. In the smart cockpit field alone, the Magic Cube model has been applied in collaboration with automobile manufacturers in the voice assistant and 3D graphics of smart cockpits. The "Nanbei Pavilion" large model by BOSS Zhipin (HKG: 2076) is a generative AI service by Beijing Huapin Borui Network Technology Co., Ltd., based on its independently developed Nanbei Pavilion large model, through the official website of the BOSS Zhipin platform, the official mobile app, WeChat mini-programs, and other related products. The "Nanbei Pavilion" large model platform by BOSS Zhipin is still in the internal testing stage and requires activation. The "Pinshang" large model is an AI large model product developed by Xianyuan Technology for scenarios such as e-commerce and FMCG. It is reported that Xianyuan Technology was established at the end of 2021 by Dr. Zhou Bowen, a tenured professor in the Department of Electronic Engineering at Tsinghua University and the Huiyan Chair Professor at Tsinghua University. The company uses cutting-edge AI technology to help businesses integrate digital and physical processes, achieving a digital transformation from demand insight to product innovation, and disruptively completing consumer-centric product innovation. Xianyuan Technology states that the Pinshang large model is one of the company's core products. With a large model scale of hundreds of billions of parameters, it provides enterprises with full-link support from market insight to marketing innovation and has been widely applied in multiple industries, including consumer goods (food and beverages, 3C, home appliances, shoes and clothing), finance, and government affairs. "Moxiaoxian" large model is developed by Xianyuan Technology and is a multimodal model technology with a parameter scale of over 3.5 billion, designed specifically for industries such as clothing, beauty, food and beverages. The model can generate highly controllable product forms, lighting effects, and model images, providing users with innovative services such as model customization and transforming ordinary figures into supermodels. It is worth mentioning that "Moxiaoxian" has introduced its self-developed Protected-SD technology, which provides copyright authorization, generation, and traceability protection for AI-created content. This is the first model AI commercial photography platform in the industry to achieve such functionality. "Xinyi Video" large model is a multimodal large model product developed by Xinyi (Beijing) Technology Co., Ltd. (Xinyi Video Company), which was released last July and claimed to be the first large model in China focused on video. It is reported that Xinyi Video Company was established in 2022 and is a domestic AI-generated video application and technology service provider. Driven by AI, it provides "generative AI" creative services in fields such as video generation, vertical industry large models for video, digital humans, VR, etc. The founder and CEO of Xinyi Video is Han Kun, who is also the founder of Yixia Technology (Weibo Miaopai, Xiaokaxiu). Cyber Intelligence's "Yiye Qingzhou" large model is a generative large language model product developed by Cyber Intelligence Technology (Beijing) Co., Ltd., including commercial inference models with parameter scales of 33 billion, 13 billion, and 7 billion. It is reported that Cyber Intelligence was established in November 2018 and is a company that develops AI technologies such as natural language processing, deep semantic understanding, and image-semantic interaction. The founder and chairman, Wang Chunhui, is a candidate for a Doctor of Management at the University of Hong Kong and holds a Ph.D. in Computational Mathematics from the University of Washington. He has worked at Oracle and IBM, with partners including Huawei, Alibaba Cloud, Mengniu, and Industrial and Commercial Bank of China. Cyber Intelligence states that the "Yiye Qingzhou" large model has become the core technology engine for digital transformation and intelligent application in enterprise, government, and consumer scenarios, and in October 2023, it became one of the first key projects to settle in the Beijing Shijingshan District General Artificial Intelligence Large Model Industry Agglomeration Area. Shenmezhidemai "AI Q&A Robot" is an AI large model algorithm product of the A-share listed company Shenmezhidemai, mainly providing applications such as product encyclopedia and brand navigation. Maimai's "Intelligent Q&A" new feature is an AI function product launched by the professional networking platform "Maimai" after the explosion of ChatGPT. It has AI Q&A capabilities and can provide real-time responses in community questions, which may help people in job hunting and acquiring workplace knowledge, etc. Zhaopin's "AI Resume Revision" new feature can reprocess and automatically generate more standardized resume content based on the user's initially filled resume version. Zhaopin, established in 1997, was listed on the New York Stock Exchange in 2014. The company's main goal is to establish a more precise and efficient connection between talent and opportunities. IReader's "Yue Ai Chat" WeChat mini-program is an AI conversational robot-style product, offering "chat about books" and "chat about people" two major scenarios. Users can converse with established character images in the system and read classics or novels in the form of conversations. Currently, IReader Technology mainly uses the technical support provided by Alibaba Cloud. According to statistics by TMTPost App, as of now, China has already had over 40 AI large model products that have received registration-based approval, and in addition, more than 240 large model and AI technology projects have been included in the "Domestic Deep Synthetic Service Algorithm Record" list. According to the "Beijing Artificial Intelligence Industry Large Model Innovation Application White Paper (2023)", as of October 2023, China had a total of 254 manufacturers and academic institutions with large models of over 1 billion parameters, distributed across more than 20 provinces, municipalities and regions, among which Beijing had 122. Several AI companies have told TMTPost App that in the first half of 2024, the company will rapidly promote the commercial implementation of super applications, corporate cooperation, and the large model ecosystem. 下载 钛媒体App 了解更多科股数据详情