But the robot’s AI motion control is even more impressive thanks to its combined model-based and learning-based locomotion systems, giving the robot a walking speed of 5.4 km per hour. Additionally, its robust control systems enable it to easily navigate confined or unpredictable human environments with remarkable precision, and with an average runtime of four hours, it’s able to tackle long-form tasks that may be dangerous or repetitive for humans. On top of this, the robot’s general adaptability is another standout feature. Designed with modularity in mind, the HMND01 can be customized with several different upper body, lower body, and end-effector configurations that are each tailored to specific tasks.

And with a payload capacity of 15 kg, the robot is first expected to be applied to manufacturing and logistics tasks, expanding these abilities over time as it learns new tasks. In terms of its price, this robot is expected to cost between $60,000 to $110,000 to start. But another humanoid is about to hit the market first, as Menti Robotics just released its MentiBot V3, a fully vertically integrated humanoid robot that combines proprietary hardware with full-stack artificial intelligence. Designed to excel in both industrial and service applications, MentiBot V3 integrates several new tech and intelligence advances to push the boundaries of robot performance, dexterity, and operational efficiency.

Standing at a height of 175 cm, MentiBot V3 is built to emulate human-like abilities but is intended to exceed them in both power and precision. In fact, one of this humanoid’s most remarkable features is the incorporation of custom-built actuators, which deliver up to three times more power compared to its competitors, with these actuators enhancing the robot’s precision and power density in order to stand out in environments where both strength and accuracy are critical. Plus, the robot’s enhanced mobility is another highlight, particularly with its redesigned torso which enables greater dexterity and agility during task execution. This improvement is even further complemented by its redesigned hands, which now feature a stronger grip, impact resistance, as well as the ability to manipulate objects with a pinch force of up to 30 N per finger, and when combined with an increased lifting and carrying capacity of up to 25 kg, these features make MentiBot V3 a highly generalizable tool for real-world applications like manufacturing, logistics, and healthcare.

Additionally, the robot has also received an upgrade to its onboard sensory suite, including enhanced head-mounted sensors along with fisheye side cameras and a rear camera in order to provide it with a full 360-degree view for complete situational awareness. This comprehensive vision system allows the humanoid to operate effectively in dynamic environments, detecting obstacles and adapting to changing conditions with ease. Another innovative feature is its hot-swappable battery system, which ensures uninterrupted operation, and with a runtime of over three hours, the robot can continue its tasks without downtime for round-the-clock industrial scenarios. Then for the robot’s artificial intelligence, it integrates two onboard Jetson Oren AGX modules for high computational power and advanced AI operations.

But Apple’s AI research division is also preparing for its entrance into robotics with the introduction of the Emotion Framework, designed to bring expressive and natural gestures to humanoid robots. Focused on enhancing human-robot interaction, Emotion combines cutting-edge artificial intelligence with advanced robotics to enable robots to interpret social context and respond with lifelike movements. It works by leveraging large language models to analyze social scenarios and generate context-appropriate motions based on human demonstrations. These motions are then translated into precise robot actions using inverse kinematics, ensuring smooth execution. But what really sets Emotion apart is its iterative refinement process, including gestures fine-tuned with real-time human feedback, allowing continuous improvement in how robots express themselves in dynamic settings.

And Emotion’s hardware partner is the Fourier GR1 robot, seamlessly integrating with Apple’s AI framework to execute nuanced gestures with remarkable accuracy. Meanwhile, generative AI also just reached the next level of realism and speed as Adobe officially launched the public beta of its Firefly AI video generator, bringing text-to-video capabilities directly into its Creative Cloud suite. After a limited testing phase last year, Firefly now empowers users to create 5-second 1080p video clips at 24 frames per second, using text prompts, images, or both. The results can then be fine-tuned through an easy-to-use web interface with controls for style, camera angles, motion, and framing.

And future updates are already in development, including a faster ideation model for quick previews and a 4K version for high-quality output. On top of this, Firefly’s integration with Adobe’s Creative Cloud tools, such as Premiere Pro and After Effects, allows creators to incorporate AI-generated clips into existing projects effortlessly. Designed to enhance workflows rather than replace them, Firefly gives professionals expanded creative flexibility. And to support it all, Adobe has addressed several outstanding ethical concerns surrounding AI by training Firefly exclusively on licensed materials, ensuring the generated content is free of copyright issues. Additionally, no customer data is used in training the AI models, reinforcing Adobe’s commitment to responsible AI practices, and Adobe is offering two subscription plans for Firefly’s premium features.

The standard plan, priced at $9.99 per month, includes 25-second videos, 6 minutes of audio translation, and unlimited access to image and vector tools. Then, the pro plan at $29.99 per month offers 70 videos, 23 minutes of translation, and the same unlimited design access. And finally, Meta just unveiled its breakthrough AI models that successfully convert brain activity into text, making mind-reading brain-computer interface technology a reality. But even more important are the two key advancements that Meta just demonstrated with this tech. The first of these was in using non-invasive brain recordings with MEG and EEG to decode up to 80% of typed characters from 35 volunteers, which were twice as accurate as traditional methods.

In fact, artificial intelligence aligns the brain signals with the proper image representations to reconstruct typed sentences, opening breakthrough possibilities for individuals who’ve lost the ability to speak. The second key advancement mapped how the brain transforms thoughts into language, revealing the neural processes that convert abstract ideas into words and actions. And while challenges such as imperfect accuracy and expensive equipment remain, Meta aims to improve the technology’s practicality and reliability for the likely purpose of developing some kind of commercial offering in the near future. Overall, this technology could bring about the next paradigm of human-to-computer interaction, opening up a brand new world of possibilities much sooner than previously expected.

[tr:trw].