JD.com's Q2 revenue was 287.9 billion; Lenovo's Q1 revenue was 90.3 billion; Gartner: AIGC has reached the peak of the bubble; Intel terminated the acquisition of Tower Semiconductor丨Daily events...


ebb67ffacd8c32c2c56165dee09a7a93.png

59125e265937290895fb0fc243028a28.png




Big data industry innovation service media

——Focus on data and change business


Entreprise's news

  • Wenxin Yiyan has launched 5 native plug-ins including Baidu search, document summary, and text-to-video conversion

At the WAVE SUMMIT Deep Learning Developer Conference held by the National Engineering Research Center for Deep Learning Technology and Application on August 16, Baidu Chief Technology Officer Wang Haifeng said at the conference that the Wenxin Yiyan large model has now launched 5 native plug-ins: Baidu Search, browse documents, E-language easy-to-map, talk about diagrams, and one-shot videos. At the same time, Wenxin Yiyan now supports the use of 3 plug-ins at the same time. The official said that it can better adapt to the needs of the scene while expanding the capability boundary of the large model.

  • Intel Announces Termination of Acquisition of High Tower Semiconductor

On August 17, Intel announced that due to the inability to obtain the regulatory approval required by the merger agreement in time, Intel has reached an agreement with Tower Semiconductor to terminate the previously disclosed acquisition agreement. Under the terms of the merger agreement, Intel will pay Hightower a termination fee of $353 million.

  • General Motors leads investment in a mysterious AI start-up company focusing on "AI + electric vehicle battery research and development"

General Motors announced on Wednesday that it is leading Mitra Chem's $60 million Series B financing project. Under the partnership agreement, General Motors and Mitra Chem will develop advanced iron-based cathode active materials, such as lithium manganese iron phosphate, to power affordable electric vehicle batteries compatible with the Ultium platform. GM's financial backing is intended to help Mitra Chem scale up its current operations and speed up the full-scale introduction of its new battery material formulations to the market.

  • Adobe's AI photo editing tool Express launched

According to news on August 17, Adobe’s AI image design and editing tool Express has ended its beta and is officially open to all users.

The official version of Express brings many new features, you can use the AI ​​generation function supported by Adobe Firefly to generate different text effects and pictures, and modify different content. In addition, the official version of Express can also combine videos, pictures and music to easily complete video production. The desktop version of Adobe Express is now officially live, and the mobile version of the application will be updated in the near future.

  • Fii: Started to develop and mass-produce high-performance AI servers such as Nvidia's H100 and H800 for customers this year

According to news on August 17, Fii stated on the interactive platform that Fii began to develop and mass-produce high-performance AI servers such as Nvidia's H100 and H800 for customers this year. In addition, the company's cloud game server has also been officially shipped in March this year. The company has been in the AI ​​field for a long time, and continues to promote the development and shipment of new products in response to customer needs. In the future, the company will continue to increase the investment in AI server research and development, continuously optimize the product structure, and lead the development opportunities of the AI ​​track.

  • Microsoft Windows 11 app store launches AI summary feature

According to news on August 17, the Microsoft App Store has launched a feature called "AI Summary", which can generate a summary based on the user's evaluation of the application.

This feature is designed to provide users with a quick and easy way to understand user reviews and opinions on the app, and provide a reference for users who want to download the app. The feature can summarize thousands of online comments into a concise paragraph in seconds, highlighting the most important points. The feature is currently only available to US Windows 11 users, but is expected to roll out to more countries and regions in the future.

  • Tencent Cloud Smart Copywriting Product Launch

According to news on August 17, at the 2023 China Digital Economy Innovation and Development Conference, Tencent Cloud Intelligence officially launched the "copywriting" product, which is based on Tencent's large-scale model capability research and development, and is designed for advertising platforms, UGC platforms, SaaS service companies, etc. Customers, aiming to provide accurate, efficient and comprehensive business copywriting tools. This new product is expected to help enterprises customize and generate more targeted copywriting content in multiple fields such as marketing, creativity, and efficiency, thereby significantly improving the quality and efficiency of copywriting creation.

  • OpenAI Acquires Digital Products Company Global Illumination

News on August 17th, in the early hours of Thursday, Beijing time, OpenAI quietly issued an announcement on its official website, announcing the acquisition of the start-up Global Illumination and bringing the entire team under its command. New members have begun to participate in the development of core products including ChatGPT. This deal is also the first public acquisition of OpenAI since its founding in 2015, and the specific terms of the deal were not made public. Global Illumination is a start-up founded in 2021 in New York by Thomas Dimson, Taylor Gordon and Joey Flynn, all three of whom worked at Zuckerberg's Instagram. The company's main business is using artificial intelligence to create smart tools, digital infrastructure and digital experiences.

