Founded in 2019, Zhipu AI has become the first domestic unicorn in the large model field with a valuation exceeding 10 billion RMB. Tencent, which once said "it's not important to bring out the light bulb a month earlier," eventually couldn't help but follow the trend. Tencent officially launched the Mix Element Large Model, but users still need to queue up to apply. The Ant Financial Large Model made its debut at the Bund Summit, claiming to solve real industry problems.
Image Source: Generated by Wujie AI
Financing
Zhishu AI's latest valuation exceeds 10 billion RMB
According to exclusive information obtained by AI Technology Review, Zhishu AI's latest valuation has exceeded 10 billion RMB, with the highest possibly reaching 15 billion RMB, making it one of the fastest companies in China to reach a valuation of over 10 billion RMB. Zhishu AI was established in June 2019, and its core team members were involved in the development of the "Wudao" project, a collaboration between Tsinghua University and the Institute of Computing Technology.
AI smart hardware company "Future Intelligence" completes tens of millions of RMB in Pre-A round financing
According to 36Kr, AI smart hardware company "Future Intelligence" recently announced the completion of tens of millions of RMB in Pre-A round financing, led by Sky Capital. This round of financing will be used for new product launches, market expansion, and other purposes. Future Intelligence is targeting the office scene and has launched multiple conference office headphones such as iFLYBUDS Pro, iFLYBUDS Air, and iFLYBUDS Nano+.
Zhongdiao Technology completes tens of millions of RMB in A2 and A3 round financing, with funds mainly used for product iteration and large model upgrades
According to the "Science and Technology Board Daily" on September 11, Shanghai Zhongdiao Information Technology Co., Ltd. recently announced the completion of A2 and A3 round financing invested by Yunqi Capital and Dongfang Datong, with a total amount of tens of millions of RMB. Shiyu Capital served as the exclusive financial advisor for this round. The funds will be mainly used for product iteration, large model upgrades, and expanding sales scale. Zhongdiao Technology was established in 2015 and focuses on the integration and aggregation of automotive big data, intelligent algorithms, and intelligent empowerment.
AI startup Imbue receives investment from NVIDIA and others, with a valuation exceeding 1 billion USD
According to IT Home on September 8, Imbue is a startup company that builds large-scale language models to drive artificial intelligence tools. The company has raised 200 million USD from investors such as NVIDIA, with a valuation exceeding 1 billion USD (approximately 7.33 billion RMB).
Silicon Valley AI startup d-Matrix completes 110 million USD in Series B financing, led by Temasek
According to the "Science and Technology Board Daily" citing Reuters, the Silicon Valley generative artificial intelligence chip startup d-Matrix recently completed 110 million USD in Series B financing. This round of financing was led by the Singapore investment company Temasek, with participation from California venture capital company Playround Global, Microsoft's M12 venture fund, Nautilus Venture Partners, Entrada Ventures, and others. The raised funds will be used for recruiting and commercializing the company's Corsair platform.
US startup Delphi completes 2.7 million USD in financing and launches AI digital cloning service
According to Webmaster's Home on September 7, the US startup Delphi raised 2.7 million USD in financing and launched a new AI digital cloning service. The core function of this service allows users to generate AI chatbots that mimic the user's personality, writing, or speaking style by uploading communication records, including emails, chat records, YouTube videos, and more. Through a collaboration with the voice cloning startup ElevenLabs, Delphi's AI clones can even mimic the user's voice.
AI chip developer "Zhonghao Xinying" completes tens of millions of RMB in Pre-B+ round financing
According to the official account of "Zhonghao Xinying Technology," Hangzhou Zhonghao Xinying recently completed tens of millions of RMB in Pre-B+ round financing, with investment from Hangzhou Huaxia Hengtian Capital. The funds will be used to promote the company's AI computing cluster business.
Gleen AI receives 4.9 million USD in funding to address AI hallucination issues using enterprise data
According to the "Science and Technology Board Daily" on September 6, the startup Gleen AI recently announced a funding of 4.9 million USD. Investors include Slow Ventures, 6th Man Ventures, South Park Commons, Spartan Group, and other venture capital companies and angel investors, including former Facebook/Meta platform product management vice president Sam Lessin. The funds will be used to continue developing an enterprise-focused anti-hallucination data layer software, which converts internal enterprise data into a vector database and uses this data to improve the quality of AI model responses.
