WAIC Multimodal and Embodied AI Updates

创业邦 · wechat · 2026-07-19

WAIC & Multimodal/Embodied AI Progress

千觉机器人 showcased the first VTLA embodied tactile model, integrating vision, touch, language, and action. They also released a visual-tactile multimodal dataset, data collection equipment, and a tactile sensor matrix, aiming to give robots stronger physical property recognition and interaction capabilities.
Rokid launched YodaOS, a native operating system for smart glasses, emphasizing graphical, immersive interactions over plain text output. It includes features like automatic itinerary logging and timed reminders.
万兴科技 exhibited its AI video creative product matrix, covering film creation, professional video workstations, and AI video commerce tools.
商汤 released its next-gen multimodal base model 日日新 U1Pro, focusing on unifying "understanding, generation, and action" and supporting native 8K output for delivery-grade design scenarios.
An industry observation noted that Indian enterprises are increasingly relying on Chinese open-source LLMs like DeepSeek, 阿里巴巴, and 月之暗面 to significantly cut AI costs.
A B2B global payments white paper suggested that cross-border payments are evolving from digitalization to "autonomization," where AI agents autonomously orchestrate and optimize payment flows upon authorization.

More from Embodied