Recently, the official website of the Cyberspace Administration of China publicly released the second batch of domestic deep synthesis service algorithm record information. So far, 151 algorithms in China have been recorded on the deep synthesis service algorithm list. Tencent, Baidu, and 11 other AI large model products have completed the record and are gradually going online, marking the formal entry of the large model industry into the productization and landing stage in China, and the large model industry will enter the fast lane of productization and landing.
Image Source: Generated by Wujie AI
Financing
Zhejiang Xingyao Completes 66 Million Yuan B1 Round Financing, Main Business Involves AI Computing Power Center Construction
According to EO Company, Zhejiang Xingyao recently completed a 66 million yuan B1 round of financing, with Zhicheng Capital as the investor, and Ruituo Capital, Yao'an China, and other old shareholders following suit. Zhejiang Xingyao is an internet industry e-commerce service provider, currently serving mainstream MCN companies nationwide to provide AI digital human computing services and e-commerce MCN company AI small model development services, helping traditional e-commerce and entertainment anchors achieve digital transformation.
AI Drawing Software Developer "Huaien Era" Completes Angel Financing of Tens of Millions
According to the "Science and Technology Innovation Board Daily" on August 29, Huaien Era recently completed angel round financing of tens of millions, led by a well-known manufacturer in Jiangsu, and the financing funds will be mainly used for AI algorithm research and development, team expansion, and the research and development and market expansion of its "Left Brain AI" technology.
AI Startup ConverSight Raises $9 Million, Using Generative AI to Accelerate Data Analysis
According to Zhanzhang Zhi Jia on August 29, the Indianapolis-based startup ConverSight announced that it has raised $9 million in Series A financing, dedicated to using generative artificial intelligence to provide better and faster data analysis for enterprises.
AI-Assisted Language Learning App Speak Raises $16 Million in Financing
According to Zhanzhang Zhi Jia on September 1, the AI-assisted language learning app Speak announced that it has raised $16 million in Series B-2 financing, led by angel investor Lachy Groom. Dropbox co-founders Drew Houston and Arash Ferdowsi also participated in this round of financing. Speak plans to use this funding to expand into more markets, including entering the US market by the end of the year.
Rockset Completes $44 Million Financing to Meet the Growing Demand for Real-Time Indexing in the Generative AI Era
According to Venture Beat on August 29, database supplier Rockset is raising $44 million in new funds to meet the growing demand for real-time indexing in the generative artificial intelligence era. It is reported that this is another round of financing after the company's Series B financing, bringing its total financing to date to $105 million. The new round of financing is led by Icon Ventures, with participation from Glynn Capital, Four Rivers, K5 Global, Sequoia, and Greylock.
Large Models
OpenAI Releases ChatGPT Enterprise Edition: Unrestricted Use of GPT-4, Higher Security and Privacy Protection
According to IT Home, OpenAI today released a new ChatGPT product specifically designed for enterprise users, providing higher security and privacy protection. In its blog post, OpenAI stated that the ChatGPT Enterprise Edition not only provides unrestricted and rapid access to the powerful GPT-4 model, but also allows for deeper data analysis to help enterprises quickly understand information and ask more complex questions to ChatGPT.
Users of ChatGPT Enterprise Edition have full control and ownership of their data, which will not be used to train the GPT-4 model. OpenAI stated that this is the first enterprise-oriented product of ChatGPT, independent of ChatGPT and ChatGPT Plus.
Tencent, Baidu, and 11 Other AI Large Model Products Pass the Record of Generative Artificial Intelligence, Will Gradually Go Online from Today
According to Beike Finance on August 31, it was learned from multiple independent sources that 11 large models in China will gradually pass the "Interim Measures for the Administration of Generative Artificial Intelligence Services" record, and the first batch will be open to the public from August 31. Among them, 5 in Beijing, 3 in Shanghai, 2 in Guangdong Province, and 1 in other provinces and cities will be gradually opened.