  • Zhiji Auto released the roadmap for the implementation of smart driving products in the next 3-5 years

On August 16, Guo Hui, Chief Scientist of Intelligent Driving of Zhiji Automobile, announced that Zhiji will officially launch the "IM AD City NOA" public test in October this year. IM AD intelligent driving will be fully data-driven and will be realized in the next 3-5 years. Autopilot in most scenarios is heading towards the end of autopilot. Judging from the road map, in the next 3-5 years, Zhiji Automobile will launch smart driving products to accelerate the launch, and intelligent driving products such as high-precision map NOA, city NOA, and commuting mode will be intensively launched. Among them, "Go to high-precision map NOA", IM AD will start the public test in September; "City NOA", the internal test has been launched in April this year, and it is progressing smoothly, and the public test will be launched in October; "Commuting mode", it is expected It will quickly cover 100+ cities across the country in 2024; in 2025, IM AD will enter the era of Door to Door (all-scenario commuting).

  • Alibaba DAMO Academy released a general model of cancer

On August 16, Alibaba Dharma Institute released a general model for multi-cancer image analysis, which can solve the problems of multiple tumor images that were difficult to detect, segment and diagnose before, and is applicable to 8 mainstream high-incidence and high-lethal cancers and tumors in related organs Subtype. This will provide radiologists with AI-assisted diagnosis support, especially in clinical scenarios such as cancer recurrence and distant metastasis.

  • Ningde era releases Shenxing supercharged battery to fully open the era of supercharged new energy vehicles

On August 16, Ningde Times released the world's first 4C supercharged battery that uses lithium iron phosphate material and can be mass-produced--Shenxing supercharged battery, which achieved the ultra-fast "charging for 10 minutes and a battery life of 400 kilometers". The charging speed and the cruising range of more than 700 kilometers greatly alleviate the user's anxiety about energy replenishment and fully open the era of overcharging of new energy vehicles.

e6b10a4df714e6b9990dcb74411a981e.png

  • OpenAI announced the introduction of GPT-4 into the content review system to shorten the work cycle

On August 16th, according to foreign media reports, OpenAI announced the use of a new generation of large-scale language model GPT-4 for content policy development and content review decisions to achieve more consistent labeling, faster feedback loops for policy optimization, and fewer of human reviewers participated. OpenAI says content moderation systems using GPT-4 can iterate policy changes more quickly, reducing work cycles from months to hours. Anyone with access to the OpenAI API can take this approach to create their own AI-assisted moderation system. According to OpenAI research, GPT-4 can in some cases perform certain content moderation tasks better than humans. The company believes its technology can help solve one of the tech world's thorniest problems: content moderation at scale.

  • Weimai releases CareGPT, the first large language model health management application in China

According to news on August 17, Weimai, a whole disease course management platform, released CareGPT, the first large language model application in the field of health management in China. According to reports, unlike general-purpose large language model products, CareGPT is mainly committed to exerting the value of health management in real medical service scenarios, and realizing the full-cycle intelligent health management capabilities of prevention, consultation, appointment, and rehabilitation. CareGPT combines language large model AI technology with a series of engineering optimization technologies and full disease course management. The current parameter scale is 7 billion, which can support multi-modal input and output in medical and health scenarios.

  • Tabnine launches AI assistant Tabnine Chat to speed up code delivery

On August 17th, Tabnine released the enterprise-level code interaction application Tabnine Chat, which can interact with its AI model in natural language to speed up code delivery. Developers can integrate the application in the IDE to interpret, search the code base and generate new code using natural language specifications. Tabnine Chat focuses on security compliance, protects private code bases, and supports multiple programming languages. It is embedded in the IDE and integrated with the code developers are writing, facilitating a seamless dialogue between developers and AI. At the same time, enterprise users can link the code base with the application, improving productivity and opening up new possibilities for advanced coding interactions.

financial report

  • JD.com: Revenue in the second quarter was 287.9 ​​billion yuan, a year-on-year increase of 7.6%

On August 17, JD.com released its financial report for the second quarter of 2023. The financial report shows that this quarter's revenue was 287.9 ​​billion yuan, a year-on-year increase of 7.6%. Among them, the revenue of electronic products and household appliances increased by 11.4% year-on-year; service revenue reached 54.1 billion yuan, a year-on-year increase of 30.1%, of which logistics and other service revenues increased by 51.5% year-on-year

  • Lenovo Group released its Q1 financial report: revenue of 90.3 billion yuan

On August 17, Lenovo Group announced its results for the first quarter of the 2023/24 fiscal year ending June 30, 2023: revenue of 90.3 billion yuan and net profit of 1.33 billion yuan. The Group's service-oriented transformation business continued to grow, and the proportion of businesses other than PC in the Group's overall revenue further increased to 41.4%.

