The research team has recently introduced an exciting new method called T-FREE, which significantly boosts the operational efficiency of large language models. Scientists from Aleph Alpha, Darmstadt University of Applied Sciences, hessian.AI, and the German Research Center for Artificial Intelligence (DFKI) have jointly released this amazing technology, which is officially named 'Sparse Representation for Memory-Efficient Embedding Without Tokenizers'. Traditionally, we use tokenizers to convert text into a numerical format that computers can understand, but

After several quarters of growth in shipments, the global smartphone market has finally emerged from its slump. This September, major brands are holding grand launch events; Apple has unveiled the highly anticipated iPhone 16, while Huawei has introduced the Mate XT. Meanwhile, Honor has made an appearance at the IFA in Berlin, announcing its entry into the 'autonomous driving era' with the industry's first AI Agent intelligent entity.

At the 2024 WAIC Scientific Frontier Main Forum, the OpenDataLab team from the Shanghai Artificial Intelligence Laboratory (Shanghai AI Lab) launched a new intelligent data extraction tool called MinerU. This tool aims to simplify the AI data processing workflow and assist AI researchers in extracting high-quality data from vast amounts of documents.

Recently, MiniMax secretly launched its first video generation large model and showcased a 2-minute video titled 'Magic Coin' generated by this model. Founder Yan Junjie revealed that the model outperforms Runway in video generation quality, but specific parameters and technical details have not yet been disclosed. MiniMax plans to continue iterating in future versions to enhance data, algorithms, and usage details, and aims to add the ability to generate videos from images and a combination of text and images in future updates. Regarding commercialization, the company will consider it once the new version reaches a satisfactory state.

Magic, a startup focused on generative AI programming, has announced the completion of $320 million in funding, bringing its total financing to nearly $465 million. This round of funding was led by former Google CEO Eric Schmidt, with participation from several well-known investors. Magic's core technology, the Long-Term Memory Network (LTM), can handle vast amounts of data and avoid bias when generating code. Its latest model, LTM-2-mini, can process 100 million tokens, equivalent to 10 million lines of code or 750 novels. In partnership with Google Cloud, it plans to establish two '

AI startup Magic has completed a $320 million funding round, with participation from Eric Schmidt, Alphabet's CapitalG, and investors like Atlassian, bringing the total funding close to $500 million. Magic focuses on AI model development aimed at generating code and automating software development tasks, featuring an exceptionally long context window of up to 100 million tokens, surpassing current commercial models. Magic plans to collaborate with Google Cloud to build two supercomputers to enhance

Apple has recently released the iOS 18.1, iPadOS 18.1, and macOS 15.1 Sequoia developer beta, introducing new features in the Apple Intelligence series. Notably, the 'AI Removal' feature for photos allows users to easily remove unwanted objects from images. Utilizing AI technology, it automatically generates backgrounds to ensure the overall effect remains unaffected. Users can remove objects through smart detection or manual selection and brushing, with the system capable of recognizing object shadows.

As the realism of AI-generated images increases, people begin to question the authenticity of videos. A video released by 'Quantum Bit' on Bilibili discusses how to identify AI-produced videos, triggering widespread interest. The video introduces some intuitive recognition methods, such as observing whether characters' movements and expressions are natural, as well as the coordination of voice, mouth shapes, and emotions. In the face of a vast number of videos, AI recognition technology becomes crucial. The traces left by AI face swapping technology during the splicing process serve as important clues for AI recognition. The research team developed classifiers from dimensions such as models, motion, and geometry.

The desktop device secretly developed by Apple combines an iPad display with a robotic arm, aiming to achieve smart home control, video conferencing, and home security monitoring through a 360-degree rotation and tilt mechanism. The device can automatically adjust the screen angle, ensuring users are always centered during video calls with its 'look at me' feature. It is equipped with a built-in Siri voice assistant that supports voice command operations. Expected to be released in 2026 or 2027, it will be priced around $1000 and run on a customized iPadOS.

Canva has officially launched its AI creation suite 'Magic Studio' in China, marking a significant step towards its mission of 'empowering the world through design'. Aimed at enhancing design efficiency and unleashing creative potential, 'Magic Studio' makes it easy for newcomers to create professional-quality works with its user-friendly features. Its core advantages include the provision of intelligent tools, such as 'Magic Write', 'Magic Eraser', 'Magic Capture', and 'Magic Animation', covering the entire process from copy generation, image processing to presentation design, simplifying the design experience.

