Boston Dynamics ATLAS 3.0 Demos 6 New AI Upgrades (SPACEO ROBOT)

Number two, AI manipulation. While handling objects comes naturally to humans, robots still struggle with acquiring this level of ingrained know-how, but now, Atlas has integrated advanced artificial intelligence to learn from extensive data sets built from real-world training and simulations. As a result, these data sets train the AI models that enable Atlas to adapt to new manipulation tasks and self-correct, like retrying if an object doesn’t budge as planned. This allows the robot to pick up, move, and place items while constantly improving its own performance, but in order to tie all its movements together into a cohesive action trajectory, this next upgrade is key.

Number three, task sequencing. Atlas now excels at sequencing, which is a logistics process from automotive manufacturing that organizes thousands of unique parts, like engine covers, into exact order for customized vehicle assembly. This update provides Atlas with a clear end goal, such as a dolly filled-in sequence plus a map of the locations of all the required parts. Atlas then autonomously identifies what each slot needs, retrieves the correct cover, and places it inside while self-adapting if a part is missing. This enhances production speed and precision, while doubling as a testbed for developing robots capable of more general-purpose tasks, requiring this next key upgrade.

Number four, advanced vision. The new Atlas integrates multiple cameras throughout its head to perceive its surroundings with superhuman detail. Using RGB cameras, it estimates object positions, tracks them over time, and maps nearby fixtures and containers to sidestep collisions. For instance, when inserting an engine cover, it senses resistance if the fit isn’t quite right, then backs up and adjusts, and then gives it another nudge to ensure it’s secure. As a result, Atlas can expertly interact with complex environments, making it more adept at navigating and manipulating objects in real time, which brings us to robot operating efficiency and upgrade number five, 360 rotation.

Atlas can now rotate a full 360 degrees, allowing it to pivot swiftly, turning its head to scan behind itself without repositioning its entire body. It’s a breakthrough because this design speeds up task transitions, whether shifting between workstations or checking its surroundings, making the robot more agile in dynamic settings. Plus, this feature cuts down on wasted motion, enabling Atlas to move back and forth faster and handle workflows with greater fluidity than its predecessor. And all of this allows for upgrade number six, autonomous architecture. Possibly the greatest breakthrough is that sequencing has been upgraded with a structured framework.

This means that Atlas receives a description of the desired outcome, such as a dolly filled with engine covers, as well as details on where to find parts. It then merges these inputs to execute the entire task from start to finish, identifying the next slot’s requirement, locating the matching cover and placing it all while adapting to hiccups like missing items. In fact, this level of autonomy showcases a serious leap in decision making, allowing Atlas to tackle complex multi-step processes with minimal oversight. As for the robot’s price, the current model is expected to cost around $150,000.

But there’s another new robot that will cost just a fraction of this as Mux Robotics just unveiled SpaceO, a next generation humanoid robot that features a more budget friendly design for AI automation. In fact, SpaceO is engineered to blend strength, precision and intelligence with the goal of operating alongside humans in varying environments, setting the pace for India’s global robotics race. And speaking of strength, SpaceO is capable of lifting a jaw-dropping 200 kilogram payload, tackling heavy duty tasks effortlessly from industrial lifting to logistical hauling. As for movement, its arms feature seven degrees of freedom for basic flexibility, while its dynamic waist rotation and adjustable height of up to eight feet makes SpaceO a humanoid that is adaptable for many sizes and flavors of challenges.

But brute strength is only half the story as SpaceO’s brain, Fusion Max Omni-modal AI, operates completely offline, ensuring real-time secure intelligence. This enables autonomous navigation and task learning, letting SpaceO adapt seamlessly to new environments without constant reprogramming. Additionally, the robot features full sensory interaction via vision and voice makes for more intuitive human-to-machine interaction. But generative AI is also pushing the envelope this week, as LTX Studio, a leading text-to-video generation platform, just announced its beta is over with the introduction of enhanced tools for video creation. One of these upgrades is improved frame control, allowing users to fine-tune camera angles and movements with precision akin to traditional filmmaking.

This feature converts static scenes into dynamic 3D environments, offering creators unprecedented command over their narratives. It all works using the LTX video model, which is an open-source innovation that generates high-quality videos, 5 seconds at 24 frames per second, in 768 by 512 resolution, in just 4 seconds on an NVIDIA H100. And with the release of LTX V0.95 for added commercial licensing, keyframe conditioning, and longer sequences with fewer artifacts, it’s quickly becoming a go-to tool for professional-grade production. Additionally, LTX Studio’s AI now excels at character consistency, maintaining identities across scenes, and integrates audio features like generated soundtracks and voiceovers.

Its offline-capable, multimodal design ensures accessibility, while multilingual support broadens its reach. From storyboarding to pitch-ready decks, LTX Studio is empowering users to craft cinematic stories effortlessly, proving that AI can rival human creativity in video production after all. And finally, Microsoft just introduced Dragon Copilot to be the leading AI assistant in healthcare, with the extensible workspace-blending natural language, ambient speech, and generative artificial intelligence together to deliver a seamless experience across mobile, web, desktop, and top electronic health records. And because it’s built on a modern architecture with enhanced security and healthcare-specific safeguards, Dragon Copilot can adapt to all specialty and care settings, starting with streamlined documentation.

Plus, clinicians gain unprecedented customization through ambient note creation, natural language dictation, custom texts, templates, and AI prompts, making record creation and upkeep easier and more precise. Beyond documentation, however, the platform still keeps critical data at clinicians’ fingertips without breaking workflow by querying notes, retrieving medical insights, and suggesting encounter recordings to ensure vital information is instantly accessible. Additionally, integration with Microsoft Fabric unlocks point-of-care data for healthcare organizations to derive large-scale insights effortlessly. And what might be the biggest feature, with a single click, Dragon Copilot summarizes evidence, preps orders, drafts referral letters, and generates after-visit summaries, dramatically slashing time spent on routine tasks.