AI productivity platform Kindo receives 7 million USD in financing
According to Webmaster's Home on September 6, the artificial intelligence startup Kindo recently announced a seed round financing of 7 million USD, which will be used to launch a secure AI productivity platform for enterprises.
Atomicwork secures 11 million USD in financing to integrate AI into employee workflows
According to Forbes on September 5, the San Francisco-based startup Atomicwork announced a seed round financing of 11 million USD to commercialize its software-as-a-service solution. Atomicwork's work assistant, Atom, based on generative AI, aims to enable employees to interact with it in a more natural way than traditional software.
Large Models
Insiders: Meta is developing a new large language model, planning to rival GPT-4
According to The Wall Street Journal on September 10, insiders revealed that Meta is developing a new artificial intelligence system that will rival the most advanced model GPT-4 provided by OpenAI. Meta's goal is for its new AI model to be several times more powerful than the model named Llama 2, which was released two months ago. The planned system will help other companies build services capable of generating complex text, analysis, and other outputs.
Baichuan Intelligence releases Baichuan 2 large model
According to the "Science and Technology Board Daily" on September 6, Baichuan Intelligence announced the official open source release of the fine-tuned Baichuan 2-7B, Baichuan 2-13B, Baichuan 2-13B-Chat, and their 4-bit quantized versions, all of which are free for commercial use.
Tencent officially launches Mix Element large language model with over a trillion parameters
On the morning of September 7, at the Tencent Global Digital Ecology Summit, Tang Daosheng, Senior Executive Vice President of Tencent and CEO of the Cloud and Smart Industries Group, announced the Tencent Mix Element large model. This large model is a self-developed large language model by Tencent with over a trillion parameters and pre-training corpus of over 20 trillion tokens (text units of large language model), possessing Chinese composition ability, task execution ability, and logical reasoning ability in complex contexts.
Baidu Cloud Intelligence announces the comprehensive upgrade release of the Qianfan large model platform 2.0
According to Webmaster's Home, at the 2023 Baidu Cloud Intelligence Conference, Baidu Cloud Intelligence announced the comprehensive upgrade release of the Qianfan large model platform 2.0, providing the most complete toolchain, the best computing power efficiency, the most large models and datasets, and enterprise-level security to help enterprises bridge the gap from large models to productivity.
Ant Financial officially releases the Financial Large Model
According to Babite News on September 8, at the Bund Summit held in Shanghai, Ant Group officially released the Financial Large Model. It is based on Ant's self-developed basic large model, customized for the financial industry, with a bottom-level computing cluster reaching a scale of tens of thousands of cards. This large model focuses on real financial scenario needs and excels in 28 categories of financial-specific tasks in five dimensions: "cognition, generation, professional knowledge, professional logic, and compliance," reaching the level of industry experts in many areas such as "judgment viewpoint extraction," "financial intent understanding," and "financial event reasoning." Currently, the Ant Financial Large Model has been fully tested on Ant Group's wealth and insurance platforms.
iFlytek's Xinghuo large model is now open to the public
According to Babite News on September 5, iFlytek officially announced that the Xinghuo cognitive large model is now open to the public, and users can download and register to use it directly from major app stores. Xinghuo is a new generation cognitive intelligent large model launched by iFlytek, with cross-domain knowledge and language understanding capabilities, able to understand and execute tasks based on natural conversation. Xinghuo was released on May 6 this year and has now been upgraded to V2.0.
360 Brain Large Model is now open to the public
According to Caixin on September 5, the 360 Brain Large Model is now open to the public. The open platforms include the 360 Brain official website, app stores, or the latest versions of 360 Security Guard, 360 Secure Browser, and 360 Search for users to experience large model services.
Liou Group: Liou Digital launches large model in the marketing field
According to Interface News on September 11, Liou Group's Liou Digital announced through its official WeChat account that it has strategically invested in the artificial intelligence startup Hangzhou Guiyi Intelligent Technology Co., Ltd., and launched the "Liou Guiyi" marketing field large model. Enterprises can achieve intelligent execution and strategy optimization of advertising placements through the AI pitcher Agent "Guiyi Miaoji" in the marketing field large model "Liou Guiyi."