Policy/Report

  • Gartner: Generative AI has reached the peak of the bubble and is about to fall into the trough

According to news on August 17, market research firm Gartner recently released a report, judging from the hype cycle (Hype Cycle) of generative AI, it has reached the "peak of inflated expectations". The agency divides the hype cycle of generative AI into Innovation Trigger, Peak of Inflated Expectations, Trough of Disillusionment, and Slope of Enlightenment. Enlightenment) and "Plateau of Productivity" (Plateau of Productivity) 5 stages. The agency believes that generative AI is currently in phase 2, and will quickly fall into a trough due to high expectations from consumers and the media. The agency predicts that it will enter the fifth stage in the next 2-5 years, when AI enhancement, cloud sustainability, causal AI, post-quantum cryptography, GitOps, etc. will flourish.

  • Chongqing: Accelerate the construction of the city's autonomous driving policy pilot zone

According to news on August 17, the Chongqing Municipal Government Information Office held the first press conference of the 2023 China International Smart Industry Expo today. Luo Li, deputy director of the Chongqing Municipal Economic and Information Commission, introduced at the press conference that Chongqing strives to build a world-class intelligent networked new energy vehicle industry cluster by 2030. Focusing on this goal, we will continue to make efforts from the following three aspects to promote the high-quality development of Chongqing's intelligent and connected new energy vehicle industry. First, strengthen the leadership of the vehicle leader. Second, speed up the clustering of core components. Third, accelerate the construction of an innovative application ecosystem.

  • Shenzhen: Cultivate characteristic industrial clusters of small and medium-sized enterprises in emerging fields such as AI

On August 16, Shenzhen officially issued "Several Measures on Promoting the Private Economy to Be Larger, Better and Stronger". Shenzhen will vigorously cultivate a group of leading private enterprises in emerging fields such as new energy vehicles, artificial intelligence, and new energy storage. Approve national and provincial small and medium-sized enterprise characteristic industrial clusters; support private enterprises to be included in the national and provincial top 100 private enterprises list, and reward specialized and new private enterprises selected at the national and provincial levels; promote small and medium-sized enterprises by the end of 2023 Policy measures for the development of characteristic industrial clusters.

Investment and Financing

  • Netthink Technology completed the A round of financing

According to the official account of Netthink Technology, recently, Netthink Technology has completed the A round of financing led by Guangzhou Industrial Investment, followed by Tianhe Fund and Jianzhi Group, with a cumulative financing amount of nearly 100 million yuan.

  • "Klarity Health" received $10 million in Series A financing

Recently, "Klarity Health", a telemedicine company headquartered in California, USA, has received US$10 million in Series A financing, led by Atypical Venture (Ruby Lu), followed by Mucker Capital and Uphonest Capital. Zhou Zhehao, founder and CEO of Klarity Health, said that this round of financing will be used for talent recruitment and marketing.

  • Guohua Investment Leads the Pre-A Round of Financing of Qi Na New Energy

Recently, according to the official account of Guohua Investment, Wuhan Qi Na New Energy Technology Co., Ltd. (hereinafter referred to as "Qi Na New Energy"), a R&D and manufacturer of sodium cathode materials, has received another tens of millions of yuan in Pre-A round financing. For Guohua Investment, GF Xinde, Dingxiang Capital, and Kaixian Capital, the financing funds will be mainly used for the construction of 10,000-ton mass production lines.

other

  • Beijing Economic Development Zone: 640 self-driving vehicles have been deployed, and the test mileage exceeds 17 million kilometers

The press conference on the construction of the "two districts" in the first half of the year held on August 17 was informed that driven by the construction of autonomous driving 3.0, the Beijing Economic Development Zone has completed the construction of digital infrastructure at 200 intersections and deployed 640 autonomous vehicles. The test mileage exceeds 17 million kilometers, and the world's first real-scenario-based vehicle-road collaborative autonomous driving data set and intelligent networked roadside operating system were released.

  • CITIC Securities: Data closed-loop capability in the era of autonomous driving 3.0 will be the winner

On August 17, CITIC Securities Research Report pointed out that as high-end autonomous driving moves towards urban mass production scenarios, the architecture of system perception, decision-making, execution, and communication will usher in a new upgrade. At present, players use "BEV+Transformer+data closed-loop" as the core architecture of the latest generation of autonomous driving mass production systems. The algorithm paradigm and technical route are temporarily converged. Data volume and data closed-loop capabilities may become the key to victory in the second half from 1 to N. . Domestic players are still far behind Tesla in terms of mass-produced car data closed-loop, and need to continue to accumulate and polish engineering experience to turn the data flywheel. OEMs with strong self-research capabilities and suppliers with rich data experience or surveying and mapping qualifications may have more advantages in terms of know-how accumulation.

