[Industrial Internet Weekly] China has released 79 large-scale models with 1 billion parameters or above; Power Computer responded to the core question; Alibaba Cloud Intelligent released audio and video AI Tongyi Listening

[Industrial Internet Weekly Report is a special product released by Titanium Media TMTpost. It will integrate the most important cutting-edge trends, major policies and industry research reports in the fields of enterprise-level services, cloud computing and big data this week. 】

Domestic News

Wu Hequan, Academician of the Chinese Academy of Engineering: 5G development has reached a critical turning point and is expected to enter the investment return period

At the main forum of the 2023 China International Information and Communications Exhibition, Wu Hequan, an academician of the Chinese Academy of Engineering, pointed out that the development of 5G has reached a critical turning point. It has now covered all counties and urban areas across the country, and the number of 5G users has accounted for more than 30% of the total mobile users. It is expected that The proportion will exceed 50% after 2024. After network deployment and construction reaches a certain scale, 5G will enter the investment return period. Therefore, it is necessary to upgrade 5G capabilities to 5G-A so that users can fully experience the advantages of 5G in consumer applications.

The prices of some Alibaba Cloud cloud server ECS products will be reduced, with the maximum decrease of 32.5%

According to Alibaba Cloud’s official website, Alibaba Cloud will adjust the prices of some ECS products. The new prices will take effect at 00:00 on June 6, 2023, Beijing time. Specifically, for the A10 c8g1 series and c16g1 series, the list price in mainland China (excluding Heyuan) is reduced by 25%, and the Heyuan city price is reduced by 32.5%; for the A10 c32g1 series and A10 bare metal instances, the list price in mainland China (excluding Heyuan) The list price was reduced by 11%, and the Heyuan Dungeon price was reduced by 19.9%.

Alibaba Cloud Intelligence releases audio and video AI Tongyi Listening, which will be officially tested on June 1st

Alibaba Cloud Intelligence released Tongyi Tingwu today and will officially launch public beta on June 1st. Zhou Jingren, CTO of Alibaba Cloud Intelligence, said that Tongyi Tingwu is an AI assistant based on Tongyi Qianwen large model and audio and video AI model. It can perform real-time speech recognition, realize real-time recording of conversations, multi-language translation, speech summary, extraction of PPT, and content. Regularity etc.

China Telecom Lingang Computing Power Intelligent Computing Public Service Platform and Domestic GPU Innovation Joint Base Released

At the Lingang New Area Intelligent Computing Conference, China Telecom’s Lingang Computing Power Intelligent Computing Public Service Platform and Domestic GPU Innovation Joint Base were officially released. China Telecom established Lingang Computing Power (Shanghai) Technology Co., Ltd. to carry out the construction of Lingang Computing Power Park and will launch 40,000 high-power racks suitable for intelligent computing and supercomputing in batches. It is reported that the establishment of China Telecom’s Lingang Computing Power Intelligent Computing Public Service Platform and Domestic GPU Innovation Joint Base will provide many enterprises in Lingang, Shanghai, and the Yangtze River Delta with better and more inclusive intelligent computing public services.

Server manufacturers: GPU shortage will continue at least until next year

According to the Shanghai Securities News, industry insiders learned that boosted by the boom in large models, the demand for computing power in the market is strong, and there is a continued shortage of computing power servers, and the distribution cycle is relatively long. Server manufacturers generally expect to wait more than six months to get the latest GPUs, and the GPU shortage will continue until at least next year. Data shows that artificial intelligence computing power, dominated by GPUs, has been growing rapidly in recent years. In 2016, the proportion of domestic artificial intelligence computing power in the entire computing power structure was only about 3%, and by 2022, this proportion will exceed 50%.

my country’s first national standard for quantum information technology was officially released

my country's first national standard in the field of quantum information technology, "Quantum Computing Terminology and Definitions", was approved by the State Administration for Market Regulation (National Standardization Administration Committee) and officially released. It will be implemented on December 1, 2023. This standard standardizes terms and definitions related to the general foundation, hardware, software and applications of quantum computing, and provides normative guidance for the preparation of scientific research reports, standard formulation, and technical document preparation in the field of quantum computing.