Online design platform Canva has announced its plan to acquire Australia-based generative AI content and research startup Leonardo.ai, as part of its goal to build a 'world-class visual AI toolkit.' While the financial terms of the deal have yet to be disclosed, Canva will gain access to a range of user-customizable text-to-image and text-to-video generators from Leonardo.ai.

Honor has launched the first AI PC, the Honor MagicBook Pro 16, which is the world's first laptop with spatial audio on the Windows platform. The integration of large AI models into terminal devices will promote the commercialization of these models. The AI PC is regarded as the best carrier for large models and represents the biggest technological transformation in the PC industry over the past decade. AI PCs will inject new momentum into the PC industry ecosystem and drive revenue growth for mid-stream OEMs and brand manufacturers.

["Microsoft has introduced a new feature called 'Generative Eraser', which uses artificial intelligence to remove objects or people from photos.","This feature is similar to the 'Magic Eraser' functionality found on Google Pixel and Samsung Galaxy phones.","The new feature will be rolled out to Windows 11 and Windows 10 users, including Arm64 devices.","Currently, the feature is available to Windows Insider users and will gradually be released to all channels.","Microsoft has not clarified whether edited photos will carry metadata or watermarks."]

At the press conference on January 10, Honor unveiled its new platform-level AI large model - Magic Model, which will be applied for the first time in the upcoming Honor Magic6 series. It also launched MagicOS 8.0, the latest integrated system that realizes a GenAI platform-level system based on user intent recognition capabilities. The intent recognition interaction speed of Honor Magic OS 8.0 is extremely fast, completing operations in just 1 second, far surpassing iOS 17's 10 seconds.

Magic搭 Community collaborates with vLLM and FastChat to provide efficient LLM inference and deployment services. Developers can use vLLM as the inference engine in FastChat, offering high-throughput model inference. FastChat is an open platform for training, serving, and evaluating LLM-based ChatBots. vLLM is an LLM service developed by researchers from the University of California, Berkeley, Stanford University, and the University of California, San Diego.

The Beijing Internet Court has held the first public trial of the country's first AI voice infringement case, where a voice actor sued Magic Sound Workshop APP, Microsoft, and four other defendants. The plaintiff discovered that their voice work had been transformed by AI and sold on the Magic Sound Workshop APP. The plaintiff accuses the defendants of infringing on their voice rights and demands an end to the infringement as well as compensation for losses. All five defendants deny the infringement, each claiming legal sources. The case is still under further trial.

The DingTalk AI Magic Wand has officially launched, featuring 17 product lines, over 60 scenarios, and nearly a hundred AI skills, all open for testing. Users can access the DingTalk AI through the 'Magic Wand' entry at the top right corner of the DingTalk homepage or the Magic Wand button on various product pages. The DingTalk AI Magic Wand offers various AI skills, including summarization, content creation, data analysis, information extraction, customer service consulting, and tool enhancement features. The summarization feature allows users to quickly understand lengthy documents, meetings, and more.

Canva has recently launched the 'Classroom Magic' educational AI tool, available for free to K-12 students and teachers. The tool includes several AI features such as 'Magic Write' and 'Magic Animate' to help students and teachers enhance their multimedia design skills. Canva has also introduced the 'Canva Shield' initiative to ensure the safe use of AI tools in the education sector.

MagicSchool AI is an educational tool developed based on OpenAI's ChatGPT technology, providing teachers with over 50 different AI functionalities for text customization, worksheet generation, scientific experiments, concept explanations, and video question generation. This tool has been developed by professionals in the education field to help improve teaching quality, particularly addressing the issue of teacher shortages. Despite the rich functionalities offered by AI tools, teachers must remain cautious of potential misinformation and biases, with decision-making authority still resting with the educators.

The Magic Community has launched a similar image generation model that can produce multiple images with similar styles but not identical. This model is suitable for fields such as artistic creation and product design, providing more options while saving time and costs. Links to the model and the creation space experience are provided for users to utilize.