AI Tool Collection uses a layout collection of 3 issues per month, updating the latest AI industry news daily. This issue is the collection of AI industry news for July Issue 1 (July 1 to July 10).
July 10
- Yao Qian, Director of the Science and Technology Supervision Bureau of the China Securities Regulatory Commission, wrote in China Finance that, overall, mainstream large models at home and abroad currently do not yet have a generational gap at the algorithm level, but there are gaps in computing power and data. He suggested vigorously supporting leading domestic technology enterprises in developing independently controllable domestic large models in general-purpose fields, while also encouraging the development of vertical industry models. (News source:East Money）
- Westlake Xinchen officially launched its self-developed general-purpose large model — the "Westlake Large Model" — and once again received a strategic investment of over 100 million yuan from Tom Cat to accelerate the application and deployment of large models. (News source:Xinchen Technology Official Account）
- KingMed Diagnostics announced today on its official WeChat account that it and Huawei Cloud officially signed a Pangu Large Model strategic cooperation agreement on July 8, and the two parties will jointly build an AI large model for the medical testing industry. (News source:KingMed Diagnostics Official Account）
- According to the Arxiv page, Google recently teamed up with Cornell University to publish a paper introducing an experimental protocol called VideoGLUE for evaluating the video understanding capabilities of large models. (News source:PingWest）
- Unity CEO John Riccitiello predicted that generative artificial intelligence (AIGC) will greatly shorten game development time, by about 5 to 10 times. (News source:Decrypt）
- Shanghai Renji Hospital upgraded its "Digital Human Customer Service" 3.0, aiming to realize anthropomorphic intelligent customer service for internet hospitals, enabling it to have voice conversation capabilities and provide patients with more convenient and considerate "smart services." (News source:IT Home）
July 9
- ChatGPT opened the Code Interpreter feature to all Plus users, providing a better interactive programming experience and data visualization capabilities. Even non-technical users can use it to complete complex programming tasks. (Read the full article:Machine Heart）
- Microsoft Bing engineering and product lead Jordi Ribas posted on Twitter that they improved the server side of the image generation tool Bing Image Creator , reducing image generation time by 30%. (News source:IT Home）
- According to reports, Google has been testing at Mayo Clinic research hospitals and other places since AprilPaLM 2a variant of the large language model, Med-PaLM 2, an artificial intelligence model designed to answer questions about medical information. (News source:The Verge）
July 8
- At today's World Artificial Intelligence Conference, China Mobile announced the release of the Jiutian artificial intelligence industry large model, divided into the Jiutian Haisuan government affairs large model and the Jiutian customer service large model. (News source:Cailian Press）
- At the recently held first Confidential Computing Summit in San Francisco, Intel announced that it will cooperate with Nvidia to provide confidential artificial intelligence solutions, strengthening AI security and privacy, and meeting customers' higher security and compliance needs. (News source:Intel Blog）
- Well-known foreign online video editing tool VEED launched an AI digital human talking-head tool, offering 50 different virtual digital avatars and choices of 75 languages and dialects including Chinese, helping users easily create localized multilingual videos. (Website address:VEED AI Avatars）
- Recently, at the World Artificial Intelligence Conference, Tencent Cloud announced that it used AI for the first time to help the world's largest radio astronomy telescope, the "China Sky Eye FAST," find 2 fast radio bursts. (Read the full article:Tencent Cloud Official Account）
- Recently, well-known cloud hosting service provider DigitalOcean announced that it would acquire New York-based cloud computing and artificial intelligence development startup Paperspace for $111 million in cash to help customers more easily test, develop, and deploy artificial intelligence applications. (News source:DigitalOcean Blog）
- Startup SpeedyBrand announced that it raised $2.5 million in a funding round led by GV (Google's venture capital arm) and Y Combinator, with a post-money valuation of $15 million. The company aims to use AIGC to create high-quality and cost-effective SEO content for small and medium-sized enterprises. (News source:TechCrunch）
- YouTube is testing AI-generated quizzes on its mobile apps for iOS and Android devices, aiming to help viewers test their understanding of topics covered in recently watched educational knowledge videos. (News source:Android Authority）
July 7
- OpenAI announced today that, effective immediately, the GPT-4 API is open to all paying OpenAI API customers, and all paying OpenAI API customers can now access the GPT-4 API with an 8K tokens context. (Read the full article:ITHome）
- At Huawei Developer Conference 2023 (Cloud), Huawei Cloud officially released Pangu Model 3.0. Pangu Model 3.0 is divided into a three-layer architecture of L0 foundation models, L1 industry models, and L2 scenario models, and will reshape all industries. (News source:Cailian Press）
- At the Alibaba Cloud MaaS Forum at the 2023 World Artificial Intelligence Conference, Alibaba Cloud CTO Zhou Jingren announced that the AI painting creation model "Tongyi Wanxiang" began invitation testing today, and images can be generated through text descriptions. (News source:TMTPost）