It is reported that the approved companies in Guangdong are Huawei and Tencent, and the products approved in other regions are related to iFlytek. The 5 large model products in Beijing are Baidu's Wenxin Yiyuan, Douyin's Yunque, Baichuan Intelligence's Baichuan Large Model, Zhipu Huazhang's Zhishu Qingyan, and the Chinese Academy of Sciences' Zidong Taichu. The 3 large model products in Shanghai are Sensetime's "SenseChat", MiniMax's ABAB large model, and Shanghai Artificial Intelligence Laboratory's Shusheng General Large Model.
APUS Announces Self-Developed Trillion-Level Multimodal AI Large Model "Tianyan Large Model"
According to China National Radio, on September 2, at the 2023 China International Fair for Trade in Services, AI company APUS announced for the first time its self-developed trillion-level multimodal AI large model "Tianyan Large Model" (also known as AiLMe Large Model).
It is reported that the Tianyan Large Model has the ability to understand and generate text, images, videos, and audio, and has achieved the third overall ranking in the internationally authoritative evaluation C-Eval. For specific application scenarios, APUS has distilled four refined models from the Tianyan Large Model, including the text model "Yique Eight", the image model "Yique Three", the audio model "Yique Six", and the video model "Yique Four".
Mianbi Intelligence Releases Trillion-Level Multimodal Large Model Luca
According to Interface News, on the afternoon of August 28, Mianbi Intelligence released its latest large model achievement Luca at the "General Artificial Intelligence Computing Power Forum" (AGICF). Mianbi Intelligence co-founder and CEO Li Dahai stated that Luca's multi-language model capabilities are comparable to ChatGPT, and emphasized its powerful multimodal capabilities and AI-native application scenarios based on collective intelligence technology. At the end of the press conference, Li Dahai announced the official start of the public beta for the multimodal intelligent assistant Luca 2.0 driven by the trillion-level base model.
Xiaomi Sound/Sound Pro Smart Speaker Opens Recruitment for Xiaoai Large Model Beta Testing
According to IT Home on September 3, today, an IT Home colleague submitted news that Xiaomi will invite some users to participate in the beta testing of Xiaomi speakers before the large model is officially open to all users. The first batch has been pushed, and subsequent openings will be irregular. According to the screenshots provided, the devices participating in the beta testing this time include the Xiaomi Sound Pro smart speaker. According to the post by the head of the "Speaker" circle in the Xiaomi community, the recruitment for beta testing started on August 30, initially only opening for the Xiaomi Sound and Xiaomi Sound Pro smart speakers, with other products to be announced later.
News Claims Tencent Will Release Hybrid Large Model in Early September, Tencent Documents and Tencent Games Have Joined the Testing
According to IT Home on September 2, it was recently learned from Tencent that Tencent will officially release the hybrid large model at the 2023 Tencent Global Digital Ecology Conference in early September, but it did not mention whether it will be open to the public.
Tencent previously stated that the hybrid large model has entered the internal application testing stage, and Tencent Cloud, Tencent Advertising, Tencent Games, Tencent Financial Technology, Tencent Meeting, and Tencent Documents, among other internal businesses and products, have already joined the testing of Tencent's hybrid large model and achieved preliminary results.
UAE Launches Open Source Arabic AI Large Language Model Jais with 13 Billion Parameters
According to IT Home on August 30, a team in the UAE recently announced a large Arabic AI model named Jais, which was developed in collaboration with a group of engineers, researchers, and a Silicon Valley chip company. It is reported that the Jais large language model contains 13 billion parameters, which are composed of a large amount of combined Arabic and English data, including some from computer code. The model was trained on the Cerebras Systems supercomputer in Silicon Valley and is a collaborative project between Cerebras, the UAE Artificial Intelligence University, and Inception, a subsidiary of the UAE G42 Technology Group focused on artificial intelligence.
DeepMind Co-Founder: OpenAI Secretly Training GPT-5
According to the "Science and Technology Innovation Board Daily" on September 4, DeepMind co-founder and Inflection AI CEO Mustafa Suleyman revealed in an interview that OpenAI is secretly training GPT-5, and within the next 18 months, the models trained by Inflection AI will be 100 times larger than the current cutting-edge models. Within the next 3 years, Inflection's models will be 1000 times larger than they are now.
