01
Important news tips
1.Intel China responds to the rumors of "Qualcomm negotiating the purchase of Intel": no comment2.Google DeepMind launches: SCoRe, which can significantly improve the self-correction ability of the model3.Alibaba Cloud's "Tongyi Lingcode" is fully evolved: a day's workload of a real programmer can be completed in 5 minutes4.Nvidia responds to rumors that some manufacturers cannot place orders for H20 chips: do not comment on the rumors5.The mainland released version 3.0 of the basic model of 10-billion-level remote sensing interpretation, "Sky and Eyes".
02
Today's headlines
Intel China responded to the rumors of "Qualcomm negotiating to buy Intel": no comment
Some foreign media reported that chip giant Qualcomm was approaching Intel about the acquisition, and this news quickly attracted widespread attention in the industry. In response to this rumor, Intel China said in an interview with reporters: "I will not comment on the rumors. ”
At present, the specific details of the transaction are not yet known, and there is still a great deal of uncertainty about whether the transaction will be successfully completed. Still, if the deal does go through, it will be one of the largest and most far-reaching mergers and acquisitions in the semiconductor industry in recent years. Intel's stock price rose significantly after the news, while Qualcomm's stock price fell. Regarding the rumors that Qualcomm negotiated to buy Intel, industry analysts believe that if the transaction can be successfully completed, it will have a profound impact on the future development of the two companies. Qualcomm is expected to strengthen its technological strength and market share in various fields such as processors, graphics processors and the Internet of Things through the acquisition of Intel, further consolidating its leading position in the semiconductor industry. Intel, on the other hand, may gain new development opportunities through this merger and acquisition to get out of the current predicament. (Global Tech)
03
Domestic information
Alibaba Cloud's "Tongyi Lingcode" has been fully evolved: a human programmer's day's workload can be completed in 5 minutes
Alibaba Cloud announced a comprehensive upgrade of its AI programming assistant "Tongyi Lingcode" at the Apsara Conference, launching an "AI programmer" that can independently complete the whole process of software development, which supports integration into Alibaba Cloud's DevOps platform "Yunxiao" and code hosting platform GitHub. This AI programmer can complete a human programmer's day's workload in a matter of minutes, significantly improving development efficiency.
Users only need to describe the requirements through natural language, and the AI programmer can automatically analyze the requirements, write code, test, and fix defects. In a hands-on demonstration, a junior high school student managed to create a countdown webpage in just two minutes, demonstrating the ease of use and power of AI programmers. In the past year, more than 5 million IDE plug-ins have been downloaded, and we have joined companies such as Hello Group, generating more than 1 billion lines of code and assisting developers to generate more than 30 million codes per day (Global Intelligence News)
阿里云和JetBrains合作推出基于通义大模型的AI Assistant
During the 2024 Apsara Conference from September 19th to 21st, Alibaba Cloud partnered with JetBrains, a global software development tool provider headquartered in the Czech Republic, to release the JetBrains AI Assistant based on Alibaba Cloud's Tongyi model.
Yuexuan Li, President of JetBrains China, said that JetBrains actively embraces market changes to ensure that its products and services always resonate with the needs of developers, and will continue to deepen its strategic cooperation with Alibaba Cloud in the future to provide powerful and intelligent tools and solutions for users in the Chinese market. Liu Weiguang, Senior Vice President of Alibaba Cloud Intelligence Group and President of the Public Cloud Business Unit, said: "Today's collision of cloud and AI has brought a new wave. We strongly believe that the application of AI will sweep and permeate all walks of life. The strategic cooperation between Alibaba Cloud and JetBrains will give full play to the unique value of bridging international technology with the Chinese market, so that more cutting-edge innovative functions can be implemented in China, and we are willing to open a new chapter in the implementation of AI in the field of development together with JetBrains." (Titanium Media AGI)
Motiff announced AI-generated UI design technology, and the price of the professional version starts at 6 yuan per month
At the Apsara Conference 2024, Motiff, a "design tool for the AI era" under Kanyun Holding Group, announced the latest progress and development process of the new AI-generated UI design technology function.
Fan Kai, head of R&D at Motiff, said that based on the self-developed "Motiff model", the first multi-modal model of the UI industry in China, users can use text descriptions to generate your "first version" design draft, and then collaborate in Motiff in one-stop, so that AI can generate the overall structure and specific content of the page according to user needs, and directly deliver a complete page design through a set of expressive UI components. At the same time, the generative UI can also make it easier for users to use Motiff's professional editing and management features, which can be effectively integrated into the overall process. "We want Motiff to be more than just a Copilot in the design field, but also an Autopilot, where we can directly complete daily design tasks and even generate complete design drafts." Fan Kai said that large models actually have very good capabilities in generating the overall structure and content, but the main lack is the control of UI details, and if you use a well-built UI component library, you can actually cover most of the UI requirements. (Titanium Media AGI)
Huazhi Large Model 5.0 was released: built by Tongfang CNKI and HUAWEI CLOUD, 7B~135B size
On September 21, the Huazhi Large Model Industry Application Seminar was held in Shanghai with the theme of "Digital Intelligence Symbiosis and Industry Win-Win", which was jointly hosted by Tongfang CNKI and HUAWEI CLOUD. At the meeting, Huazhi Model 5.0 was released, and new applications such as intelligent PPT, AI technology novelty search, Huazhi App, and 3D holographic interactive digital human were launched.