Qiu Xipeng, head of the MOSS system at Fudan University: There will be plans to launch MOSS models with greater computing power and stronger reasoning capabilities

Qiu Xipeng, head of the MOSS system at Fudan University, said that the MOSS system currently ranks high in downloads in the open source community, and has more than 10,000 following users on the Github platform. "So far, the MOSS backend has received applications from hundreds of enterprises, most of which are small and medium-sized enterprises, covering traditional industries, intelligent information services, finance, medical care, education, etc. The main function of the MOSS model currently focuses on helping enterprises Apply large models to specific fields at a lower cost," Qiu Xipeng said. Judging from the feedback on the use of the MOSS model, Qiu Xipeng bluntly said that the current chat function of the model has achieved a good interactive experience, but it still needs further improvement in handling complex problems. Qiu Xipeng revealed that Fudan will plan to launch a MOSS model with greater computing power and stronger reasoning capabilities.

Baode Computer responded to the question: Baoxin’s first CPU is a customized CPU product launched with the support of Intel

Power Computer Systems Co., Ltd. issued a statement on May 31 about the first CPU of Baoxin. The document stated that "Borex's first CPU is a customized CPU product launched with the support of Intel Corporation; Baoxin CPU is mainly used for branded PC terminals in the commercial market and has not declared any projects or subsidies to relevant government departments."

Previously, after the release of the Baoxin processor in early May, there was constant controversy about the Baoxin processor copying x86 and defrauding subsidies. GK5 testing showed that it was actually a replica of the i3-10105. In the face of doubts, Li Ruijie, chairman of the Board of Directors, responded on Weibo: "Standing on the shoulders of the Intel giant and going to the next level, we will have our own way to solve the original code problem, not just building a bridge." He also said , let those who fish in troubled waters calm down and stick to the plan. I believe it will be a new force in China's chips.

Guodun Quantum: Released the "Zuchongzhihao" quantum computing cloud platform, connected to a 176-bit quantum computer

Guodun Quantum released a new generation of quantum computing cloud platform, which is connected to the same 176-bit superconducting quantum computer as "Zuchongzhihao". According to reports, the system added 110 coupling bits of control interfaces based on the original 66-bit chip of the "Zuchongzhihao", setting a new record for the number of bits of a superconducting quantum computer in China's cloud platform.

Qinhuai Data Q1 2023: Net profit increased by 167.5% year-on-year

Chindata Group (NASDAQ: CD) announced its unaudited financial report for the first quarter of 2023. The quarterly report shows that Chindata’s revenue in the first quarter was 1.444 billion yuan, a year-on-year increase of 56.8%; net profit was 253 million yuan, a year-on-year increase of 167.5%, and the net profit margin was 17.5%; adjusted EBITDA was 814 million yuan, a year-on-year increase of 64.6%, and adjusted EBITDA Profit rate is 56.4%; ROIC (return on assets) is 18.7%. At the same time, Qinhuai Data raised its full-year guidance for adjusted EBITDA to 3.1-3.22 billion yuan. At the operational level, as of March this year, Chindata has 33 data centers in operation or under construction globally, with an operating capacity of 639MW; the total IT capacity is 898MW, and the total capacity is expected to exceed 1,000MW in the next year.

360 releases 360 smart brain vision model, Zhou Hongyi: The combination of multi-modal large models and the Internet of Things will become the next trend

360 (hereinafter referred to as "360") Smart Life Group released new visual large-scale models and AI hardware products. Zhou Hongyi, founder of 360 Group, said at the meeting, "The original AIoT was only vertical AI, not general AI. After large-scale models Empowered AIoT is 'real AI'." Zhou Hongyi believes that all software, APPs, websites, and all industries are worthy of being reshaped with large models, and smart hardware is a hardware-based APP. "Large models will bring about a new industrial revolution." Judging from the development trend of large models, multi-modality is the only way for the development of large models. The most important change of GPT-4 is that it has multi-modal processing capabilities. Therefore, Zhou Hongyi predicted that the combination of multi-modal large models and the Internet of Things will become the next trend.