Tencent Health announces medical large model
According to Jinshi News, on September 8, Tencent Health announced the medical large model, releasing a multi-scenario AI product matrix including intelligent Q&A, family doctor assistant, and intelligent medical imaging platform. The drug discovery platform "Yunshen" (iDrug) revealed a new protein structure prediction algorithm framework tFold. Currently, Tencent's medical large model includes scene large models such as text generation, intelligent Q&A, structured and retrieved medical records, imaging reports, and auxiliary diagnosis, which can be embedded in the entire medical process.
According to the "Science and Technology Board Daily" on September 9, Tencent Meeting announced the integration of Tencent Mix Element large model and the release of Tencent Meeting AI Assistant. Based on the Tencent Meeting AI Assistant, participants can complete various complex tasks such as information extraction, content analysis, and meeting management through simple and natural commands. Tencent Meeting also announced commercial data: the number of users has exceeded 400 million, and the number of paid users has increased fivefold year-on-year.
The first domestic large model intelligent financial assistant appeared, Ant Wealth released Zhibao 2.0
On September 8, at the "Bund Summit" held in Shanghai, Ant Wealth platform released the first intelligent financial assistant using large model technology, Zhibao 2.0. The Zhibao 2.0 version is based on the recently released financial large model developed by Ant Group, and has significantly improved in knowledge, expertise, language, and security, providing financial institutions with high-quality market analysis, position diagnosis, asset allocation, and professional services such as investment education, with knowledge levels comparable to industry experts. Zhibao 2.0 is currently in internal testing and will be launched after completing relevant filing work.
Taobao launches AI large model application "Taobao Wenwen" for internal testing
According to 36Kr on September 8, Taobao has launched the AI large model application "Taobao Wenwen" for internal testing. Taobao Wenwen can understand consumer needs across categories and domains, and can fulfill requirements and scenarios such as information search and creative content generation. Currently, the application can be tested by submitting an application through the Taobao App by searching for "Taobao Wenwen" or by participating in testing through an invitation code. The official full launch date of the application has not been disclosed.
Morgan Stanley to launch AI chatbot co-developed with OpenAI this month, providing investment insights
According to Sina Finance on September 7, after testing with 1,000 financial advisors, Morgan Stanley will launch a generative artificial intelligence (AI) chatbot later this month. This AI chatbot was co-developed by Morgan Stanley and ChatGPT developer OpenAI.
Alibaba's Tmall releases two new AI advertising products "Tmall Bailin" and "Wanxiangtai Wujie Edition"
According to 36Kr, on September 6, at the 2023 m Summit, Alibaba's Tmall released two new AI advertising products, the one-stop brand management platform "Tmall Bailin" and the one-stop marketing delivery system "Wanxiangtai Wujie Edition," to create a growth "double engine" for merchants.
Kingsoft Office: WPS AI officially open to the public
According to Babite News on September 5, Kingsoft Office officially announced that WPS AI is now open to the public, with AI capabilities initially applied to WPS Intelligent Documents. More WPS AI capabilities will be available in other components. Users can experience WPS Intelligent Documents on the latest version of the WPS client/app, Kingsoft Document mini-program/official website.
It is said that WPS AI, as the first landing generative AI application in the Chinese collaborative office track, has been integrated into WPS Text, Presentation, Spreadsheet, PDF, and Kingsoft Document product lines, solving users' daily office problems in content generation, content understanding, and command operation. Kingsoft Office positions WPS AI as an application of large language models, and in the future, it will develop in three strategic directions: AIGC (content creation), Copilot (smart assistant), and Insight (knowledge insight).
Regulation
Revelation: Several small and medium-sized AI drawing websites using AIGC technology in China have been shut down
According to the public account "Vision Future" on September 7, several small and medium-sized AI drawing websites in China that used AIGC technology, such as 4B3.com, liblibai.com, and tusiart.com, were shut down without warning or transferred to overseas domains, and are currently inaccessible in China.
