[Industrial Internet Weekly Report is a featured product released by titanium media TMTpost, which will integrate the most important cutting-edge trends, heavy policies and industry research reports in the fields of enterprise-level services, cloud computing, and big data this week. 】
Last week (July 10-16), the Cyberspace Administration of China and seven other departments issued the Interim Measures for the Management of Generative Artificial Intelligence Services, which will come into force on August 15, 2023. The Interim Measures stipulate that generative AI services shall be subject to inclusive, prudent and categorical supervision. Compared with the Measures for the Administration of Generative Artificial Intelligence Services (Draft for Comments) issued on April 11, the Interim Measures delete "content generated by generative AI shall be true and accurate, and measures shall be taken to prevent the generation of false information", "user profiling shall not be carried out", "user real-name system", and "generated content found during operation or reported by users that does not meet the requirements of these Measures shall be prevented from being regenerated within 3 months through model optimization training and other means".
Director of the Digital Economy and Legal Innovation Research Center of the University of International Business and Economics also participated in the formulation of the Interim Measures. From the perspective of licensing, the Interim Measures weaken the tone of the Draft for Comments, which emphasizes supervision, and Article 3 clarifies the combination of equal emphasis on development and security, innovation and governance according to law, and puts "development" and "innovation" before "security" and "governance" in the order of terms. "There were different opinions during the formulation of the Interim Measures, and in the end, development was taken as the main keynote and development was put first, which was recognized and adhered to."
Domestic news
China Electronics and Huawei decided to merge the Kunpeng ecosystem and PKS ecosystem
Recently, China Electronics Information Industry Group Co., Ltd. ("China Electronics") and Huawei Technologies Co., Ltd. ("Huawei") decided to merge the Kunpeng ecosystem and PKS ecosystem to jointly create a "Pengteng" ecosystem that supports both Kunpeng and Feiteng processors.
Megvii resumed the issuance registration process
According to the website of the Shanghai Stock Exchange, Megvii Technology Co., Ltd. has resumed the issuance registration process. Previously, Megvii suspended its issuance registration procedure because the financial information recorded in the listing application documents had expired and needed to be supplemented.
The Tsinghua ChatGLM-6B and ChatGLM2-6B models allow free commercial use
Zhipu AI and Tsinghua KEG Lab announced on the 14th that from now on, the ChatGLM-6B and ChatGLM2-6B weights are completely open to academic research, and free commercial use is allowed after completing the enterprise registration and obtaining authorization.
Huawei Releases New AI Storage Products in the Large Model Era
For the application of large models in different industries and scenarios, Huawei released OceanStor A310 Deep Learning Data Lake Storage and FusionCube A3000 Training/Push Super Convergence Appliance, which provide storage methods for basic model training, industry model training, and model training and inference in subdivided scenarios
Chongqing's first self-developed large language model "Zhaoyan" was released
The "Zhaoyan" large model release conference of Chongqing Institute of Artificial Intelligence of Shanghai Jiao Tong University was held in the Western (Chongqing) Science City on the 13th. At the meeting, the latest research and development results of the institute - "Zhaoyan" large model product were released, which is the first self-developed large language model in Chongqing, and it is also the first release of results of the institute since landing in Science City. "Zhaoyan" can present accurate and professional text output for users such as academic research, business analysis, and life consulting.
Jingdong Group's large model "Yanxi" was released today and is expected to be officially launched in August
JD.com-SW(09618.HK) held the 2023 JD.com Global Technology Explorer Conference and JD Cloud Summit. At the meeting, the Jingdong Yanxi model was officially unveiled. It is understood that the Yanxi model integrates 70% of general data and 30% of the native data of digital intelligence supply chain, and will be used in knowledge-intensive and task-oriented industrial scenarios such as retail, logistics, finance, health, and government affairs to solve real industry problems. JD.com Group also released a development platform and tool that matches the Yanxi model - Yanxi AI Development and Computing Platform, which has launched appointment registration and is expected to be officially launched in August.
Alibaba announced the launch of its multi-platform DAAS product "Antelope One", which will soon be connected to the large model
At the Antelope Digital Intelligence New Potential Energy Conference, Alibaba's enterprise intelligent service company Lingyang announced the launch of Antelope One's multi-platform, multi-channel one-stop DAAS product. Provide one-stop data services for analysis, performance, marketing and customer service in the process of enterprise operation. According to reports, the entire range of Antelope products will soon be fully upgraded to access the large model, and enterprises using Antelope One will be able to use the large model technology.
DingTalk launches enterprise-level HR SaaS integrated suite "DingTalk HCM+"
DingTalk released the enterprise-level HR SaaS integrated suite product "DingTalk HCM+", covering products and solution services such as "DingTalk OKR, DingTalk Organizational Brain and Enterprise Change Consulting", covering the entire link of organization management. Among them, some products are provided by ecological partners, DingTalk open base capabilities, and joint ecosystem to serve customers. DingTalk HCM+ is mainly aimed at DingTalk's strategic key customers, aiming to help large enterprises carry out organizational governance, better implement business strategies, activate organizations and individual employees, and achieve enterprise cost reduction and efficiency increase. DingTalk HCM+ will also be connected to AI capabilities.