Zuoyebang Releases Galaxy Large Model, Ranking First on C-Eval and CMMLU Lists
According to 36Kr on September 2, Zuoyebang released its self-developed Galaxy large model. It is a large education model covering multiple subjects, stages, and scenarios, integrating AI algorithms and educational data. Data shows that the Galaxy large model ranks first on both the C-Eval and CMMLU lists, making it the first education large model to rank first on both of these authoritative lists. It is understood that the model will soon be applied in Zuoyebang's app, learning tools, smart hardware, smart books, and digital education business scenarios.
Venustech Releases "Fengyunwei" Large Model
According to 36Kr on September 1, Venustech released the Venustech Fengyunwei large model. It is a large model in the security industry vertical field, built on a large amount of security professional knowledge, and has already started trial use of related products.
Innovative Qizhi Launches "Qizhi Kongming AInnoGC" Industrial Large Model Product Matrix
According to Sina Technology on September 1, Innovusion announced the launch of the "Qizhi Kongming AInnoGC" industrial large model product matrix, releasing an industrial large model (Alnno-15B) with a parameter volume of over 150 billion, a large model service engine, and three generative AI application products based on large models - "Qizhi Mingda ChatRobot" generative industrial robot task scheduling application, "Qizhi Mingshu ChatBl" generative enterprise private domain data analysis application, and "Qizhi Mingrui ChatDoc" generative enterprise private domain knowledge Q&A application.
Baijiayun Releases Baiwen Large Model
According to 36Kr's report on August 31, the all-in-one AI video technology service provider Baijiayun held a strategic launch event and officially released the Baiwen large model, a one-stop AIGC scenario platform created by the company.
"Zhipu AI" Officially Launches the First Generative AI Assistant Zhipu Qingyan
According to Babite on August 31, "Zhipu AI" officially launched the first generative AI assistant, Zhipu Qingyan, based on the ChatGLM2 bilingual dialogue model independently developed by Zhipu AI. Users can experience it by downloading the application from the app store or using it in the WeChat mini-program.
Chuangyebang Launches Investment Large Model BangChat
According to Zhanzhang Zhi Jia's report on August 30, Chuangyebang has launched the investment large model BangChat by deeply integrating proprietary data with artificial intelligence technology and has made a new version of the website and mobile client. It is reported that Chuangyebang's data intelligence drives entrepreneurs' decision-making in business planning, market forecasting, and other areas, while BangChat provides intelligent analysis and decision-making capabilities through generative AI tools.
iFlytek Introduces Shanhai Large Model 2.0 with a Parameter Scale of Hundreds of Billions
According to Zhanzhang Zhi Jia's report on August 31, iFlytek announced the release of Shanhai Large Model 2.0, with a parameter scale reaching hundreds of billions and an increased amount of pre-training language data, with training data (Tokens) reaching two trillion (2.0T). Its performance in the C-Eval global large model comprehensive evaluation exceeded that of GPT-4, entering the top three with a score of 70.
Maishang Consumer Releases "Tianjing" Retail Financial Large Model
According to Zhanzhang Zhi Jia's report, on August 28, Maishang Consumer Finance Co., Ltd. released the first retail financial large model "Tianjing" at the Financial Large Model Development Forum. Maishang Consumer is focusing on the development of large models in six core areas, including basic language characteristics, logic and reasoning, semantic understanding, generation and creation, financial capabilities, and security and compliance, to promote financial digital transformation.
Anheng Information Releases "Hengnao" Security Vertical Large Model
According to 36Kr's report on August 28, Anheng Information held a strategic upgrade release conference based on large models in Hangzhou with the theme "Making AI Safer and Security Smarter." At the event, the "Hengnao Security Vertical Large Model" was officially released, providing development momentum for Anheng Information's data security strategy, MSS strategy, and talent strategy.
