Wang Xingxing of Yushu Technology reports a dramatic 90% success rate in AI-generated code, highlighting rapid advancements in AI capabilities. He also predicts explosive growth in China's humanoid robot industry, with market expansion of 50-100% in H1 2025.
SenseTime introduced its advanced DayDayNew 6.5 large language model and Mynie embodied intelligence platform at WAIC. The upgrades focus on multimodal reasoning and cost efficiency, while the robotics platform demonstrated autonomous presentation capabilities, signaling breakthroughs in AI applications.
SenseTime introduced its 'Wuneng' embodied intelligence platform at the World AI Conference (WAIC) 2025, showcasing advanced capabilities in robotics and smart devices. The platform leverages SenseTime's world model and computing infrastructure to enhance perception, navigation, and interaction, promising transformative applications across industries.
At the 2025 World Artificial Intelligence Conference, Zhihui Jun showcased the humanoid robot Lingxi X2 and announced the open-source plan for 'Zhiyuan Lingqu OS,' the industry's first embodied intelligence operating system. The initiative aims to foster collaboration and innovation in robotics, with a layered framework supporting hardware abstraction, intelligent agent services, and developer tools. The system is set to launch in Q4 2025.
Tencent showcased its latest AI advancements at the 2025 World Artificial Intelligence Conference, including the MixFormer 3D World Model and embodied intelligence platform Tairos. The company introduced a comprehensive '1+3+N' system architecture, highlighting upgrades to its intelligent agent development platforms and diverse industry applications.
China's first open-source ocean large model, OceanGPT (Cangyuan), has been launched by Zhejiang University. The model processes marine data, interprets sonar images, and generates robot control code through natural language commands. It aims to enhance underwater robotics and multi-robot collaboration while reducing errors common in general AI models.
China's first open-source oceanic large model, OceanGPT (Cangyuan), debuted in Hangzhou on June 26. Developed by Zhejiang University, it enhances marine research with multi-modal data interpretation and robot control capabilities, reducing errors through its 'slow thinking' mechanism.
Google DeepMind has launched Gemini Robotics, an on-device AI model enabling robots to perform complex tasks like tying shoelaces without internet connectivity. The system uses visual-language-action architecture for precise control and is compatible with multiple robot platforms. Developers can customize functions with minimal demonstrations, while built-in safety mechanisms prevent accidents.
Orbbec has announced major upgrades to its Tian Gong Pro robot, featuring Gemini330 depth cameras for enhanced perception. The company also revealed a new industrial stereo vision solution at Automate2025, signaling advancements in smart logistics and factory automation.
Google DeepMind has unveiled Gemini Robotics On-Device, a new AI model that enables robots to perform complex tasks locally without cloud dependency. The technology features low-shot learning and multi-task capabilities, potentially transforming industries from manufacturing to logistics.