Inspur and Intel released the AI server NF5698G7, which supports 8 Gaudi2 accelerators
Recently, Inspur and Intel released a new generation of AI server NF5698G7 for generative AI. The AI server supports eight OAM high-speed interconnected Gaudi2 accelerators, which have the advantages of high performance, high scalability, high energy efficiency, and open ecosystem, and will provide AI customers with powerful large model training and inference capabilities. It is reported that NF5698G7 adopts a high-expansion design, supports 12 PCIe5.0 X16 IO expansion, and supports a maximum interconnection bandwidth of 4800Gbps between nodes, which meets the communication requirements of parallel and data parallel of large model pipelines, and can be extended to Vancard AI clusters.
Wang Xiaochuan's Baichuan Intelligent released the 13 billion parameter general large language model Baichuan-13B series
Baichuan Intelligence, founded by Sogou founder Wang Xiaochuan, today officially released Baichuan-13B-Base, a general large-language model with parameters of 13 billion, Baichuan-13B-Chat and two quantitative versions of its INT4/INT8. This is the second general large language model released by Baichuan Intelligence, and on June 15, Baichuan Intelligent has launched its first Chinese and English language model with 7 billion parameters, Baichuan-7B.
Overseas news
Musk officially announced the establishment of an xAI artificial intelligence company, with team members from DeepMind, OpenAI and so on
Musk's team announced the creation of xAI Artificial Intelligence Company, which aims to understand "the true nature of the universe." The xAI team will hold a launch event on July 14 in the Twitter space to share more information. xAI is separate from Twitter's current parent company, Company X, but will work closely with Company X and Tesla.
xAI said on its website that its team will be led by Musk, and its executives have previously worked at a number of companies at the forefront of artificial intelligence, including OpenAI, Google's DeepMind, Microsoft, Tesla, Twitter, and academics such as the University of Toronto. Stakeholders have worked on well-known projects such as DeepMind's AlphaCode and OpenAI's GPT-3.5 and GPT-4.
US media: The CEOs of the three major US chip giants will go to Washington to lobby Biden to abandon new restrictions on exports to China
Bloomberg quoted people familiar with the matter as saying that the chief executive officers (CEOs) of the three major US chip giants will travel to Washington next week for talks with US government officials and lawmakers to lobby Biden to abandon new chip export restrictions to China. According to people familiar with the matter, the three chip giants are Intel, Qualcomm and Nvidia, and while they do not expect to stop all relevant actions of the US government, they perceive a window of opportunity and hope to convince the Biden administration that the escalation will hurt the White House's current engagement with China and diplomatic efforts to build a more productive relationship.
The United States and the Netherlands plan to further restrict cooperation with China? US media broke the news: ASML's cooperation with China faces stricter restrictions
Bloomberg quoted people familiar with the matter on the 14th to break the news that as the conflict between the United States and China further intensifies, the cooperation between Dutch lithography machine giant ASML and Chinese customers faces stricter restrictions. According to sources, Dutch export control rules will prohibit ASML from providing maintenance, repair services and supplying spare parts for restricted equipment without government approval. In addition to the above-mentioned controls introduced by the Dutch government, the U.S. government is expected to ban ASML from selling last-generation DUV (deep ultraviolet) equipment to about six Chinese factories without Washington's approval, the people familiar with the matter, who requested anonymity. (Global Network)
Musk said that Tesla is developing its own chips, which will never be called GPUs or H100s
Musk said at the xAI conference that Tesla is developing its own chip, but it will never be called a GPU, or described by 100s, H100s, etc. He also said that Dojo2 will make a move towards the big model.
KPMG will spend billions of dollars using Microsoft cloud and AI services over the next 5 years
KPMG and Microsoft recently announced a significant expansion of their global partnership, according to KPMG's official WeChat message. KPMG announced that it will spend billions of dollars using Microsoft cloud and AI services over the next 5 years. KPMG expects this partnership to generate US$12 billion in potential revenue growth opportunities for KPMG.
The Associated Press has a news data-sharing agreement with OpenAI to train generative AI
The Associated Press and OpenAI, the developer of ChatGPT, have reached a two-year agreement to share some news content and technology. This move is likely to set a precedent for journalism to collaborate with generative AI. As part of the agreement, OpenAI will have access to portions of the AP's text archives since 1985 to help train its AI algorithms. In return, the AP will gain access to OpenAI's technology and product expertise.
Financing mergers and acquisitions
It is reported that Silicon Valley chatbot startups Character.AI negotiate a new round of financing
According to reports, people familiar with the matter revealed that Character.AI is in talks with investors about a new round of financing. In March, the Silicon Valley company, founded by two former Google employees, said it had raised $150 million in a new funding round led by A16z (Andreessen Horowitz), a well-known venture capital firm, with a valuation of $1 billion. The company launched an early version of its software last September, allowing users to create and interact with chatbots that can play characters or celebrities like Musk.
