New AI Hardware Lets Monster Humanoid Robot Do This (AI NEWS)

So beginning with Paxini’s hands, which feature not only some of the toughest materials in any robot seen yet, but also some of the highest resolution tactile sensors on any robot hand. Featuring 1,140 intelligent tactile processing units per fingertip, and it comes in four-fingered and five-fingered versions. So this brings us to Paxini’s destruction testing. How strong do fingertips need to be to survive a real world wear and tear? So here they demonstrate its slash resistance, flame resistance, brush resistance, abrasion resistance, as well as steam resistance, and it even features chip-level stray field immunity to prevent random magnetic disturbances.

In fact, its fingers are so perceptive of fine textures that it can even identify materials just by touching them. Like here, it’s predicting that it’s touching gauze with a 97.5% level of confidence, or here it’s touching leather wear and predicting its makeup. And the hands feature an 8 megapixel HD AI hand camera in each palm, and this helps the robot estimate how to grasp different objects that it encounters, which I’ll show you in just a second. But first, here are some of the examples of its current speed and precision, and tell us in the comments how this compares to other robots.

But this is the Dex H13, and it features 13 active degrees of freedom, with a payload capacity of 5 kilograms or about 11 pounds. Now, let’s take a look at the robot itself. This is Paxini’s new Tora Double 1, which is a general-purpose humanoid built for both home and industrial use. But let’s take a look at its specs first here. So in terms of the robot’s whole-body perception, it features 3D LiDAR, a head-mounted depth camera, a fisheye camera, as well as a waist-mounted depth camera. And let’s see what it can do around the house thanks to these sensors.

So it’s able to accomplish different tasks using its in-house developed omnivision tactile language action model, which is just a fancy way of saying that the robot’s AI system fuses vision, touch, language, and action together into a unified understanding, and then it uses the robot’s PX-Core control architecture to carry out these actions in the real world. And now for its folding chassis design. So before they release a two-legged version of the Tora Double 1, it’s currently using a four-wheeled base on a hinge. And this features a 21.5 centimeter reach for stair climbing, which is about eight and a half inches.

And tell us whether you’d prefer a two-legged version or a wheelbase for this robot in the home, because in the factory, this could make more sense with everything having a flat surface. For example, in this industrial demo, we can see the robot using multi-point vision coordination for pick and place tasks in the warehouse. And in terms of the robot’s height, it stands at up to 1.6 meters, and it can work at different levels. Next, they show off the robot fetching a package just after being shown this QR code for its tracking.

But what else would you want to see this robot do out of the box to justify its price tag? And how much do you think it should cost? Next, what if your robot could make money right now? Well, a laundromat in San Francisco is already proving that automation is ready and profitable for business. And we’ll show you in just a second what it can do at home as well. But here you can see the robot folding clothes autonomously in a shop. And it’s actually serving a little bit of marketing here too, if you see the girl looking through the window, as well as various people walking by filming and likely sharing to social media.

So would this justify the cost for business and what can it do in the home? Here we also see a video of the Isaac robot doing various tasks around the house. But tell us whether you’d rather hire this robot or a human if you owned a laundromat, for instance. Meanwhile, Google’s Gemini just released its brand new live speech-to-speech translation app. And this is quite impressive because it holds two-way conversations in over 70 languages with 2,000 language pairs. And here’s how it works. Okay, so as I was saying, the Gemini model now does translation live.

It works on the Google Translate app with any headphones. As you can hear, I am being translated in real time. Let’s see what it’s like when I ask for advice for my next trip for my colleagues. And soon enough, these kinds of humanoids are going to be trained by synthetic data that’s generated by the likes of Google’s upcoming Gemini 3 Flash, which allegedly leaked this video, as Google is currently publicly testing two unreleased models on Ella Marina, codenamed Fierce Falcon and Ghost Falcon. And this testing follows their established pattern of pushing preview versions before public versions.

Plus, it appears to be time to counter OpenAI’s upcoming models, which have also been quietly surfacing. And when it comes to the next level of robot training, Mark Riberts of Boston Dynamics just unveiled the Ultra Mobile Vehicle, which is a robot that taught itself bunny hops and spins via simulation. And it looks just like a human riding a BMX bicycle, but it transferred these skills to reality without any programming. And its split mass design mimics a human rider shifting its weight for balance. Unitree just released its own app store for humanoid robot actions.

And it’s essentially an action app store, which allows you to find pre-programmed functions, dances, stances, different things that you can have the robot do in the real world. But tell us in the comments which actions need to be pre-programmed into the robot out of the box. And finally, as a warning, viewer discretion is advised for this next story. So TouchLab just demoed a robotic hand performing a medical procedure on a cucumber. And you tell me down below if you recognize what exactly is going on here. Some of you might not know what you’re watching.

In fact, this demo used a shadow hand robot fitted with TouchLab’s proprietary fingertips in order to perform this delicate manipulation. And the company says its sensors can withstand repeated use while delivering fine grain pressure feedback needed for threading thin medical instruments. And the system maintained what TouchLab describes as a secure yet gentle grip throughout the procedure. And TouchLab states the technology could advance robotic systems for complex manipulation tasks. Though the company explicitly notes that this research is for a demo only and it’s not a medical device approved for human procedures.

But would you trust a robot doing a medical procedure on you or would you rather a human? Anyways, like and subscribe. Check out this video here for more AI news. And thanks for watching. Bye. [tr:trw].