"Geek Headlines" - the news circle for technicians!
Good morning, readers and friends of CSDN, "Geek Headlines" is here, let's see what are the important news worthy of our technical people's attention today.
Organize | Meng Yidan
Listing | CSDN (ID: CSDNnews)
A quick look at news points in one minute!
-
Meituan and Baidu take the lead, and big factories start recruiting for the class of 2024
-
Alibaba Dharma Institute launched the AI framework HQTrack, which can achieve high-quality video tracking
-
Huawei to restart 5G mobile chip production as early as this year
-
Geely will release the world's first full-scenario AI model of the automotive industry
-
Ali Ye Jieping: The Mota community has over 2 million developers
-
HUAWEI CLOUD Pangu Meteorological Large-scale Model Online
-
Didi Pay: Domestic and foreign users can pay in digital renminbi for online car-hailing
-
Hackers display malicious AI tool FraudGPT, which can automatically generate a variety of network attack codes
-
Nvidia's special chip supply soared to 500,000 yuan
-
Twitter rebrands to X in Apple Store
-
Language inequality in large AI models: English is the cheapest, other languages are much more expensive
-
AWS will charge for IPv4 public addresses
-
Brave says it doesn't support Google's Web Environment Integrity
domestic news
Meituan and Baidu take the lead, and big factories start recruiting for the class of 2024
On July 31, Meituan launched the 2024 campus recruitment. This recruitment is aimed at fresh graduates from domestic and overseas colleges and universities who will graduate from November 2023 to October 2024. It is expected to recruit 6,000 people, covering technology, products, business analysis, operations, finance, supply chain, functions, marketing, There are 10 categories of positions and more than 100 types of positions including design, sales, customer service and support. Baidu also launched the 2024 school recruitment this month, and the positions cover four categories: technology, products, government-enterprise solutions and services, professional services, and management support. At present, Internet companies that have launched school recruitment include Xiaohongshu, Sina and Weibo, Netease Game Leihuo, Mihayou, DJI, etc., but the specific recruitment scale has not been announced. (Interface News)
Alibaba Dharma Institute launched the AI framework HQTrack, which can achieve high-quality video tracking
According to the GitHub page, Ali Dharma Institute recently announced that it has cooperated with Dalian University of Technology to launch an AI framework called HQTrack. The framework mainly consists of a Video Multi-Object Segmentator (VMOS) and a Mask Optimizer (MR), aiming to achieve high-quality tracking of any object in a video. According to the research team, HQTrack is based on the open source frameworks DeAOT and HQ-SAM, and some improvements have been made on top of it to realize the possibility of segmenting each target in the video. The research team said that the AI framework can be applied to fields such as face recognition, intelligent driving, monitoring and tracking, and is expected to land in future commercial applications. (Pin Wan)
Huawei to restart 5G mobile chip production as early as this year
Recently, it was revealed that Huawei is working with China's largest chipmaker SMIC to develop a 5G-enabled chip that could be mass-produced within a few months. Further details on Huawei's plans to reboot the chip are unclear, but it is almost certainly utilizing chiplet technology. (Nikkei Asia)
Geely will release the world's first full-scenario AI model of the automotive industry
At yesterday's internal meeting, Gan Jiayue, CEO of Geely Automobile Group, revealed that the company already has a full-stack self-developed AI large-scale model technology, which will be released in the second half of the year. It is reported that this model will be the world's first full-scenario large-scale model of the automotive industry, and it is also the AI large-scale model with the most abundant application scenarios. Full coverage of large data models at the level, tens of billions, and hundreds of billions. Among them, AI intelligent interactive technologies such as voice interaction outside the car, AI music MV, AI children's picture book, WOW wallpaper version 1.2, etc., have already been installed on the Geely Galaxy L6 for the first time, and the model is expected to be launched in September. (Financial Association)
Ali Ye Jieping: The Mota community has over 2 million developers
Ye Jieping, vice president of Alibaba, said that the Mota community has gathered more than 1,000 AI models and more than 2 million AI developers. The models have been downloaded more than 45 million times, and has launched academic cooperation with more than 20 universities such as Tsinghua University, Peking University, and Zhejiang University. There are 500,000 developers from universities. At present, the third-party models in the community include the ChatGLM series models jointly developed by Tsinghua University and Zhipu, the Unidiffuser series models developed by TSAIL, an artificial intelligence innovation team of Tsinghua University, the Jiang Ziya series models developed by IDEA Research Institute, and the NLP and knowledge graph team developed by Zhejiang University. Intellectual analysis models, etc. In terms of computing power, Alibaba Cloud can provide a single cluster with a maximum scale of 10,000 GPU cards, which can carry multiple trillion-parameter large models for simultaneous online training. (Science and Technology Innovation Board Daily)
HUAWEI CLOUD Pangu Meteorological Large-scale Model Online
