AI News

Wang Xingxing, founder of Yushu Technology, recently stated that companies that first integrate large models with robotic hardware will lead global AI development, and their technological breakthroughs are of great significance, even enough to win the Nobel Prize. He believes that the integration of large models and robots is a key indicator of whether an AI company has reached world-class standards.

Meta adjusts its policy for third-party AI bots on WhatsApp, allowing developers in specific regions to rejoin the service through a 'per-message billing' model under pressure from Italian regulators. Previously, Meta had blocked third-party bots like ChatGPT to protect its own AI products.

OpenAI CEO Sam Altman showcased his primary phone, the Apple iPhone Air with a MagSafe wallet, at a public event, praising its 'very cool' design and expressing a desire to upgrade.....

China's 2025 'Top 10 Tech Buzzwords' highlight AI trends, including 'Intelligent Agents,' 'Embodied AI,' and 'AI Governance,' selected via data analysis and expert review.....

ByteDance is accelerating its AI hardware layout, planning to launch the second generation of Douyin phones in 2026 and develop them with Nuoyibia. At the same time, the company adopts a self-research and development + cooperation strategy, expanding ecological collaborations such as ride-hailing.

Nvidia, Amazon, and Microsoft are in talks to invest around $600 billion in OpenAI to compete with Google, with Nvidia potentially contributing $300 billion, Amazon considering $100-200 billion, and Microsoft continuing its support, highlighting tech giants' strategic focus on AI.....

Kunlun Wanwei's open-source SkyReels-V3 integrates three key functions: reference image-to-video conversion, video extension, and audio-driven virtual avatars. It supports 1-4 reference images, accurately preserving subject identity and composition, advancing video generation into a high-fidelity, multimodal era.....

Apple App Store exposed for major loopholes, allowing dozens of 'AI nude' apps to generate fake explicit images without user consent, including minors.....

Ant's LingBot-World, an open-source interactive world model, offers a high-fidelity virtual training environment for embodied AI and autonomous driving. It simulates physics for low-cost digital training and transfers learned causal behaviors to real-world applications, addressing data scarcity and high training costs.....

Mushroom IoT's 'Lingzhi AI' vertical large model has been registered with the National Internet Information Office, becoming a benchmark AI product in the industrial auxiliary energy sector. It focuses on general industrial equipment and auxiliary energy, addressing industry needs through four core capabilities to enhance industrial intelligence.....

MiniMax launches Music 2.5, aiming to break through the bottlenecks of control and authenticity in AI music. The new version supports 14 paragraph-level tags for control, such as prelude and bridge, allowing creators to precisely arrange structures and achieve professional-level composition.

Microsoft's latest financial report shows that its AI strategy is entering a period of returns. In the previous quarter, the investment in OpenAI brought $7.6 billion in net profit, effectively responding to external doubts. The deep collaboration between the two companies has been transformed into a stable cash flow. After OpenAI restructured as a non-profit company, it reached a $250 billion Azure cloud service purchase agreement with Microsoft.

Samsung confirms its next-gen AR glasses will launch in 2026, positioned as a 'multimodal AI carrier' with immersive AI experiences via voice, vision, and smart assistance.....

OpenMind launches a robot-specific app store to address slow software adaptation, enabling legged and humanoid robots to update apps like smartphones via its OM1 modular OS, enhancing real-world flexibility.....

The internal leak of Google's Gemini series AI model 'Snow Bunny' has sparked industry attention. The model is described as a revolutionary checkpoint that can build a complete application in one go, showcasing its powerful capabilities.

Google's next-generation AI model, Gemini 3.5, has been leaked. The internal test version named Snow Bunny demonstrates strong engineering capabilities, allowing the construction of complete applications at once and supporting the generation of up to 3,000 lines of executable code in a single prompt. Google also prepared multiple specialized models, such as Fierce Falcon, focused on specific fields, indicating significant breakthroughs in programming.

Samsung Electronics posted strong 2025 results, driven by surging chip demand from the global AI race. Full-year operating profit rose 33.2% to KRW 43.6 trillion, with sales up 10.9% to KRW 333.6 trillion and net profit increasing 31.2% to KRW 45.2 trillion. Q4 performance was particularly strong, with operating profit hitting a record high.....

Moonshot推出的Kimi K2.5 model, its Kimi Agent office capabilities have significantly improved, entering the "Mastering Office" era. This AI can now deeply process Excel, Word, PDF, and PPT files, directly generating professional-level results. The core upgrade lies in an exponential leap in efficiency, shortening mechanical tasks that originally took hours or even days to just a few minutes, such as extracting key information from hundreds of pages of PDF.

OpenAI is secretly developing a new social platform using biometrics to eliminate bot accounts, aiming for a human-only environment and promoting ChatGPT and Sora, seen as a counter to Musk's X platform.....

On January 29, JD Cloud announced the launch of Moltbot Cloud Service on its lightweight cloud servers, achieving a leap from local deployment to one-click activation in the cloud. Users can complete the configuration in just three steps. As a recent hot open source Agent project, Moltbot breaks down application barriers with its unique interactive logic, allowing users to interact directly without opening specific websites or apps.