Welcome to the 【AI Daily】 column! 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 to help you gain insights into technological trends and innovative AI product applications.

Fresh AI Products Click to Learn More: https://top.aibase.com/

1. DeepSeek Daily Active Users Exceed 20 Million, Two Companies Named DeepSeek Established in Hong Kong

Within just 20 days of its launch, DeepSeek has surpassed 20 million daily active users, demonstrating its significant potential and market demand in the artificial intelligence sector. Meanwhile, the parent company of DeepSeek has registered two new companies in Hong Kong, marking the beginning of its internationalization process.

image.png

【AiBase Summary:】

📈 DeepSeek has surpassed 20 million daily active users within 20 days of launch, indicating strong market demand.

🏢 Two companies named DeepSeek have been established in Hong Kong, marking the start of its international expansion.

🔍 DeepSeek has registered multiple trademarks across various fields, showcasing its diversified development strategy.

2. Google Launches the Gemini 2.0 Model Series

Google recently launched its latest Gemini 2.0 model series, including the basic version Flash, the economical version Flash-Lite, and the experimental version Pro, designed to meet different user needs. These models provide a good balance between performance and cost, especially the Pro version, which excels in accuracy for mathematical and complex problem-solving. Adjustments to API pricing make the usage costs more competitive for users with mixed workloads, and the functionalities of the new models are continuously expanding, with plans to support more output formats in the future.

image.png

【AiBase Summary:】

🔹 Google has launched three new Gemini 2.0 models to cater to diverse needs.

📊 Gemini 2.0 Pro performs excellently in mathematical and accuracy tests, scoring higher than its predecessors.

💰 Adjustments to API pricing make mixed workload usage costs more competitive, with Flash-Lite filling a market gap with better performance.

Details link: https://blog.google/technology/google-deepmind/gemini-model-updates-february-2025/

3. OpenAI Unveils Free ChatGPT Search Accessible Without Login

OpenAI has recently announced a significant update, making the search function of ChatGPT freely available, allowing users to access it without logging in. This change enables users to conveniently access a powerful search engine, enjoying rapid information retrieval and parsing capabilities. Compared to traditional search engines, ChatGPT can provide the latest information within minutes and has optimized its user interface for a more friendly experience.

image.png

【AiBase Summary:】

🚀 Users can now use the ChatGPT search engine directly without logging in.

⚡ The search function of ChatGPT is capable of rapidly retrieving and parsing real-time web information.

🖥️ The updated user interface is closer to traditional search engines, enhancing the user experience.

4. World-Class AI Scientist Xu Zhuhong Joins Alibaba

Alibaba Group has welcomed Xu Zhuhong, a top global AI scientist, who will serve as Vice President, focusing on the research and application of multimodal foundational models and agents for AI To C business. Xu's joining will drive innovation and development in Alibaba's AI field, enhancing the end-to-end capabilities of C-end products. His rich academic background and research achievements will inject new vitality into Alibaba's AI business, with expectations for greater breakthroughs in the future.

image.png

【AiBase Summary:】

🚀 Xu Zhuhong serves as Vice President at Alibaba, responsible for developing multimodal foundational models and agents for AI To C business.

📈 He will focus on enhancing Alibaba's AI applications' end-to-end capabilities for C-end products.

🌍 Xu's extensive academic background and research achievements will bring new development opportunities to Alibaba's AI business.

5. Costing Less Than $50! Researchers Train AI Inference Model s1 Comparable to OpenAI's o1

A research team from Stanford University and the University of Washington has successfully trained an AI inference model named s1, with training costs under $50, and performance on par with top models. Using distillation techniques, they extracted reasoning capabilities from existing models, showcasing the potential for low-cost innovation. However, this has raised concerns about the commercialization of AI models, especially as large AI labs face the threat of low-cost model replication.

image.png

【AiBase Summary:】

🌟 The training cost of the s1 model is under $50, performing comparably to top inference models.

🛠️ The research team quickly and efficiently trained the model by extracting reasoning capabilities from existing models using distillation techniques.

🚀 Large AI labs express concerns about the situation of low-cost model replication, with future investments focusing on AI infrastructure.

Details link: https://arxiv.org/pdf/2501.19393

6. First in the Country! Baidu Smart Cloud Successfully Activates Kunlun Chip Generation III Ten Thousand Card Cluster, Significantly Reducing Unit Computing Costs

Baidu Smart Cloud has recently successfully activated the country's first self-developed Kunlun Chip Generation III Ten Thousand Card Cluster, marking an important breakthrough in the field of AI computing power. The ultra-large-scale parallel computing capabilities of the Ten Thousand Card Cluster significantly enhance training efficiency, especially in training complex models. Through self-developed chips, Baidu has not only resolved its own computing power issues but also provided new development ideas for the industry, promoting rapid iteration and innovation in AI applications.

image.png

【AiBase Summary:】

⚡ The Ten Thousand Card Cluster has ultra-large-scale parallel computing capabilities, significantly improving training efficiency.