Huawei officially announced tonight that the HUAWEI CLOUD Pangea Meteorological Model is officially launched on the official website of the European Center for Medium-Range Weather Forecasting. On the website, global weather forecasters, weather enthusiasts and the general public can view Pangu's global weather forecast for the next 10 days for free. The European Medium-term Meteorological Center (ECMWF) also published a comparative test report of the Pangea Meteorological Large Model and the European Numerical Model from April to July this year. advantage. The report pointed out that the AI method represented by Pangu will change the bottleneck of slow improvement in the accuracy of numerical weather prediction in recent years. (IT House)
Didi Pay: Domestic and foreign users can pay in digital renminbi for online car-hailing
In order to help guarantee the payment service of the Universiade, Didi Pay provides a digital RMB payment function for travel scenarios. After downloading the Didi Chuxing App, users at home and abroad can use digital renminbi to pay for online car-hailing. It is understood that in the Didi Chuxing scene, express cars, special cars, and luxury cars have opened digital RMB payment services. Domestic and foreign users can choose "digital renminbi" for payment in the order payment link of Didi Chuxing App, and domestic users can also open wallet fast payment for Didi to enjoy the convenient payment experience of digital renminbi. (Central Broadcasting Corporation)
international news
Hackers display malicious AI tool FraudGPT, which can automatically generate a variety of network attack codes
According to the security company Netenrich, another hacker has created a malicious AI tool called FraudGPT, which is said to be "capable of generating a variety of network attack codes" and "more than 3,000 buyers have placed orders in less than a week." Rakesh Krishnan, a researcher at Netenrich, claimed that FraudGPT has been circulating on the dark web since July 22, and the subscription fee is $200 per month (approximately RMB 1429.46), $1000 (approximately RMB 7147.3) for six months, and $1700 per year. US dollars (about 12150.41 yuan).
Nvidia's special chip supply soared to 500,000 yuan
According to foreign media reports, due to the sudden surge in demand and export controls, the price of NVIDIA's AI GPU has been soaring all the way. AI chips that used to cost tens of thousands or hundreds of thousands of yuan have now risen to more than 200,000, 300,000 or even 500,000 yuan, almost becoming a wealth management product. According to supply chain news, the A800 and H800 chips specially supplied by Nvidia have increased from the original 120,000 yuan to 250,000 or even 300,000 yuan, and even as high as 500,000 yuan. Due to the increasing demand for AI graphics cards from major companies, Nvidia CEO Huang Renxun confidently stated in an interview before that the computer era has changed, and data centers will need to use CPUs less and less. 10,000 CPUs, instead buying millions of GPUs. (China Semiconductor Forum)
Twitter rebrands to X in Apple Store
The Twitter application has been renamed X in the Apple App Store on July 31, which means that the company may have obtained special approval from Apple, because the App Store has clearly stipulated that the application name requires at least two letters. Product designer Nick Sheriff previously stated on Twitter that Apple does not allow any application to use a single character as a name on iOS. But if Twitter gets special approval, it would mark the first time a company has received such permission since the App Store was founded. (IT House)
Language inequality in large AI models: English is the cheapest, other languages are much more expensive
Recently, a study showed that due to the way server costs are measured and billed by services such as OpenAI, English input and output costs are much lower than other languages, with Simplified Chinese being about twice as expensive as English, Spanish The Shan language of Myanmar is 1.5 times that of English, while the Shan language of Myanmar is 15 times that of English.
Twitter user Dylan Patel shared a photo showing a study conducted by the University of Oxford that found that it took 198 tokens for an LLM to process a sentence in Burmese, while the same sentence in English only required 198 tokens. 17 tokens are required. Lexograms represent the computational cost of accessing LLMs through APIs such as OpenAI's ChatGPT or Anthropic's Claude 2, meaning that a Burmese sentence is 11 times more expensive to use this service than an English sentence. (IT House)
AWS will charge for IPv4 public addresses
Cloud computing giant Amazon AWS announced that it will start charging for IPv4 public network addresses next year to encourage users to use IPv6 addresses. The reason is that IPv4 addresses are increasingly scarce and the cost of obtaining them is increasing. The IPv4 address space has been exhausted since 2019. Since then, the acquisition of IPv4 addresses mainly depends on recycling and the release of unused address segments. The cost of acquiring a single IPv4 address has more than tripled in the past five years. For the cloud computing giant, now is the time to pass the cost on to customers. Beginning February 1, 2024, AWS will charge $0.005 per address per hour for all IPv4 public addresses. The new policy applies to services including AWS EC2, Relational Database Service (RDS) and Elastic Kubernetes Service (EKS). (Solidot)
Programmer Zone
Brave says it doesn't support Google's Web Environment Integrity
After Vivaldi, Brave, another Chromium-based browser, also stated that it would not support Google's controversial Web Environment Integrity (WEI) API. WEI API is considered to be equivalent to the DRM of the Web, allowing websites to refuse to provide services to specific browsers because of browser installation of advertising plug-ins or other reasons. Brave co-founder and CEO Brendan Eich said that Brave will not support WEI. Google has added many privacy-invading mechanisms to Chrome/Chromium code in the past, and Brave usually chooses not to support such mechanisms. At present, the other two major browser developers, Apple and Microsoft, have not made any statement on this. (Solidot)