DingTalk slash “/” launches invitation test for enterprise users

For enterprise users, DingTalk slash "/" officially launches the invitation test. This invitation test opens up scenarios such as DingTalk group chat, documents, video conferencing and application development, involving Wen Sheng texts, Wen Sheng pictures, abstracts and summaries. With various AI capabilities such as application generation, enter "/" to allow AI to automatically develop applications, write articles, draw, organize minutes, etc.

International News

Microsoft Office 365 AI Assistant is priced 40% higher than the regular version, and 100 customers have already paid US$100,000 each

According to market news, more than 600 of Microsoft's largest customers, including Bank of America, Walmart, Ford and Accenture, are expected to trial the artificial intelligence capabilities in Microsoft Office 365, and at least 100 of them have already added an additional 1,000 customers. Each subscription account pays an annual fee of up to $100,000. Compared with the classic version, the price of this AI version of Office 365 is at least 40% higher, and it has value-added functions such as automatically writing text in Word documents and automatically creating PPT. With 600 customers each paying for 1,000 accounts, Microsoft will earn $60 million in revenue.

The Rothschild family, a century-old financial giant, "reduces its holdings" in NVIDIA

While the world is going crazy about AI, the Rothschild family, a century-old European financial giant, is quietly reducing its positions. The global chief investment officer (CIO) of Edmond de Rothschild, an asset management institution, revealed that the company has been overweight Nvidia since the end of 2020, but has now taken some profits and its position is now "much smaller." Like Edmond de Rothschild, who also chose to cash out at high prices is the famous contemporary investment valuation expert Aswath Damodaran, known as the "Godfather of Valuation". According to reports, Damodaran has been investing in Nvidia since 2017 and has now sold his shares in the company.

NVIDIA launches new AI model Neuralangelo, which can convert videos into high-precision 3D models

NVIDIA announced the launch of an AI model called Neuralangelo that can convert video footage into 3D models with rich levels of detail. NVIDIA emphasizes that Neuralangelo is accurate and universal. Accuracy means that the model can accurately represent the texture of complex materials, such as roof tiles, glass panes, and smooth marble; universality means that creators can create realistic objects simply by using video footage shot with smartphones , and integrate it into their projects.

Microsoft signs artificial intelligence computing power agreement with startup CoreWeave

Microsoft signed an artificial intelligence computing power agreement with the startup CoreWeave, which may be worth at least billions of dollars. Previously, Nvidia had invested US$100 million in CoreWeave, which was valued at US$2 billion in April. CoreWeave announced Wednesday it has secured $200 million in financing.

Broadcom’s revenue in the second fiscal quarter was US$8.733 billion, and net profit increased by 34% year-on-year

Broadcom today released the company’s second quarter financial report for fiscal year 2023. The report shows that Broadcom’s net revenue in the second fiscal quarter was US$8.733 billion, an increase of 8% compared with US$8.103 billion in the same period last year, but a decrease compared with US$8.915 billion in the previous quarter; net profit was US$3.481 billion. , an increase of 34% compared with US$2.590 billion in the same period last year, and US$3.774 billion in the previous quarter; net profit attributable to ordinary shareholders was US$3.481 billion, compared with US$2.515 billion in the same period last year, and US$2.515 billion in the previous quarter. Quarterly was $3.774 billion.

