Welcome to the 【AI Daily】 section! This is your daily guide to exploring the world of artificial intelligence. Every day, we present you with hot topics in the AI field, focusing on developers, helping you gain insights into technological trends and understand innovative AI product applications.
Fresh AI Products Click to learn more: https://top.aibase.com/
1. Luchen Upgraded Free Video Generation Model Video Ocean V2.0, create 20-second high-definition videos with one click
In today's thriving AI technology landscape, Luchen Technology's launch of Video Ocean V2.0 marks another significant leap in video creation. This tool is completely free and has achieved notable improvements in image quality, dynamic performance, and creative freedom. Users can generate high-quality videos through simple text prompts or reference images, with a user-friendly interface suitable for both professionals and casual users.
【AiBase Summary:】
🌟 Enhanced image quality: Achieves ultra-realism in facial details, light and shadow changes, and environmental interactions.
🏃 Action representation: Supports smooth and natural presentation of everything from everyday actions to extreme sports.
🎨 Diverse creative styles: Covers various styles including 3D realism, 2D animation, cinematic quality, and cyberpunk.
Experience link: video.luchentech.com/zh-CN
2. Surprise! DeepSeek APP officially launched, available on both iOS and Android
DeepSeek has quietly released its new APP, supporting both iOS and Android platforms. Features include deep thinking and online search. The app utilizes the DeepSeek-R1-Lite model for chain reasoning analysis, helping users solve problems step by step. Meanwhile, the DeepSeek V3 model supports quick access to internet information, generating concise answers. The app also offers smart conversation, accurate translation, creative writing, efficient programming, and intelligent problem-solving, significantly enhancing user productivity.
【AiBase Summary:】
🧠 Deep thinking feature analyzes problems through chain reasoning to draw conclusions.
🌐 Online search feature supports quick information retrieval and generates accurate answers.
📄 The app supports file uploads, providing smart conversation and various practical features.
3. OpenAI's Robotics Project Hiring, Plans to Develop Custom Sensors and Humanoid Robots
OpenAI recently announced the launch of its robotics project, aimed at exploring advancements in robotics technology. This project will develop robots equipped with custom sensors and is hiring for multiple key positions, including electronic perception engineers, robotic mechanical design engineers, and technical project managers. Annual salaries for each position range from $340,000 to $440,000. OpenAI aims to create versatile, adaptive, and multifunctional robots, potentially including humanoid forms.
【AiBase Summary:】
🔍 OpenAI launches robotics project, hiring for positions like electronic perception engineers.
🤖 Plans to develop versatile, adaptive robots, potentially including humanoid forms.
💡 OpenAI continues to advance robotics research by investing in humanoid robotics technology companies.
4. Researchers Open Source Inference AI Model Sky-T1, Training Cost Less Than $450
The Sky Computing Lab team at the University of California, Berkeley has launched Sky-T1-32B-Preview, an open-source inference AI model that significantly reduces the cost and difficulty of developing inference AI. The model performs excellently on multiple benchmark tests, with training costs only $450, far below the previous costs of millions of dollars.
【AiBase Summary:】
💰 The training cost of Sky-T1 is only $450, significantly lowering the barriers to developing inference AI.
📊 In benchmark tests like MATH500, Sky-T1 outperformed earlier versions of OpenAI's models.
🔍 The NovaSky team plans to continue developing more efficient open-source models to enhance inference performance and accuracy.
5. Google Releases Open Source Time Series Forecasting Model TimesFM 2.0, Usable for Retail Sales and Stock Trends Prediction
Google's research team has released TimesFM 2.0, a pre-trained model designed for time series forecasting, aimed at improving prediction accuracy. This model supports univariate predictions with up to 2048 time points and can handle any time span. Through open sourcing and scientific sharing, TimesFM 2.0 provides users with a rich dataset covering multiple domains, advancing the development of artificial intelligence.
【AiBase Summary:】
🌟 TimesFM 2.0 is Google's new time series forecasting model, focusing on enhancing prediction accuracy.
🔧 The model supports predictions for up to 2048 time points and can handle any forecasting time span.
📊 Users can freely choose prediction frequency based on different time series characteristics, increasing prediction flexibility.
Details link: https://huggingface.co/google/timesfm-2.0-500m-pytorch
6. Shenzhen's New Humanoid Robot SE01 Gains Popularity Overseas, Netizens Exclaim: Is This Really Not Generated by Sora?
The SE01 humanoid robot launched by Shenzhen has quickly gained popularity due to its human-like gait and flexibility, attracting a large audience and sparking discussions overseas. Despite occasional falls, its technological background and market prospects have left consumers excited about the future of robotics technology. Zhongqing Robot Company is dedicated to combining high technology with market demand, showcasing strong market competitiveness.
【AiBase Summary:】
🚶♂️ The SE01 robot boasts excellent gait, speed, and stability, walking like a model on a runway.
💰 Production costs are controlled at 100,000 yuan, with an expected selling price of 150,000 yuan, aiming for small-scale commercial use in the future.
📸 Equipped with Intel RealSense depth cameras and 360-degree LiDAR, it has good environmental perception and obstacle avoidance capabilities.
