Daxiao Robot unveils Kaiwu World Model 3.1 at WRC
Two threads in today's AI news: the embodied-AI race has a new open contender, and a top law school is drawing a hard line on classroom AI use.
Daxiao Robot debuted the Kaiwu (Kairos) World Model 3.1 at the 2026 World Robot Conference in Beijing, pairing it with an open-source data-collection stack and three industry solutions for instant retail, hotels, and logistics.
Kairos 3.1 fuses generative, physical, and cognitive intelligence in a single architecture, compressing visual feeds, language instructions, force-touch signals, and motion trajectories into one latent space so a robot can understand a scene, predict how the world will move, and plan an action inside the same model. Daxiao says it tops global embodied-AI benchmarks for world-model video generation and state prediction — beating Nvidia's Cosmos 3 and others — and has open-sourced the model to the industry. The harder bet is the data flywheel underneath it: the company's environment-style collection kit, built around a sense glove that records finger forces down to 0.01 newton and joint angles within two degrees, is meant to turn real commercial deployments into the high-density, physically-grounded training data that world models actually need. If the retail and hotel pilots feed clean data back into the model, Daxiao gets a self-reinforcing loop — but those benchmark wins are vendor-run, and the real proof is whether the robots keep getting smarter once they leave the demo floor.
UC Berkeley Law has set an AI-use policy that, by default, bars students from using AI in class unless an instructor explicitly opts in.
Effective Summer 2026, the policy treats AI as a tool that must stay coupled to human reasoning rather than a shortcut — a notable pivot after two years in which most universities lurched toward tolerating generative tools. Berkeley's own framing is that future lawyers will need to use AI fluently, but the current state of the technology demands that use be paired with the cognition it's meant to support, not replace. The move arrives as law schools and bar examiners are still fighting over how to let AI into legal practice; by defaulting to skepticism and pushing the burden onto professors to permit tools, Berkeley is betting that restraint now beats a permissive free-for-all later.
What to watch: whether other elite law schools follow Berkeley's opt-in default, and whether Daxiao's open-source world model draws real adoption outside China.
Should a top law school default to banning AI in the classroom, or does treating it as forbidden leave the next generation of lawyers at a disadvantage? Tell us in the comments.