💡 Self-developed chips effectively solve computing power supply issues, providing a reference for other enterprises.

🔧 The upgraded Baibei AI heterogeneous computing platform 4.0 promotes the effective utilization of computing power.

7. Apple Launches New App Apple Invites for Easy Creation and Sharing of Invitations

Apple has recently launched a new app called "Apple Invites," aimed at helping users easily create personalized invitations for various gatherings and events. The app not only allows users to manage RSVPs and share photos but also integrates with Apple Music to provide collaborative playlists. iCloud+ users can enjoy more advanced features, while everyone can participate in RSVPs. The app is user-friendly, suitable for iPhone devices running iOS 18 and above, and is now available for free on the App Store.

image.png

【AiBase Summary:】

🎈 Users can easily create personalized invitations for birthday parties, family gatherings, and more using the "Apple Invites" app.

📸 The app supports sharing photos and videos, allowing participants to collaboratively create a shared album to preserve precious memories.

🎶 Apple Music subscribers can create collaborative playlists to provide a unique musical backdrop for events, enhancing the party atmosphere.

Details link: https://apps.apple.com/ca/app/apple-invites/id6472498645

8. Firefox Browser Sidebar Introduces AI Chatbot Feature for Direct Access to Popular AIs like Claude

The latest version 135 of the Firefox browser has officially introduced an AI chatbot sidebar feature, allowing users to directly access various popular AIs, including ChatGPT and Claude, within the browser. Users need to update the browser and log into the respective AI accounts to use this feature, which supports both free and paid accounts. The launch of this feature not only enhances user experience but also showcases Mozilla's ongoing efforts in browser innovation.

image.png

【AiBase Summary:】

🆕 The new AI chatbot sidebar feature allows users to directly access multiple popular AIs.

🔑 Users need to log into their respective AI accounts for the first use, with core features available for both free and paid versions.

🔄 The sidebar is movable and closable, allowing users to easily switch between different AIs and view detailed information.

9. OpenAI Applies for Humanoid Robot Trademark, Strategizing for the Future Smart Era

OpenAI has recently submitted a trademark application to the trademark office, indicating its potential entry into the humanoid robot field, planning to launch user-programmable robots with communication and learning capabilities, aimed at providing assistance and entertainment for people. Meanwhile, the company is actively recruiting team members for robotics, promoting the development of general robotic technology. CEO Altman stated that superintelligence will transform scientific discovery and innovation, leading to greater prosperity, and the future will gradually shift towards more advanced intelligence.

image.png

【AiBase Summary:】

🤖 OpenAI plans to launch user-programmable humanoid robots, entering the humanoid robot market.

🛠️ The company is recruiting new team members for robotics, committed to advancing general robotic technology.

🚀 CEO Altman noted that superintelligence will accelerate scientific discovery and innovation, driving human prosperity.

10. Google Tests New "AI Mode" Search Based on Gemini 2.0 Technology

Google is conducting a test of a new feature called "AI Mode," aimed at providing users with a persistent space to pose open-ended and exploratory questions. This mode is based on Gemini 2.0 technology and aims to enhance search intelligence, helping users better organize information and explore related content. It is currently being tested among employees and is expected to be launched before 2025, allowing users to interact with the system through a more intuitive interface, ask follow-up questions, and enjoy a richer search experience.

image.png

【AiBase Summary:】

🌟 Google is testing a new "AI Mode" designed to handle open-ended and exploratory questions.

🤖 This mode utilizes Gemini 2.0 technology to provide users with smarter information organization and link exploration capabilities.

📅 The "AI Mode" is expected to be launched before 2025, currently undergoing employee testing.

11. On the First Day Back to Work, DeepSeek Refuses Interviews from Multiple Media Outlets, Focuses on Its Own Research

On the first day back to work after the holiday, the office of the AI company DeepSeek attracted significant media attention, particularly from Japanese and Korean outlets. However, despite the media's enthusiasm, DeepSeek chose to decline all interview requests, emphasizing its focus on its own research and development.

image.png

【AiBase Summary:】

📅 On the first day back to work, DeepSeek's office attracted many media outlets for interviews.

🔒 The property has enhanced security, requiring all individuals entering DeepSeek's building to undergo verification.

🚫 DeepSeek declined interview requests, emphasizing its commitment to its own research and development.