South Korea's SK hynix, the world's second-largest maker of memory chips, said Thursday that it will set up a new division called AI Infra to be responsible for artificial intelligence (AI) semiconductor-related businesses, as part of the company's strategy to focus more on high-demand high-end chips. It is reported that the division will also lead the development of artificial intelligence technologies such as a new generation of HBM chips, and find and develop new markets. SK hynix said Kim Juseon, the current head of the global sales and marketing department, will lead the new division.
On December 7, TCL Huaxing released the world's first semiconductor display vertical large model Xingzhi X-Intelligence. According to reports, the large model is developed by TCL Huaxing in conjunction with Zhipu AI and the National New Display Technology Innovation Center, and has been privatized and deployed.
McDonald's has announced a partnership with Google in 2024 to introduce generative artificial intelligence (AI) to optimize the operations of its stores around the world. As part of this partnership, McDonald's will carry out a large-scale hardware and software upgrade, covering multiple systems including ordering kiosks and mobile apps, with the aim of providing faster, fresher food.
McDonald's says that with generative AI, customers can expect to experience more AI-based automated services in more drive-in restaurants in the future. However, McDonald's did not provide details on how the AI will be used, saying only that it will make hardware and software upgrades, both local in stores and services provided through Google Cloud. This updated system will help restaurant management "quickly identify and adopt solutions to reduce business disruption."
Tianya is on holiday today, and on December 6, the mascot image "Long Chenchen" of the 2024 Spring Festival Gala of the Year of the Dragon of the Radio and Television Station was officially released. In response to this image, some netizens suspected that it was synthesized by AI. In this regard, the official Weibo of the Spring Festival Gala issued a statement in the early hours of this morning, saying, "'Long Chenchen' must have imperfections, but it is really drawn by the design teacher." Ornamentation, color, dragon face, etc., the design scheme of each detail has also been changed one version after another, and the design teacher's head has been bald one after another, so that there is the 'Long Chenchen' that everyone sees today." In addition, the Spring Festival Gala also attached two design drafts of "Long Chenchen"**, which proves that it is not synthesized by AI.
Meta has launched a new, standalone AI image generator, Imagine with Meta, which allows users to create images from natural language descriptions.
According to reports, the new artificial image generator is powered by Meta's existing EMU image generation model to create high-resolution images based on text prompts. It's currently free to use for English-speaking users in the United States (whether there will be a fee for the future is unknown), and each prompt generates four images.
Previously, Meta image generation models have faced controversy over racially biased image stickers. To address these issues, Meta says it will begin adding watermarks to images generated by Imagine With Meta, which will be generated by AI models and can be detected by the appropriate models to increase content transparency.
On December 6, Google officially announced the release of Gemini 10, which is what it considers to be the largest and most powerful AI model that can process information in different content forms such as **, audio, and text. According to reports, Google has released three different versions for different scenarios: Gemini Ultra, the largest and most powerful model for highly complex tasksGemini Pro, the best model for all kinds of tasks;Gemini Nano, the most efficient model on the device.
In addition, Google evaluated the performance of both models on a variety of tasks, from natural images, audio, comprehension, to mathematical reasoning, and Gemini Ultra has outperformed GPT-4 on 30 of the 32 commonly used academic benchmarks. Whereas, on the MMLU (Massive Multitasking Language Understanding) test, the Gemini Ultra scored 90A high score of 0%, becoming the first model to outperform human experts on MMLU. The MMLU is said to use a combination of 57 subjects, including mathematics, physics, history, law, medicine, and ethics, to test world knowledge and problem-solving skills.
Tianyancha APP shows that recently, AutoNavi Software has applied for the registration of a number of trademarks such as "AutoNavi Yunrui", "AutoNavi Yunzhi" and "AutoNavi Seconds", and the international classification involves scientific instruments, transportation and storage, ** services, etc., and the current trademark status is pending substantive examination. According to reports, "Yunrui" is a large industrial model independently developed by AutoNavi Yuntu, a subsidiary of AutoNavi Map"AutoNavi second delivery" is the errand express delivery business that AutoNavi has recently begun to test.
Recently, Alibaba Cloud and the Institute of Intelligent Industry (AIR) of Tsinghua University jointly launched the research on "Large Model-based Evolvable Agents" to explore the common basic technologies of model agents and improve the multilingual and multimodal basic capabilities of large models. It is reported that this is the first special research on the systematic integration of industry, university and research of evolvable agents in China, and it is also an important milestone in the continuous evolution of artificial intelligence model capabilities.
On December 7, SenseTime announced that Raccoon, an intelligent programming assistant based on SenseTime's self-developed large language model, will be open for public beta from now on, supporting 30+ mainstream programming languages such as Python, J**A, J**Ascript, C++, Go, SQL, and mainstream IDEs (integrated development environments) such as VS Code and IntelliJ Idea.
According to reports, **Little Raccoon covers software requirements analysis, architecture design, ** writing, software testing and other links to meet the needs of users ** writing, data analysis, programming Xi and so on. In practical applications, it can help developers improve programming efficiency by more than 50%.