Nvidia CEO Jen-Hsun Huang: H100 will be manufactured exclusively by TSMC and will not consider adding a second wafer foundry

NVIDIA CEO Jen-Hsun Huang said that H100 is produced exclusively by TSMC and will not consider adding a second wafer foundry. The main reason is that in the entire design foundry process, "it is very difficult to do it once, and it will be difficult to do it separately." It’s more difficult for two OEMs to do it twice.” Nvidia and TSMC have begun 3/2nm cooperation planning, and Huang Renxun also confirmed for the first time that the next-generation AI chip has placed an order with TSMC. At least in the AI ​​GPU series, TSMC will continue to "monopolize large orders" within a few years. In addition to NVIDIA's A100/A800 and H100, which have all placed orders with TSMC, the RTX 40 series has also fully returned to TSMC's 4nm process, while the RTX 30 series uses Samsung's 8nm process. The current cooperation between NVIDIA and Samsung is mainly focused on memory.

Character.AI downloads exceeded 1.7 million in the first week: benchmarking ChatGPT, created by former Google AI experts

Recently, Character.AI, an artificial intelligence application that claims to compete with ChatGPT, announced that in less than a week after it was launched for global iOS and Android users on May 23, its App installations have exceeded 1.7 million . In comparison, the ChatGPT mobile app was downloaded more than 500,000 times in its first six days.

It is understood that Character.AI is a chatbot platform that provides customizable artificial intelligence assistants with distinctive personalities and the ability for users to create their own characters. The platform is built by Character Technologies, a Palo Alto-based startup founded in November 2021 by artificial intelligence experts Noam Chazer and Daniel de Freitas, who previously led Google’s A research team that built LaMDA (Language Model for Dialogue Applications).

Salesforce’s first-quarter revenue was US$8.25 billion, analysts expected US$8.18 billion

Salesforce’s revenue in the first quarter was US$8.25 billion, analysts expected US$8.18 billion; adjusted earnings per share in the first quarter was US$1.69; adjusted operating profit margin in the first quarter was 27.6%; second quarter revenue is expected to be US$8.51 billion to US$8.53 billion, analysts expected US$8.49 billion; full-year revenue is still forecast to be US$34.5 billion to US$34.7 billion; full-year adjusted earnings per share are expected to be US$7.41 to US$7.43.

C3.ai’s fourth quarter revenue was US$72.4 million, analysts expected US$71.4 million

C3.ai’s fourth quarter adjusted loss per share was $0.13, analysts expected a loss of $0.17; fourth quarter revenue was $72.4 million, analysts expected $71.4 million; first quarter revenue was expected to be $7,000 million to US$72.5 million, analysts expected US$72.1 million; revenue for fiscal 2024 is expected to be US$295 million to US$320 million, while analysts expected US$317 million.

Microsoft plans to launch a version of ChatGPT that runs on dedicated cloud servers at a price ten times that of the regular version

According to reports, people familiar with the matter revealed that later this quarter, Microsoft's Azure cloud server division plans to sell a version of ChatGPT that runs on a dedicated cloud server, and the data of this version will be kept separately from the data of other customers. This was done to reassure customers that their secrets would not be leaked to ChatGPT's main system, people familiar with the matter said. But the cost of this product may be as much as 10 times that of the regular version of ChatGPT that customers currently use.

NVIDIA: GH200 chip is a "giant GPU", the first batch will be provided to Google Cloud, META and Microsoft

Nvidia CEO Jensen Huang said that the GH200 chip is a "giant GPU". Google Cloud, META and Microsoft will be the first to get the GH200.

Financing M&A

News: AI startup MiniMax is close to completing a new round of financing exceeding US$250 million, with Tencent-related entities participating in the investment

According to market news, people familiar with the matter said that MiniMax, a Chinese AI startup focusing on general large models, is close to completing a round of financing exceeding US$250 million, and its valuation is expected to reach nearly US$1.2 billion. The company is developing AI solutions similar to ChatGPT. MiniMax’s latest round of financing attracted new investors, including entities linked to Tencent, two sources said. It is reported that MiniMax attracted miHoYo to participate in its previous financing. Other early investors include Yunqi Capital and Mingshi Capital.