- In addition to Tongyi Wanxiang, Alibaba Cloud also launched China's first large-model invocation tool,ModelScopeGPT, which can receive user instructions and, through the "hub model," invoke other AI models in the ModelScope community with one click. Large and small models work together to complete complex tasks, and it can converse with users, generate images, generate videos, provide voice dubbing, and more. (Read the full article:Qianjiang Evening News）
- Ant Group and Tsinghua University jointly released the AI safety detection platform "Yijian 2.0." Yijian 2.0 can use generative AI capabilities to detect generative AI models, and can identify hundreds of risks across three major categories: data security, content security, and technology ethics, covering multiple data and task types such as tables, text, and images. (News source:PingWest）
- The "SenseTime SenseNova" large-model system has been comprehensively upgraded. The 2.0 version of the hundred-billion-parameter natural language processing model "SenseChat" supports Arabic and Cantonese, and model versions with different parameter scales, such as small models, have been launched to meet application needs across different terminals and scenarios such as mobile devices. (News source:The Paper）
- Enterprises such as 360 Group, Baidu, Huawei, and Alibaba were awarded by the China Electronics Standardization Institute of the Ministry of Industry and Information Technology as lead units of the "National Artificial Intelligence Standardization Overall Group Large Model Special Group," actively promoting the construction of the national standards system for large models and helping the development of China's large model industry. (News source:Tencent）
- Japanese Minister of Economy, Trade and Industry Yasutoshi Nishimura said at a press conference that the Japanese government will provide SoftBank Group with a subsidy of 5.3 billion yen (about 270 million RMB) for the development of generative artificial intelligence. (News source:Jiemian News）
- The next-generation "world's fastest" supercomputer, El Capitan, has begun installing components that will be fully operational next year. A key component of the system will be AMD's upcoming "industry's most advanced" Instinct MI 300 AI GPU. (Read the full article:ITHome）
July 6
- Today, a new-generation digital Earth technology platform developed by Chinese researchers—the StarMap Earth Smart Brain Engine—was released. The system uses intensive "intelligent computing" to provide users with intelligent processing of Earth data, intelligent perception of Earth information, and intelligent reconstruction of Earth scenarios. (Read the full article:CCTV News）
- The 2023 World Artificial Intelligence Conference (WAIC 2023), themed "Intelligent Connectivity, Generating the Future," opened this morning at the Shanghai Expo Center. Numerous well-known industry figures attended, including Turing Award winner Yann LeCun, Musk, AMD CEO Su Fengzi, and Midjourney founder David Holz. (News source:The Paper）
- OpenAI is forming a new team led by its chief scientist and co-founder Ilya Sutskever, investing 20% of its total computing power to develop methods for guiding and controlling "superintelligent" AI systems. (Read the full article:QbitAI）
- Recently, the internationally top academic journal Nature published in its main issue the research results of the Huawei Cloud Pangu Model R&D team—"A Three-Dimensional Neural Network for Accurate Medium-Range Global Weather Forecasting"—making weather forecasts more accurate. (Read the full article:ITHome）
- Former Twitter vice president Richard Rabbat and former chief technology officer of the GIF image library "Giphy," Anthony Johnson, founded an AI company called Lighty AI, an AI-driven personal assistant startup. (News source:The Information）
- Recently, two American writers filed a lawsuit against ChatGPT parent company OpenAI, claiming that OpenAI used their works to train its artificial intelligence without obtaining copyright authorization. (Read the full article:Cailian Press）
- Recently, the domestic L4 autonomous driving company "WeRide" obtained the first national-level autonomous driving license in the Middle East approved by the United Arab Emirates. Subsequently, various types of WeRide unmanned vehicles, including Robotaxi, Robobus, and unmanned sanitation vehicles, can be tested and operated locally. (Read the full article:QbitAI）
July 5
- Midjourney announced today on its official Twitter that it has launched a new content expansion feature called Pan, allowing users to expand image content in a fixed direction after generating an image. (News source:PingWest）
- OpenAI announced that ChatGPT has suspended access to Bing's web search feature, because users could use this feature to bypass the paywalls of some websites and directly obtain the full text of articles. To respect the rights of content owners, OpenAI decided to suspend access to this feature. (News source:ITHome）
- Inflection AI recently announced that it is developing a supercomputer equipped with 22,000 NVIDIA H100 GPUs, which can meet the development needs of generative AI. (News source:WCCF Tech）
- Researchers from Tsinghua University Shenzhen International Graduate School, Tencent AI Lab, and Peng Cheng Laboratory jointly published a research paper on DreamDiffusion, which uses the powerful generative capabilities of pretrained text-to-image models to directly generate high-quality images from electroencephalogram signals. (Read the full article:Machine Heart）
July 4
- Tencent Cloud released China's first AI-native vector database, supporting an industry-leading vector retrieval scale of up to 1 billion and controlling latency at the millisecond level. Compared with traditional standalone plug-in databases, the retrieval scale is increased by 10 times, and it also has a peak capacity of millions of queries per second (QPS). (Read the full article:Tencent Cloud）
