[Collection of Big Model Daily Reports on November 9th] ChatGPT suffered DDoS hacker attacks and experienced periodic interruptions; developers’ “second brains” attacked, GitHub Copilot was updated, and human development participation was further reduced; big models took shortcuts to “brush the rankings”? The problem of data pollution deserves attention

The new features launched by OpenAI are so powerful that the server was instantly overcrowded

Link: https://news.miracleplus.com/share_link/11563

The new features released by OpenAI developers are so popular that the servers are overcrowded. Starting at around 6 a.m. Pacific time on November 8, the ChatGPT server was down for more than 90 minutes, and users who accessed it would receive the message “ChatGPT is at capacity right now”. Subsequently, OpenAI issued two “server outage” warnings in succession – a partial outage and a full line outage, and said that it was investigating the cause of the outage, repairing and monitoring it.

Developers’ “second brain” is coming, GitHub Copilot is updated, and human development participation is further reduced

Link: https://news.miracleplus.com/share_link/11564

What makes Andrej Karpathy lament is that the contribution of humans in directly writing code in the software development process will become smaller and smaller, and the role of direct input and supervision will become more abstract. Eventually, the role of humans will simply be to do basic review and validation, rather than being the primary programmer and developer. Turns out it’s GitHub’s newly released Copilot Workspace, which reimagines developers’ internal processes. If AI development tools are developers’ second pair of hands, then Copilot Workspace will be developers’ “second brain.” In the process of coding, the biggest headache is encountering an unfamiliar software repository, programming language or framework. The difficulty caused by solving these problems may delay your completion of the task, or even make it impossible to complete it at all. It is not easy to master these quickly and regain momentum in a short period of time. But Copilot Workspace may be able to help you get twice the result with half the effort, and even help you complete larger and more complex tasks.

GitHub CEO: AI assistant has become a cash cow for the company

Link: https://news.miracleplus.com/share_link/11565

In the latest news, GitHub CEO Thomas Domke revealed in an interview with the media that Copilot has “become a money-making tool.” Domke even bluntly said that the cost to users of Copilot is less than the price GitHub charges for it.

Using language to align multi-modal information, Peking University, Tencent and others proposed LanguageBind, refreshing many lists

Link: https://news.miracleplus.com/share_link/11566

Researchers from Peking University, Tencent and other institutions have proposed a multi-modal alignment framework – LanguageBind. The framework has achieved excellent performance in downstream tasks of five different modalities, including video, audio, text, depth map and thermal image, and has topped multiple evaluation lists. This marks the direction of “grand unification” in the field of multi-modal learning. “The concept has taken an important step forward.

Is the big model taking shortcuts to “beat the rankings”? The problem of data pollution deserves attention

Link: https://news.miracleplus.com/share_link/11567

In the first year of generative AI, everyone’s work pace has become much faster. In particular, this year everyone is working hard to roll out large models: Recently, domestic and foreign technology giants and startups have taken turns to launch large models. As soon as the press conference started, all of them made major breakthroughs, and each one refreshed the important Benchmark list. Either ranked first or first echelon. After being excited about the rapid progress of technology, many people found that there seemed to be something wrong: Why is everyone at the top of the rankings? What is this mechanism? As a result, the issue of “list brushing” has also begun to attract attention. Recently, we have noticed that there are more and more discussions in WeChat Moments and Zhihu communities on the issue of “swiping the rankings” of large models. In particular, a post on Zhihu: How do you evaluate the phenomenon that the Tiangong Large Model Technical Report pointed out that many large models use data in the field to boost rankings? It aroused everyone’s discussion.

Dimensity 9300, the first generative AI mobile chip: capable of running large models with 33 billion parameters

Link: https://news.miracleplus.com/share_link/11568

2023 is the first year of generative AI, and the mobile devices in our hands are also accelerating into the era of large models. On the evening of November 6, MediaTek officially released its annual flagship SoC Dimensity 9300. This chip has a 4+4 full-core design and completely surpasses competing Android and Apple products in terms of performance and energy consumption. What’s even more noteworthy is that this year’s flagship chip is defined as a 5G generative AI mobile chip, providing superior features of high intelligence, high performance, high energy efficiency, and low power consumption that far exceed those of the past.

China’s second batch, 11 large model registrations were approved

Link: https://news.miracleplus.com/share_link/11569

The second batch of large-scale AI models that have been registered in China include 11 companies, and some of them have opened their services to the whole society. Including the first batch of more than 10 large models, more than 20 large models have been registered. The new batch of registration lists include: NetEase Youdao (“Ziyue” large model), Ant Group (Bailing large model), Wall-facing Intelligence (“Luca”), Mobvoi (“Sequence Monkey”), Kunlun Vanguard Wei (“Tiangong” large model), Meituan (model), Zhihu (“Zhihaitu AI” model), Dark Side of the Moon (moonshot), Kingsoft Office (WPSAI), TAL (MathGPT large model), etc. .

China Literature CEO Hou Xiaonan: AIGC promotes large-scale export of online articles overseas, reducing translation costs by an average of 90%

Link: https://news.miracleplus.com/share_link/11570

On November 9, the 2023 World Internet Conference Cross-Strait and Hong Kong and Macao Internet Development Forum was held in Wuzhen, Zhejiang. Focusing on the theme of “Shared Development and Win-Win Future”, new opportunities in the Internet industry brought about by technological changes such as AIGC have become the focus of heated discussions among guests. In the era of digital intelligence, how to further promote cultural exchanges and industrial development? “What we insist on are two sentences: stick to good stories and embrace new technologies.” Hou Xiaonan, CEO and President of Yuewen Group, said.