Guochuang Semiconductor will transfer its IC and SiC component business to Hon Hai for NT$204 million

Hon Hai and Yageo stated that the joint venture between the two parties, Guochuang Semiconductor, will transfer its integrated circuit (IC) and silicon carbide component/module product business to Hon Hai’s newly established IC design subsidiary for NT$204 million. . Hon Hai will focus on automotive ICs, while Yageo will focus on developing MOSFET products. Yageo and Hon Hai will adjust their shareholding ratios in Guochuang Semiconductor to 55% and 45% respectively. Chen Taiming, chairman of Yageo, was appointed chairman of Guochuang Semiconductor and Fuding Advanced.

Grayscale Security completed tens of millions of yuan in Pre-A round of financing

Beijing Huidu Technology Co., Ltd. (hereinafter referred to as Huidu Security) completed tens of millions of yuan in Pre-A round financing. The investor in this round is Anheng Information, and Voyage Capital serves as the exclusive financial advisor. This round of financing will be mainly used for product research and development, expert knowledge research and market expansion.

Policy and Data

Minister of Industry and Information Technology Jin Zhuanglong: Will accelerate the construction of virtual private networks in the 5G industry and comprehensively promote 6G technology research and development

Minister Jin Zhuanglong of the Ministry of Industry and Information Technology said at the 31st China International Information and Communications Technology Exhibition held today that the Ministry of Industry and Information Technology will accelerate the construction of virtual private networks in the 5G industry, in-depth implementation of the 5G Application Sail Action, further enrich and expand 5G application scenarios, and deepen the integration of industrial Internet , improve the industrial Internet technology system, standard system, and application system, build a number of 5G factories, and vigorously promote the intelligent, green, and integrated manufacturing industry. At the same time, we will accelerate the cultivation of emerging industries to seize new advantages in the future, continue to enhance the advantages of the entire industry chain in mobile communications, optical communications and other fields, consolidate the leading position in advantageous industries, proactively deploy the next generation Internet and other cutting-edge fields, comprehensively promote 6G technology research and development, and create good development environment, guide and support enterprises to increase investment in research and development, and accelerate the development of emerging digital industries such as artificial intelligence, blockchain, cloud computing, and network security.

Report: China has released 79 large models with more than 1 billion parameters

At the "Artificial Intelligence Large Model Development Forum", a parallel forum of the 2023 Zhongguancun Forum, Zhao Zhiyun, director of the China Institute of Scientific and Technological Information, released the "China Artificial Intelligence Large Model Map Research Report". Zhao Zhiyun explained that various technical routes for China's large models are making breakthroughs in parallel, especially in aspects such as natural language understanding, machine vision, and multi-modality. Several large models with influence in the industry have emerged. According to incomplete statistics, so far, China has released 79 large-scale models with more than 1 billion parameters.

Bloomberg Industry Research: The artificial intelligence market is expected to reach US$1.3 trillion by 2032

The release of consumer-focused artificial intelligence (AI) systems such as OpenAl Inc.’s ChatGPT and Google’s Bard will fuel a decade-long boom that will propel the generative AI market from $40 billion last year to 2032 .3 trillion US dollars per year. According to a new report from Bloomberg Intelligence analysts led by Mandeep Singh, the industry is likely to grow at a compound annual growth rate of about 42% over the next decade, driven first by training infrastructure and then by large language model inference machines, advertising and other service promotions. The report said Amazon's cloud unit, Google parent company Alphabet Inc., Nvidia, and Microsoft, which has invested billions of dollars in OpenAI, may be among the biggest winners of the artificial intelligence boom.

IDC: The global industrial security market will reach US$6.7 billion by 2026

The IDC report predicts that by 2026, the global industrial security market will reach US$6.7 billion, with a five-year compound growth rate of 28.4%. China's industrial Internet security market will also achieve rapid development driven by policies and demand.

The above is the detailed content of [Industrial Internet Weekly] China has released 79 large-scale models with 1 billion parameters or above; Power Computer responded to the core question; Alibaba Cloud Intelligent released audio and video AI Tongyi Listening. For more information, please follow other related articles on the PHP Chinese website!