AI Morning News, April 18, 2025
1. Alibaba's Tongyi Wanxiang has open-sourced the "First and Last Frame Video Generation Model" with 14 billion parameters. It supports the generation of 720p high-definition videos and can achieve special effect changes and camera control.
2. Doubao 1.5 Deep Thinking Model has been released, enhancing capabilities in mathematics, programming, and creative writing. Combined with visual understanding functions, it can assist in travel and project management, among other things.
3. OpenAI has introduced the Flex processing API option, offering lower model prices, but with slower response times.
4. The Shanghai Artificial Intelligence Laboratory has released the "Shusheng・Wanxiang 3.0" multimodal large model, improving its basic capabilities and demonstrating excellent performance in fields such as architectural drawing understanding.
5. Google's Gemini Live feature is now available to all Android users, supporting real-time recognition of camera and screen content, and enhancing the user interaction experience.
6. Microsoft Copilot Vision has been launched for free on the Edge browser and can interpret screen content through voice commands.
7. Fudan University has developed the "PoX (Break of Dawn)" picosecond-level flash memory device, with a read-write speed of 400 picoseconds, surpassing SRAM technology.
No comments:
Post a Comment