Applications
Tencent Develops AI Design Platform AI Design
According to Zhanzhang Zhi Jia's report on September 4, Tencent independently developed the AI Design intelligent design platform, which uses artificial intelligence technology to generate brand logos, aiming to help individual users (now supporting QQ and WeChat users) efficiently and cost-effectively inspire brand ideas. It is easy to use, requiring only 3 steps (entering the brand name, selecting keywords and preferred colors) to obtain multiple logo options.
Baidu Input Method Launches "Super Writing" AI Assistant, Based on the Wenxin Large Model
According to IT Home on September 2, following the announcement of the opening of Wenxin to the public, Baidu Input Method recently received an update to version 11.7.19.9 for Android. The upgraded Baidu Input Method added a full-scenario AI creative assistant called "Super Writing," based on the Wenxin large model, which can help users with "high EQ communication," "divine comments," rewriting copy, and more. Additionally, the feature supports providing users with short video captions, popular titles for Xiaohongshu, and creative Weibo posts.
Daguan Data Launches "Caozhi" Large Model Integrated Machine
According to the "Science and Technology Innovation Board Daily" on September 1, on the eve of the 2023 INCLUSION - The Bund Conference, Daguan Data launched the private "Caozhi" large model integrated machine, which is equipped with its independently developed "Caozhi" large model and supports third-party open-source model libraries for deploying large model applications within enterprises.
Meitu's AI Digital Avatar Generation Tool DreamAvatar Goes Online, Introducing the "AI Actor" Digital Human Service
According to Babite on August 29, Meitu announced that its AI digital avatar generation tool, DreamAvatar, has recently gone online, introducing the "AI Actor" digital human service. It is AI-driven and serves productivity scenarios such as video content creation, film and video processing, and editing.
Huawei Releases Mate 60 Pro, Integrating the Pangu AI Large Model
According to Babite, Huawei's official store today released a letter to Huawei users, launching the "HUAWEI Mate 60 Pro Pioneer Plan" to allow some consumers to experience the most powerful Mate phone in history. The Mate 60 Pro is the world's first mass-market smartphone to support satellite calls, enabling users to make and receive satellite calls even without ground network signals. It also features AI remote control, intelligent payment, and non-screen gaze functions. Additionally, it integrates the Pangu AI large model to provide consumers with a smarter interactive experience.
DingTalk AI Used for the First Time in the Hangzhou Asian Games
According to Babite, on August 28, DingTalk announced that the Hangzhou Asian Games and DingTalk jointly created the world's first large-scale sports event integrated intelligent event management platform, "Asian Games Ding." This is the first time DingTalk AI has been used in the Hangzhou Asian Games.
Samsung Electronics to Introduce Bixby-Based Generative AI in All Home Appliances Next Year
According to the "Science and Technology Innovation Board Daily" on September 4, Miyoung Yu, Vice President of Samsung Electronics' home appliance division, announced at the IFA 2023 exhibition that Samsung Electronics will introduce Bixby-based generative AI functionality in all home appliances next year. If the plan goes smoothly, Samsung Electronics' home appliances are expected to not only adopt AI services in high-end product lines such as Bespoke but also apply and implement them in entry-level product lines.
"80s" Female Programmer Uses AI to Assist Writing, Signs Contracts for 16 Novels in Six Months
According to Changjiang Cloud News, an "80s" female programmer named Lanmao used AI assistance and signed contracts for 16 novels with online novel platforms in six months. The highest-rated novel received a score of 8.2 from netizens. Lanmao mentioned that she could use AI assistance to write a 5,000-word novel in one to two hours, and a 10,000-word novel in three to four hours. She also mentioned that AI can provide unlimited inspiration, text, and plot directions, and can compensate for her lack of emotional appeal due to her overly rational thinking.
Regulation
151 Algorithms Completed Record Filing, Including Wenxin, Xunfei Xinghuo, and Other AI Large Models and AIGC Products
According to AI Black Horse's report on September 4, 151 algorithms have completed record filing on the list of deep synthesis service algorithms, including 110 from the second batch and 41 from the first batch of the "Domestic Deep Synthesis Service Algorithm Filing List." This includes AI large models and AIGC products such as Wenxin, Xunfei Xinghuo, 360 Zhinao, Hybrid Large Model, Pangu Large Model, ChatGLM, Youdao Ziyue Large Model, Xiaobingdao, WPS, Douyin, Moyin Workshop, Sense Large Model, Meitu Show, Quark, Congrong Large Model, Meituan, DingTalk, Tencent Meeting, and QQ Music.
