IRON Humanoid Robot Tests 720 AI Vision Act-Two AI Animation New AI Agents Comet AI Browser

➡ Xpeng’s iron robot, equipped with advanced AI features, is now performing tasks in the company’s manufacturing facility and appearing in showrooms in China. Meanwhile, Tesla’s Optimus humanoid robot is making its debut in showrooms across the U.S., with plans to use them in factories. Ultra Robotics has deployed intelligent warehouse robots that adapt quickly to real-world conditions. Daxo Robotics has unveiled a new robotic hand, Muscle V0, capable of ultra-fine movements. Runway’s Act 2, a motion capture technology, and Rechi Mini, an open-source robot for human-robot interaction, have also been launched. A new AI web browser, Comet, aims to make online experiences more intelligent and responsive. Lastly, Anthropic has introduced a directory of applications and tools that can be connected to its Claude platform.

Today on AI News, Xpeng’s iron robot just introduced a series of advanced features that tie directly into the company’s automotive expertise. To start, the iron robot runs on newly updated Turing AI chips, plus it’s equipped with Xpeng’s AI Eagle Eye perception system, which is the same system used in their P7 Plus vehicles. This allows the robot to scan its environment in a full 360-degree view, and even map its surroundings in 720-degree, three-dimensional space. On top of this, development efforts have focused on giving the robot strong walking and movement planning abilities using its reinforcement learning to adapt and improve.

As for intelligence, Xpeng robots and cars share powerful end-to-end models in the company’s AIOS operating system, with both being trained on massive datasets. Currently, these AI robots are already performing tasks at Xpeng’s manufacturing facility while simultaneously appearing in showrooms around China with capabilities such as conversation, memory and reasoning. In fact, Xpeng says their future plans involve thinking robots, and in the United States, Tesla’s Optimus humanoid robot is now making its debut in showrooms across the country, with display units having been installed in various retail locations. Tesla’s goal is to use the curiosity generated to boost engagement with its electric vehicles and other products in its stores.

But the catch is that these showroom units are still static, non-functional prototypes. And so far Tesla has shared clips of Optimus robots working in their labs, but a hard commercial release date remains unannounced. Nevertheless, Tesla plans to utilize its Optimus robot in its own factories, with long-term goals to deploy thousands of humanoid robots for manufacturing tasks. But there’s another contender that may be even closer to the finish line called Ultra Robotics, and they’ve already deployed their own intelligent warehouse robots in the real world, to do real work. These robots can be set up in just a few hours, much faster than traditional systems that often require weeks of deployment.

Plus, they adapt quickly to the unpredictable conditions of real warehouses. On top of this, Ultra’s robots are already packaging e-commerce orders in third-party logistics centers across the United States. Furthermore, customers get immediate value thanks to remote teleoperation on day one, and the robots become fully automated over time as they learn from real-world data. While these robots move slightly slower than humans, their average throughput is already comparable. And taking a peek into the future of robot dexterity, Daxo Robotics just unveiled its new Muscle V0 robotic hand with some next-generation engineering tweaks. First off, unlike traditional rigid robotic hands, Muscle V0 is built on soft, continuum-based mechanisms that allow for theoretically infinite degrees of freedom.

This means it is capable of ultra-fine movements with actuated control points in the hundreds, which is well beyond conventional designs. Additionally, its architecture incorporates not just the shape of the human hand, but also the underlying principles of human muscle control, such as compliance, redundancy, and built-in proprioception. And Muscle V0 is designed from the ground up for machine learning applications, where rich sensing enables high-dimensional force feedback, with the system supporting both teleoperation and simulation for data collection. But what’s more, is that the team even developed custom deformable body simulators to handle the complexities of such a soft system.

As for design, Daxo Robotics put an emphasis on durability, robustness, and a user-friendly graphical interface to serve as a practical tool for real-world tasks and research in dexterous manipulation. And in the latest generative AI breakthroughs, a major leap forward in motion capture technology has just arrived with the launch of Runway’s Act 2, an advanced model designed to deliver high-fidelity animation across head, face, body, and hand tracking. Building on the foundation set by its predecessor, Act 2 now offers substantial improvements in consistency and motion realism. And to pull it off, the system only requires a driving performance video and a reference character.

Plus, Act 2 stands out for its ability to animate virtually any character, regardless of environment, style, or art direction, without sacrificing performance quality. The model even supports comprehensive facial expression capture, upper body movement, intricate hand gestures, and background dynamics, allowing for a new level of creative freedom. As for availability, Act 2 has already released to all enterprise customers and creative partners, with plans to expand access to everyone else soon. But there’s another new contender that has emerged in the world of accessible robotics called the Rechi Mini, and it’s an expressive, open-source robot built for human-robot interaction, creative coding, and AI experimentation.

At just 28 centimeters tall and weighing 1.5 kilograms, Rechi Mini arrives as a compact, desktop-friendly kit with both light and fully autonomous options, plus a starting price of $299. As for what it can do, Rechi Mini is fully programmable in Python, with support for JavaScript and Scratch on the way, opening doors for educators, developers, and families looking to dive into robotics. And with 15-plus plug-and-play behaviors available at launch as well as seamless integration with the Hugging Face community, users can upload, share, and download new robot behaviors as the platform grows. As for its hardware, it features motorized movement, a camera, microphones, and speakers for rich multimodal interaction.

Hugging Face is hoping that the robot’s open-source hardware, software, and its intuitive simulation software developer kit can all help Rechi Mini become a go-to tool for hands-on AI learning, research, and creative exploration. Meanwhile, a new AI web browser named Comet is aiming to redefine how humans use the Internet, focusing on making online experiences more intelligent, fluid, and responsive. Instead of simply browsing or searching, Comet is designed as a thinking companion for users, an assistant that blends curiosity, action, and accuracy into one seamless platform. At the heart of Comet is an integrated assistant that keeps track of context, enabling users to ask questions anywhere online, compare products, or find additional information without switching tabs or losing their workflow.

And for even more power, Comet allows users to turn questions into actions, such as booking meetings, sending emails, or making purchases, all within the browser itself. On top of this, Comet solidifies its trustworthiness and accuracy by using perplexity’s reputation for reliable answers. By combining all of this, each web page allows users to highlight any text for instant explanations or explore related ideas, all tailored to their unique interests. Comet has already been released for use by perplexity Mac subscribers, with invite-only access expanding over the summer. And finally, Anthropic has just introduced a comprehensive directory of applications and tools that can be connected to its Claude platform with a single click.

Users can now easily browse and link Claude to services such as Canva, Figma, Linear, Notion, Stripe, and many more. Plus, all available integrations can be explored on Claude.ai directory. On top of this, remote integrations are accessible for paid subscribers on web and desktop, while local desktop extensions are open to all users via Claude desktop. Anyways, make sure to like and subscribe for more AI news and check out this video here. [tr:trw].