It is reported that similar shutdowns were accompanied by interviews with website responsible persons by the Cyberspace Administration of China and other supporting measures. These interviews were conducted in accordance with the relevant provisions of the "Administrative Enforcement Procedures of the Cyberspace Administration" and the "Interim Measures for the Management of Generative Artificial Intelligence Services," and the main reason was "calling overseas ChatGPT product interfaces without evaluation to provide intelligent dialogue and question-and-answer functions, posing significant security risks."
Previously, the Cyberspace Administration of China, together with the National Development and Reform Commission, the Ministry of Education, the Ministry of Science and Technology, the Ministry of Industry and Information Technology, the Ministry of Public Security, and the State Administration of Radio, Film, and Television, issued the "Interim Measures for the Management of Generative Artificial Intelligence Services" on July 10 and implemented it from August 15.
Big Companies
NVIDIA releases TensorRT-LLM, which can increase the inference performance of H100 by up to 8 times
According to IT Home on September 9, NVIDIA recently announced the release of TensorRT-LLM, a deeply optimized open-source library that can accelerate the inference performance of all large language models on AI GPUs such as Hopper. In performance tests, based on the A100, NVIDIA compared the H100 with and without TensorRT-LLM enabled, and found that in GPT-J 6B inference, the inference performance of H100 was 4 times that of A100, while the performance of H100 with TensorRT-LLM enabled was 8 times that of A100.
Media: ARM's oversubscription may exceed fivefold
According to bankers participating in ARM's roadshow, ARM's US IPO may bring in oversubscriptions exceeding fivefold. The British chip design company expects the AI boom to drive accelerated revenue growth in the future. ARM CEO Rene Haas stated in a promotional video, "AI will be ubiquitous, and all AI will run on ARM's software. Our opportunities are limitless, and with the advent of the AI era, global computing demand is insatiable."
Viewpoints
Robin Li: Entrepreneurs focusing on large models is meaningless, focusing on application opportunities is more significant
According to 36Kr on September 5, Baidu's founder, chairman, and CEO Robin Li stated at the Baidu Cloud Intelligence Conference that more than 70 large models with over 1 billion parameters have been released in China. However, the models themselves do not directly generate value; the applications developed based on the basic large models are the real significance of these large models. Therefore, for entrepreneurs, focusing on large models is meaningless, and focusing on application opportunities is more significant, just like in the era of mobile internet, where there were only Android and iOS operating systems, but many successful applications.
Wang Minghui, Managing Partner of Jianting Capital: AIGC is expected to bring economic benefits of $2.6 trillion to $4.4 trillion to enterprise operations
According to a report in the Shanghai Securities News, at the "New Wave of AI Frontiers and Applications" sub-forum at the 2023 Inclusion Bund Summit from September 7 to 9, Wang Minghui, Managing Partner of Jianting Capital, predicted that AIGC would release tremendous value in the critical functions of enterprises. It is expected that in 16 business functions of enterprises, AIGC can bring total economic benefits of $2.6 trillion to $4.4 trillion, with the impact concentrated in four functions, including customer operations, marketing and sales, software engineering, and product development, accounting for 75% of the total value.
Li Zhifei, CEO of Rokid: Computing power itself does not constitute a competitive barrier for large models
According to the "Science and Technology Board Daily" on September 8, at the 2023 Tencent Global Digital Ecology Summit, Li Zhifei, CEO of Rokid, a generative AI and voice interaction AI startup, stated that the computing power of large models is a consumable and does not constitute a barrier in itself. Rational and moderate investment is required. Data, especially the data accumulated by users in their own products, is more critical. The data flywheel is difficult to establish, but once established, it is a more reliable barrier than models and computing power.
Tencent's Tang Daosheng: It is too early to discuss the specific business prospects of large models, and the competition in the large model market is still in the early stages
According to The Paper, on September 7, Tang Daosheng, Senior Executive Vice President of Tencent and CEO of the Cloud and Smart Industries Group, stated that it is too early to discuss the specific business prospects of large models, and the competition in the large model market is still in the early stages. Different scenarios require different models, and it is not necessary to use a single solution to meet the demand. Tencent hopes to provide the most suitable tools for different users.