Perspectives
OpenAI Expected to Generate Over $1 Billion in Revenue in the Next 12 Months
According to The Information's report on August 30, OpenAI is expected to generate over $1 billion in revenue in the next 12 months through the sale of artificial intelligence software and its computing power. This far exceeds the revenue forecast previously disclosed to shareholders by the company. Earlier this year, when investors bought shares from existing shareholders, the company's book value was $27 billion. Before starting to charge for ChatGPT last year, OpenAI only generated $28 million in revenue.
Amid the AI Boom, Anguilla Expects to Generate Up to $30 Million in Revenue with Its ".ai" Domain
According to IT Home's report on September 2, Anguilla, a small island in the Caribbean Sea and a British territory, has owned the top-level domain ".ai" (country and region class) since the 1990s. Anguilla expects to generate up to $30 million in revenue with its ".ai" domain.
The global AI boom over the past year has made the ".ai" top-level domain extremely attractive to tech companies. Industry giants such as Musk's "X" company, Google, Microsoft, and Meta own URLs such as X.ai, google.ai, and microsoft.ai, which can direct users to their AI service webpages.
Alibaba Cloud Chairman Zhang Yong: In the AI Era, China May See a New Generation of SaaS
According to Geek Park, on August 23, Alibaba Cloud, in collaboration with Founder Park, invited over 20 outstanding entrepreneurs in the field of large models, including model layer, tool layer, and application layer, for a face-to-face closed-door exchange at the West Lake Wetland in Hangzhou. At the closed-door meeting named "Xixi Forum" by Alibaba Cloud Chairman Zhang Yong, Zhang expressed that the development of AI technology has brought new possibilities to the ecosystem. He expressed regret that while the past decade saw rapid development in China's cloud computing infrastructure, the SaaS industry in China did not see substantial improvement. In contrast, SaaS companies in the United States are exploring the integration of AI into their platforms, taking a different path from Chinese companies. He believes that in the AI era, China may see a new generation of SaaS, which will be a new intelligent service driven by data and intelligence, different from the process-driven SaaS of the past, and it may not even be called SaaS.
Research Reports
CITIC Securities: The Large Model Industry Will Enter the Fast Lane of Productization and Implementation
According to Jiemian News, CITIC Securities' research report on September 4 stated that the first batch of 8 large models approved for launch marks the official entry of the large model industry in China into the stage of productization and implementation. Subsequently, large model products from companies such as Alibaba, Tencent, and iFlytek are also expected to be approved, further accelerating the productization and implementation of AI products. The domestic large model product ecosystem will gradually be constructed, and services will be standardized, leading the large model industry into the fast lane of productization and implementation. The accelerated entry of large models into the stage of productization and implementation directly benefits large model service providers, companies with existing large models, or those with large model development capabilities. The productization of large models will accelerate iteration speed, prioritize efficiency and accuracy, and further increase the demand for computing power, benefiting the domestic computing power sector.
CCID Consulting: As of July, China Has Seen a Total of 130 Large Models
According to China Economic Net's report on August 28, CCID Consulting recently released the first phase of the "IT 2023" series of in-depth research, showing that in 2023, China's domestic large models have experienced explosive growth. In just the first 7 months of 2023, a total of 64 large models were released. The research indicates that China is closely following the international frontier in large models. In 2021, China started a wave of large model releases, gradually emerging a group of large models with industry influence. In 2023, driven by ChatGPT, the development of large models has entered a new stage, and domestic large models have shown explosive growth. As of July 2023, a total of 130 large models have been released in China.
The research report stated that in 2021, the number of large model releases abroad was 38, with a growth rate of 171%. As of July 2023, the cumulative number of large model releases abroad reached 138, with 114 large models in the United States.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