7. NeuralSVG: Generate Editable Vector Graphics from Text
NeuralSVG is an innovative approach that utilizes text prompts to generate vector graphics, providing artists and designers with a flexible and efficient tool. Its core relies on implicit neural representation and score distillation sampling techniques, capable of generating multi-layer structured SVG files. Users can dynamically adjust colors, backgrounds, and proportions during the generation process, greatly enhancing design personalization and flexibility.
【AiBase Summary:】
🖼️ NeuralSVG can generate multi-layer structured vector graphics from text prompts.
🎨 Users can dynamically adjust the color and proportion of generated graphics for personalized design.
✏️ The system can generate sketches with varying stroke counts, showcasing its strong adaptability.
Details link: https://sagipolaczek.github.io/NeuralSVG/
8. Open Source Action Estimation Model ViTPose: Can Estimate Each Frame's Actions and Label Them
ViTPose is an open-source action estimation model focused on human pose recognition. Its uniqueness lies in using a simple and efficient visual transformer technology instead of complex convolutional neural networks. The model structure is straightforward and easy to adjust, finding a balance between performance and speed. ViTPose performs excellently on multiple datasets, even surpassing many complex models, demonstrating the powerful potential of simple models.
【AiBase Summary:】
🦵 ViTPose uses a pure visual transformer architecture, simplifying the model structure and improving efficiency.
📈 The model can be resized according to needs, flexibly adapting to different tasks and datasets.
🌍 ViTPose performs excellently on datasets like MS COCO, showcasing its strong pose estimation capabilities.
Details link: https://huggingface.co/spaces/hysts/ViTPose-transformers
9. OpenAI's Frequent Crawling Causes 3D Model Website to Crash
Recently, the Ukrainian 3D model website Trilegangers experienced severe traffic attacks due to frequent data scraping by OpenAI's crawler GPTBot. Although Trilegangers prohibits unauthorized scraping in its usage agreement, the website's failure to properly set the robots.txt file led to server overload and a crash. To address this crisis, Trilegangers has taken measures to change the robots.txt file and use Cloudflare to block crawler access, ensuring the website operates normally.
【AiBase Summary:】
🛑 The Trilegangers website crashed due to frequent scraping by OpenAI's GPTBot, leading to server overload.
🔧 The website's failure to properly set the robots.txt file did not effectively prevent crawler access, resulting in a traffic attack.
🌐 Trilegangers has taken measures, including setting the correct robots.txt file and using Cloudflare to block crawlers, ensuring normal website operation.
10. LG Announces Entry into Humanoid Robot Market, Plans to Launch Subscription-based AI Agent Q9
LG Electronics announced at CES 2025 that it will increase investment in humanoid robot research and development, planning to launch a humanoid robot based on the AI agent Q9 to enhance household productivity. LG has also established a strategic partnership with Microsoft to promote AI technology applications in various scenarios including homes, automobiles, and hotels. Facing competitors like Tesla, LG aims to secure a place in this field through technological innovation and improved user experience.
【AiBase Summary:】
🤖 LG plans to launch a humanoid robot based on the AI agent Q9 in 2025, aiming to enhance household productivity.
🤝 LG has established a strategic partnership with Microsoft to promote AI technology applications across various scenarios, including homes, automobiles, and hotels.
🏃♂️ The humanoid robot competition is intensifying, with competitors like Tesla actively developing related technologies, presenting broad market prospects.
11. AI Giants OpenAI, Google, etc. Purchase Unpublished Videos from Creators, Up to $4 Per Minute
Recently, tech giants like OpenAI and Google have been directly purchasing unpublished videos from content creators at prices ranging from $1 to $4 per minute for training their AI models. This trend not only provides creators with a new source of income but also sparks fierce competition in the market. High-quality video materials, such as 4K and drone footage, receive higher bids, while ordinary unused videos are priced relatively lower.
【AiBase Summary:】
💰 OpenAI, Google, and other companies purchase unpublished videos from creators at prices ranging from $1 to $4 per minute, especially offering higher bids for high-quality materials.
📹 Contracts include protection clauses ensuring AI companies cannot create digital replicas of creators, and creators can control AI's use of their public videos.
🌐 This trend brings unexpected benefits to content creators, allowing many unpublished materials to generate actual income through sales.
12. Four Departments: Formulating National Occupational Standards for AI Training and Data Annotation
Recently, the National Development and Reform Commission and other departments jointly issued implementation opinions aimed at promoting the high-quality development of the data annotation industry, seeking to advance the professionalization and intelligence of the data annotation industry and foster innovation and application of artificial intelligence. The policy emphasizes talent introduction and training, promotes the integration of industry, academia, and research, and supports the connection of occupational qualifications and skill levels in the data annotation field, aiming to achieve significant improvements in industry scale and innovation capabilities by 2027, with an expected annual compound growth rate exceeding 20%.
【AiBase Summary:】
🎓 Cultivating high-end professionals and formulating national occupational standards to enhance the professional level of the data annotation industry.
📈 By 2027, promoting the scale of the data annotation industry, with an expected annual compound growth rate exceeding 20%.
🤝 Strengthening industry cooperation, promoting public data annotation demand, especially in the fields of government services and urban governance.
