[News] Good Morning deepin, News Overview - July 26, 2024
Tofloor
default avatar
Yoloy_Lan
Moderator
2024-07-26 09:27
Author

News Highlights

  • CrowdStrike Apologizes for Outage with $10 Gift Card, Marked as Fraudulent Activity
    This week, CrowdStrike issued a $10 apology gift card to global partners in response to the global Windows computer outage. The gift card can be used on the food delivery platform Uber Eats. CrowdStrike acknowledged that the "July 19 incident" created extra work for its partners. Some recipients of the gift card reported on Wednesday that when they tried to redeem it, they were informed that the card "had been canceled by the issuer and is no longer valid." CrowdStrike spokesperson Kevin Benach confirmed that gift cards were distributed to employees and partners, and stated that due to high usage, Uber flagged these vouchers as potentially fraudulent. (Financial News)

  • OpenAI Directly Challenges Google with New AI Search Feature
    On July 25, local time, OpenAI launched a new search feature called "SearchGPT," marking its most direct challenge to Google, which has long dominated the search industry. OpenAI announced that registration for the new tool is now open; it is currently in the prototype phase and being tested with a small group of users and publishers. The new feature will have a beta version released in web browsers. The company plans to integrate the search tool's functionality into ChatGPT in the future. In a blog post, OpenAI stated that SearchGPT will provide summarized search results with source links in response to user queries. Users can also ask follow-up questions and receive contextual responses. (Interface)

  • Loongson 3C6000 Server CPU Successfully Taped Out
    Hu Weiwu, chairman of Loongson Technology Co., Ltd., announced that the company's server CPU, Loongson 3C6000, has recently completed tape-out. Test results indicate that its general processing performance has significantly improved compared to the previous generation server CPU, Loongson 3C5000, reaching the level of Intel's mid-to-high-end Xeon Silver 4314 processor.
    The Loongson 3C6000 is a CPU product aimed at the server market, featuring 16 cores and 32 threads on a single die, supporting dual, quad, and octa-socket direct connections. This CPU uses Loongson's self-developed instruction set architecture, "LoongArch," without requiring foreign authorization. From the overall architecture planning to the functional definitions of each component, and the detailed encoding, naming, and meaning of each instruction, Loongson has redesigned everything with full autonomy. The Loongson 3C6000 processor will also introduce LoongChain 1.0 for inter-die interconnection, significantly reducing inter-die access latency for the first time.

  • Stack Overflow 2024 Developer Survey Report
    In the past year, 62.3% of respondents used JavaScript, followed by HTML/CSS (52.9%) and Python (51%). Among the programming languages that developers "have not tried yet but want to try the most next year," Python ranked first, surpassing JavaScript. Additionally, Python is the most popular programming language among those learning to code, with this group indicating that they hope to use an average of 12 different languages next year.


  • Mistral AI Releases 123 Billion Parameter Large 2 Model
    Following Meta's release of the strongest open-source large model, Llama 3.1, French AI startup Mistral AI has also entered the competition, launching its new flagship open-source model, Mistral Large 2, today.
    This model has 123 billion parameters and can run at high throughput on a single H100 node, competing with the latest cutting-edge models from OpenAI and Meta in code generation, mathematics, reasoning, etc., while providing stronger multilingual support and advanced function calling capabilities.
    Mistral stated that one of the training focuses was to minimize the model's hallucination issues. The company claims that the training for Large 2 makes its responses more discerning; when it does not know something, it will admit its lack of knowledge instead of fabricating seemingly reasonable information.
    Mistral Large 2 features a 128k context window and supports dozens of languages, including Chinese, as well as over 80 programming languages. The model achieved an accuracy of 84.0% on the MMLU benchmark and shows significant improvements in code generation, reasoning, and multilingual support. It supports languages such as French, German, Spanish, Italian, Portuguese, Arabic, Hindi, Russian, Chinese, Japanese, and Korean. In terms of coding, it supports over 80 programming languages, including Python, Java, C, C++, JavaScript, and Bash.


  • Node.js Adds Native Support for TypeScript
    Node.js has merged a PR that adds experimental support for TypeScript. According to the PR description, developers can execute TypeScript files by configuring the experimental flag --experimental-strip-types.
    Node.js will automatically transpile TypeScript code into JavaScript code. During the transpilation process, type checking will not be performed, and types will be discarded—this process is referred to as "type stripping" by the Node.js team.


  • OpenAI Offers Free Fine-Tuning for GPT-4o Mini Model
    Shortly after Meta launched the open-source Llama 3.1 model, OpenAI announced that it would provide free fine-tuning for its GPT-4o Mini model, allowing users to train the model with additional data at no cost to achieve better performance for specific use cases.
    The fine-tuning for GPT-4o Mini is available to developers in OpenAI's Tier 4 and 5 usage levels, which are the highest-priced tiers in OpenAI's program. OpenAI plans to gradually extend free fine-tuning to all tiers. Free fine-tuning will be available from now until September 23.
Reply Favorite View the author
All Replies

No replies yet

New Thread

Popular Events

More
国际排名
WHLUG