AI startup Alphasense raised money at a valuation of $2.5 billion
AI-driven market research and data company Alphasense Inc. is reportedly raising money with a valuation of $2.5 billion, according to people familiar with the matter. The New York-based company is in talks to raise about $150 million, which will be used for the company's business growth opportunities, including potential acquisitions. The round of talks, which has not yet concluded, could be announced in a few weeks. Previously, Alphasense raised $100 million in April at a valuation of $1.8 billion from investors including Viking Global Investors, Alphabet Inc. and a subsidiary of Goldman Sachs.
Data security governance vendor "Hawin Technology" received tens of millions of yuan in Series A financing
Data security governance vendor "Haoin Technology" completed tens of millions of yuan in Series A financing. The investor in this round of financing is Qianxin Group and Sichuan Development Fund jointly established, with Voyage Capital acting as the exclusive financial advisor. According to reports, this is the second financing of this year after receiving Hefei high investment financing in April. Starting from data storage and traditional data governance services, Horn Technology is now positioned as a company that provides platform-level AI engine + data governance + data security integrated data services.
Big data technology service company "High-tech Security" completed Series B financing
Gaochuang Security completed the B round of financing, the specific amount of which was not disclosed, and the investor of this round of financing was Investment Control Donghai. This financing will provide financial support for high-innovation security and further promote development and innovation in the field of security. High-Tech Security is a typical security-based big data technology service enterprise.
"Guoweixin", a domestic digital EDA full-process solution provider, received the first round of hundreds of millions of yuan in external financing
Guoweixin, a domestic digital EDA full-process solution provider, announced the completion of the first round of hundreds of millions of yuan of external financing, which was jointly led by Anxin Qianhong and Guangzhou Lifeng. The financing funds will be mainly used for SMIC's in-depth construction of the full-process EDA tool system for domestic digital chips, and accelerate the realization of large-scale mass production, application and strengthening of core technical team construction.
Policy & Trends
State-owned Assets Supervision and Administration Commission: Promote the integration of manufacturing with the Internet, big data, artificial intelligence and other technologies, and strengthen the construction of new digital infrastructure
According to the website of the State-owned Assets Supervision and Administration Commission, from July 13 to 14, the State-owned Assets Supervision and Administration Commission of the State Council held a seminar for heads of central enterprises. The meeting pointed out that it is necessary to adhere to the efforts of strategic emerging industries and traditional industries at both ends and accelerate the construction of a modern industrial system. It is necessary to go all out to develop strategic emerging industries, form a systematic layout as soon as possible on the basis of scientific demonstration, adhere to long-term and stable investment, strengthen coordination and cooperation with various ownership enterprises in the industrial chain while developing, increase major investment, industrial mergers and acquisitions, technology research and development, standard formulation, etc., and step up efforts to build a number of strategic emerging industrial clusters and industrial leaders with international competitiveness. It is necessary to increase the intensity of transformation and upgrading of traditional industries, promote the high-end, intelligent and green of traditional industries, deeply implement industrial infrastructure reengineering projects and high-quality development of manufacturing industries, accelerate the renewal and transformation of key processes and core equipment, promote the integration of manufacturing with the Internet, big data, artificial intelligence and other technologies, strengthen the construction of new digital infrastructure, effectively and orderly promote the implementation of carbon peaking plans, and accelerate the formation of spatial patterns, industrial structures and production methods that save resources and protect the environment.
The global computing power report was released: the manufacturing industry surpassed the financial industry for the first time, and the computing power level rose to the second place
The "Global Computing Power Index Assessment Report" jointly compiled by Tsinghua University and other institutions was released. The report points out that despite the general slowdown in global GDP growth, the digital economy remains strong, and computing power has become an important driver of economic growth. Data show that from 2022 to 2023, the proportion of digital economy in GDP in major countries in the world will continue to increase, and it is expected to reach 54% by 2026. The report shows that at present, the top three industries in the global computing power level are the Internet, manufacturing, and finance. In the past year, the level of computing power in the manufacturing industry surpassed the financial industry for the first time and rose to second place. From China's perspective, the average annual growth rate of manufacturing IT spending in the next five years will reach 16.2%, significantly higher than that of the rest of the world.
Institutions: The global RAN market is expected to decline at a CAGR of 1% over the next five years
According to the latest report from market research firm Dell'Oro Group, after growing by 40%-50% between 2017 and 2021, the global RAN market revenue flattened out in 2022, and this trend continued until the first quarter of 2023. The agency also said that 5G-Advanced is not expected to trigger a new capital expenditure cycle. The global RAN market is expected to decline at a CAGR of 1% over the next five years.
IDC: China's data volume will grow to 76.6 zettabytes by 2027
According to IDC's latest report, China's data volume will grow from 23.88 zettabytes in 2022 to 76.6 zettabytes in 2027, with an average annual growth rate of 26.3%, of which government, media, professional services, retail, healthcare, and finance are the main distribution areas.
IDC: China's IoT market spending is gradually rising, and it is expected to rank first in the world by 2027
IDC predicts that China's IoT spending will approach $300 billion in 2027, ranking first in the world and accounting for about 1/4 of the total global IoT investment. In addition, China's IoT IT spending grew steadily at a five-year CAGR of 13.2%, exceeding the global average.