  • U.S. Senator Warner sends letter to tech giants on AI safety

On August 17, U.S. Senator Warner sent letters to the CEOs of companies such as Google, Meta, and Microsoft on the issue of artificial intelligence security. At the same time, he also sent letters to companies such as Apple, Amazon, and OpenAI on the issue of artificial intelligence security measures.

  • Chinese scientists successfully built a general-purpose brain-like artificial intelligence engine

On August 17th, it was learned from the Institute of Automation, Chinese Academy of Sciences that the researchers of the Institute successfully built a full-pulse neural network brain-like cognitive intelligence engine "Zhimai", and fully open sourced it. The platform will provide basic support for exploring brain-inspired intelligence research for general artificial intelligence, and help explore the computational nature of natural intelligence and the development of a new generation of artificial intelligence. The relevant research results were published in the journal "Patterns" of Cell Press in the form of a cover article.

  • Tianda successfully developed a high-performance millimeter-wave chip set

Recently, the entrepreneurial team of doctoral students from the School of Microelectronics of Tianjin University "Sinling Technology" successfully developed a high-performance 5G multi-band multi-standard compatible millimeter-wave chip set. The chip set is the first in the world to realize multi-band and multi-standard integration, and realize multi-band and multi-standard coverage of mainstream communications in 5.5G/6G international communication standards.

  • In the first half of the year, Beijing built 16,500 new 5G base stations

On August 17, at the "Two Districts" press conference, Wang Lei, a member of the party group and deputy director of the Beijing Municipal Bureau of Economics and Information Technology, said that in the first half of the year, Beijing built 16,500 new 5G base stations and 45.6 new gigabit fixed network connections. 10,000 households, achieving full coverage within the Fifth Ring Road and precise coverage in key areas and typical scenarios outside the Fifth Ring Road. At the same time, Beijing has built a national industrial Internet big data center and a top-level node command and operation center, with more than 55,000 connected companies. Haidian and Chaoyang artificial intelligence computing power centers plan to build 4000P computing power. Beijing Ascend Artificial Intelligence Computing Center, as the first approved national new-generation artificial intelligence public computing power open innovation platform in North China, has been officially put into operation.

  • LLM-Adapters: Integrating various adapters into large language models

On August 17th, researchers from Singapore University of Technology and Design and Singapore Management University published a paper entitled "LLM-Adapters: An Adapter Family for Parameter-Efficient Fine-Tuning of Large Language Models". This study introduces a family of adapters called LLM-Adapters, designed to enable efficient parameter fine-tuning of large language models. Through this adapter series, the number of parameters required for fine-tuning can be reduced without affecting the performance of the model, thereby improving the efficiency of fine-tuning.

4f8c3c21a0336144783c0165aa16e714.png

  • Canadian government plans to seek public comment on 'voluntary code of conduct' for generative AI

On August 17th, according to foreign media reports, the Canadian government plans to solicit public opinion to develop a "voluntary code of conduct" for artificial intelligence companies. The guidelines were developed by Innovation, Science and Economic Development Canada (ISED) to ensure companies adopt security measures, testing protocols and disclosure practices. The initiative aims to build trust in the new generation of artificial intelligence systems and prepare for the upcoming transition of the regulatory framework. The guidelines will be developed in consultation with companies, academia and civil society to ensure their validity and comprehensiveness. The guidelines encourage avoiding behaviors in which AI systems may harm users, and emphasize the importance of content labeling and disclosure of inner workings. Big tech companies such as Google, Microsoft and Amazon have said they will participate in the consultation process. In addition, the Canadian government has also proposed the Artificial Intelligence and Data Act, which is expected to take effect in two years.

  • Meta's new language model Humpback follows instructions and surpasses LLaMa model

News on August 17: Researchers at Meta have developed a new technique called "instruction back translation", which aims to improve the instruction tracking ability of large language models. This technology fine-tunes models such as LLaMa to follow instructions through self-augmentation and self-management methods, without the need for expensive human annotation or more powerful models. In the self-augmentation phase, the model generates candidate instruction-response pairs from an unlabeled text corpus and predicts the instruction that elicited the response. During the self-management phase, the model scores and filters candidate pairs, retaining the highest-scoring subset. The Humpback65B model developed by researchers at Meta achieved state-of-the-art results on the instruction-trace benchmark. They plan to extend this method further for even better results.

847a93a2ac7aeb53ab50d9ea7cd55db7.jpeg





d36058029e672347a724d286ba95ac68.png

c09b651108583686495f07bf4dc15cd0.png

Guess you like

Origin blog.csdn.net/YMPzUELX3AIAp7Q/article/details/132353082