- Peking UniversityChatExcelThe team recently released the Chinese legal large model ChatLaw, providing inclusive legal services to the public. The model supports file and voice output, and also supports legal document writing, legal advice, and legal aid recommendations. (Read the full article:QbitAI）
- Japan and the European Union held a ministerial meeting in the digital field in Tokyo yesterday, and a joint statement was compiled stating that they will cooperate in the formulation of international rules for generative artificial intelligence (AI). In order to formulate the rules, the two sides will advance the "Hiroshima AI Process" proposed at the G7 summit in May. (News source:Jiemian News）
- According to reports, ByteDance wants to build robots. At present, its robotics team has about 50 people and plans to expand to hundreds by the end of the year, and to produce some robots that serve ByteDance's own e-commerce fulfillment needs and can sort and pack goods in warehouses. (News source:LatePost）
- Recently, BMW Brilliance and Baidu held a strategic cooperation memorandum signing ceremony. The two sides will share advantageous resources, explore integrated innovation of AI technology and all-scenario automotive manufacturing, and advance cooperation in digital and intelligent technologies. (News source:Baidu AI Cloud）
July 3
- Baidu officially launched the iOS version ofERNIE BotApp, which is now available on the Apple App Store. Users can download and install it for free. The App size is 46.1MB. (App Store address:ERNIE Bot on the App Store）
- The latest version of Mathematica Wolfram has officially introduced a large language model (LLM). By simply inputting a text description, scientific calculations can be performed. (News source:QbitAI）
- Data from Similarweb, a website traffic monitoring service provider, shows that since the beginning of this year, ChatGPT's month-over-month growth rate in visits was 131.6% in January, and has declined month by month since then. In May, this figure had fallen to 2.8%, and June is expected to show month-over-month negative growth. (News source:CCTV News）
- An "AI 2.0 company" personally led and incubated by Sinovation Ventures Chairman and CEO Kai-Fu Lee made its first public appearance, officially announcing the new brand "01.AI". The official website (01.AI) was launched simultaneously. (News source:Geek Net）
- The multilingual NLP team of Alibaba DAMO Academy released the first multilingual multimodal benchmark M3Exam, covering a total of 12,317 questions, to promote the evaluation of large models. (News source:Machine Heart）
- Kingsoft Office cooperated with the NVIDIA team to improve the inference efficiency of image document recognition and understanding through NVIDIA Tensor Core GPUs and TensorRT; with the deployment of NVIDIA Triton Inference Server, it successfully optimized GPU utilization and provided high inference throughput. (News source:IT Home）
- Tianyancha and Huawei Cloud jointly released the world's first business investigation large model — the "Tianyanmei" trusted business assistant. This product faces users in an anthropomorphic interactive form and provides them with accurate business information. It is the industry's first controllable large-model digital product for individuals. The product has completed internal testing and is expected to open invitation trials to users at the end of July. (News source:36Kr）
- Snap recently jointly announced a large model product called Magic123 with King Abdullah University of Science and Technology and researchers from the University of Oxford, which can convert 2D images into 3D models. (News source:PingWest）
July 2
- Elon Musk recently announced that in order to prevent AI companies from massively scraping Twitter data and affecting the experience of real users, he decided to limit Twitter reading volume. Unverified accounts can only read 600 tweets per day, while newly registered unverified accounts can only read 300. (News source:ITHome）
- The Artificial Intelligence Summit Forum of the 2023 Global Digital Economy Conference was held in Beijing today. With the theme of "Intelligent Emergence, Reshaping the Future," more than 10 top experts and scholars in the field of artificial intelligence and representatives of enterprises and institutions shared profound insights and thoughts on the currently hottest topic of large model development. (News source:Tencent）
- Valve, the operator of the Steam platform, said in response to reports that it refused to list games containing AI-generated content that it is learning and adapting to the application of AI technology in game development and trying to incorporate it into its existing review policies. (News source:ITHome）
- At the Artificial Intelligence Summit Forum of the 2023 Global Digital Economy Conference held today, Zhang Xin, deputy general manager of China Telecom Digital Intelligence Technology Branch, announced China Telecom's ChatGPT product, the TeleChat large model, which supports outputting code, writing speeches, and more. (News source:TMTPost）
July 1
- Humane, a popular AI company founded by former Apple employees, today revealed its first product, the Humane Ai Pin. The company says it is a "clothing-based wearable device" that will go on sale later this year, is equipped with Qualcomm's Snapdragon platform, and uses a series of sensors to achieve "contextual and environmental computing interaction." (News source:The Verge）
- Industrial software service provider Aipuhua recently released a new generation AI+MOM (Manufacturing Operations Management) industrial software "Tianxuan," and also launched the MOM intelligent copilot "Aiwen" based on the Fourth Paradigm Shishuo large model. (News source:36Kr）