Ant Group's Vice President Wang Xiaohang: All key aspects of the financial industry are worth transforming with large model technology
According to Babite News on September 8, at the Bund Summit held in Shanghai, Wang Xiaohang, Vice President of Ant Group and head of financial large models, stated that all key aspects of the financial industry are worth transforming with large model technology. Financial large models need to ensure the rigor of domain knowledge and professional logic, as well as the controllability of content generation, in order to truly bring industrial value. This is also the real problem that financial large models need to solve.
Wang Minghui, Chairman of Jianting Capital: In the end, only one or two companies may dominate the large model market for consumers
According to Meta Universe NEWS on September 7, Wang Minghui, Founding Partner and Chairman of Jianting Capital, believes that large models for consumers face the dual challenges of regulatory rules and technological breakthroughs. From a business perspective, in the end, only one or two companies may succeed. For large models for businesses, the main focus is on serving business users, and it is expected to empower real industries and improve industry efficiency. However, the demands of business customers are more rigorous, and large models need to become reliable scene tools. Wang Minghui stated that the vigorous development of large models this year will definitely improve the level of enterprise digitization. In fact, data elements, large models, and the metaverse form an organic whole.
Ni Yili, Chairman of McKinsey China: Generative AI could have a positive impact of up to $7.9 trillion on the global economy
On September 7, at the 2023 Tencent Global Digital Ecology Summit, Ni Yili, Chairman of McKinsey China, stated that generative AI could have a positive impact of up to $7.9 trillion on the global economy, but the overall broad AI could reach $25 trillion. In addition, Ni Yili claimed that due to generative AI, the time when half of work activities are replaced by automation has been brought forward by ten years compared to previous predictions.
AMD's Lisa Su: The AI accelerator market in data centers is expected to reach $150 billion by 2027
According to the "Science and Technology Board Daily" on September 6, Lisa Su, CEO of AMD, recently stated that the demand for AI semiconductors is still strong, and AMD's data center business has been very strong in the second half of the year. She reiterated that the AI accelerator market in data centers is expected to reach $150 billion by 2027, and confirmed that AMD's upcoming MI300 AI accelerator product will be launched in the fourth quarter.
Microsoft: AIGC Tools Will Be the Fastest Software Business to Reach $10 Billion in Revenue
According to the "Science and Technology Board Daily" on September 8, Microsoft's Chief Financial Officer Amy Hood recently stated that the suite of generative artificial intelligence tools will be the fastest software business for the company to reach $10 billion in revenue. She said, "The opportunity is right in front of us, and AI enterprise software can attract customers." Microsoft has released a series of AI-driven products, including the $30 per user per month 365 Copilot service. Hood stated that customers have responded positively to 365 Copilot and feel that it has changed the way they work. She added that the target market for the 365 Copilot service is particularly large, as Microsoft 365 has 160 million users.
Research Report
Report: Generative AI Will Replace 2.4 Million Jobs in the US by 2030, Mainly Affecting White-Collar Workers
According to IT Home on September 10, according to the latest report from the analysis company Forrester, generative artificial intelligence will replace 2.4 million jobs in the US and affect 11 million jobs by 2030. These jobs are mainly white-collar, including technical writing, social science research, proofreading, and copywriting. Generative artificial intelligence refers to AI technology that can generate text, images, and other content, such as OpenAI's GPT-4 and ChatGPT.
Microsoft and OpenAI's Supercomputers Consume Significant Water Resources in Iowa
Microsoft disclosed in its latest environmental report that its global water consumption increased by 34% from 2021 to 2022, reaching nearly 1.7 billion gallons (equivalent to more than 2500 Olympic-sized swimming pools), which is a significant increase compared to previous years. Shaolei Ren, a researcher at the University of California, Riverside, believes that most of the growth is attributed to artificial intelligence, including significant investments in generative artificial intelligence and collaboration with OpenAI.
As a hub for Microsoft's cloud computing services, West Des Moines, Iowa, is the main source of water consumption. Additionally, the supercomputing data center used to train GPT-4 by OpenAI is also located in the area. External researchers estimate that asking 5 to 50 questions using ChatGPT consumes about 500 milliliters of water each time.
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。