The Huazhi model 5.0 released this time achieves:
- The ability of "full series, multi-modality, strong knowledge, and high trustworthiness" has been improved, and 7B, 13B, 38B, and 135B multi-size models from the end side to the hundreds of billions of levels have been trained for the construction needs of diversified large models.
- The multi-modal large model with cross-modal and multi-layer semantic fusion can realize the understanding of academic images and statistical tables, as well as the intelligent recognition and segmentation of entities in complex scenarios.
- Through the integration of multi-source vector library with strong logic control and the original multi-level thinking chain enhancement technology, the controllable generation of strong logical content of thinking chain + knowledge framework is realized, which effectively inhibits the illusion of large models, and further improves the professionalism and credibility of the content.
Tongfang CNKI said that at present, the Huazhi model has been successfully applied in the fields of education and scientific research, industry and agriculture, government affairs and finance, medical law and other fields. Based on Huazhi's large model, the "CNKI AI Learning and Research Assistant, the first personalized and generative professional knowledge service platform in China", has thousands of institutional users. (IT House)
China released version 3.0 of the 10-billion-level remote sensing interpretation base model "Sky and Eye".
At the first Aerospace Information Technology Conference, the Aerospace Information Innovation Institute of the Chinese Academy of Sciences announced that the world's first 10-billion-parameter air-space integrated remote sensing interpretation basic model independently developed by the research team of Fu Kun and Pengcheng Laboratory - "Aerospace Eyes" version 3.0 was officially released.
Version 3.0 provides an effective solution for space-air-ground integration applications. According to reports, since March this year, the research team has cooperated deeply with Pengcheng Laboratory, based on the hub node of "Pengcheng Cloud Brain II" of China Computing Network, using 512 Ascend 910 graphics cards, and completed the pre-training on 400 million samples of localized multimodal aerospace remote sensing data in several months, and the scale of basic model parameters exceeded 10 billion for the first time. VisionEye 3.0 uses a heat-transfer-based architecture to dramatically accelerate learning and inference, with computational speeds up to more than 2.4 times faster than traditional models. Based on the data of multi-sensor satellites such as visible light, synthetic aperture radar, thermal infrared, and multispectral, the model has been extended to the automatic and high-precision interpretation and processing of observation data from different aviation platforms such as aerial remote sensing aircraft and unmanned aerial vehicles. (IT House)
The University of Hong Kong pushes AI to predict the risk of cardiovascular and cerebrovascular diseases in Chinese people
On 22 September, the Li Ka Shing Faculty of Medicine of the University of Hong Kong (HKU) led the development of the first Personalised Cardiovascular and Cerebrovascular Disease Risk Assessment Model (P-CARDIAC) for Chinese people, which will launch a two-year research project to use artificial intelligence to assist medical staff in tailoring the most appropriate health management plans for about 3,000 patients in six public hospitals, which is expected to revolutionize the prevention and management of Chinese cardiovascular and cerebrovascular diseases, and greatly reduce the medical burden in Hong Kong.
P-CARDIAC is an AI model trained by integrating different data, including the use of the Hong Kong Hospital Authority's massive patient medical record data, referring to more than 120 risk factors, and taking into account the unique characteristics and medical history of each patient, in order to accurately predict the risk of cardiovascular and cerebrovascular diseases in Chinese people, and provide a comprehensive personalized cardiovascular and cerebrovascular disease risk score. Clinical Professor of Cardiology at HKU Faculty of Medicine, Mr Yiu Kai-hang said that in Hong Kong, heart and cerebrovascular diseases have always been the top five causes of death, so there is an urgent need to develop more effective disease prevention and management strategies. (Gelonghui)
Microsoft Research Asia and the Shanghai Mental Health Center have brought a new training method to patients with cognitive impairment with the help of multimodal large models
According to the official disclosure of Microsoft Research Asia, the institute and the Shanghai Mental Health Center jointly carried out joint research, based on the multi-modal large model in Microsoft Azure OpenAI service, developed a personalized cognitive training framework using a variety of modal data (such as voice, text, pictures, music and other information), bringing new possibilities for cognitive training for patients with cognitive impairment.
Based on the advice of experts from the Shanghai Mental Health Center and the needs of patients, researchers from Microsoft Research Asia developed a personalized cognitive training framework called ReMe based on multiple dimensions such as ease of use, user-friendliness, functional specialization, and ease of entry. Based on Microsoft's Azure OpenAI service, the tool has real-time interactive response functions, supports input and output of text, images, speech and other modalities, and provides users with a new cognitive training experience in the form of a conversational robot. At the same time, Microsoft Research Asia is also trying to record important nodes and events in daily life through wearable devices in the form of opportunistic sensing based on sensor technology, so as to promote more convenient and personalized personal memory training. (IT House)
After 14 rounds of financing, Westwell Technology launched an A-share IPO
According to the official website of the China Securities Regulatory Commission on September 19, Shanghai Xijing Technology Co., Ltd. has registered with the Shanghai Securities Regulatory Bureau, and the counseling agency is China International Capital Corporation Limited, which is classified as C39 computer, communication and other electronic equipment manufacturing.
Founded in 2015, Westwell Technology is an artificial intelligence solution service provider for the industrial and logistics industry, and its main business includes Qomolo new energy autonomous driving solutions, Cosmo digital intelligent scenario solutions, Loopo intelligent platform and LinkAIge intelligent consulting. As one of the first companies in China to develop driverless heavy trucks based on forward-looking self-development, its scenarios have now covered seaports
![Tesla and other car companies responded to an overseas company's illegal surveying and mapping; League of Legends team layoffs[fig]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACwAAAAAAQABAAACAkQBADs=)