First Ready AI Robot: ATLAS vs KAI vs GENE 26.5 (AI NEWS)

Third, there’s the hand’s soft contact physics, which are possible due to the hand being covered in biomimetic materials, which replicate the friction and give of the human skin and muscle. And fourth, there’s direct mapping, which allows the human hand motions to be mapped directly to the robot, enabling near lossless information transfer from human demonstrations. But let’s take a look at this cooking demo, as what you’re seeing here isn’t just an AI model, but a full stack system, meaning Genesis co-designed the hardware, data engine, and control stack to ensure the brain and body are totally synchronized, starting with gene 26.5’s robotics native brain with these four key features.

Number one is its unified multimodal model, meaning it processes vision, language, and body position, also known as proprioception, as well as tactile touch, all simultaneously. Number two is flow matching, meaning it predicts multimodal futures, allowing the robot to reason through different ways of physical interaction might play out. Number three is general purpose dexterity, meaning the same model and hardware can pivot between various tasks, using either one or both of its hands for different subtasks. And number four is instant deployment. So the system is capable of zero shot generalization, allowing it to enter a new environment and perform work with as little as 20 minutes of fine tuning data.

And it all centers around five pillars of manipulation, with the first being spatial precision, or accuracy in contact and tool placement. The second being temporal composition, or its timing and speed for complex dynamics. The third being contact richness, allowing the hand to manage multiple touch points, like from its fingertips to its palms. The fourth being contact coordination, so it can synchronize multiple contacts into one coherent behavior. And the fifth being tool mediated interaction, allowing it to use objects to extend its own reach and capability. But to scale this system, the key is its human centric data engine.

And this starts with its tactile gloves, which capture tacit knowledge, or human intuition via EMF tracking and dense tactile sensors. Next is the fact that it uses its work as data, because the gloves are worn during real world tasks, which turned the actual labor into training data without interruption. And finally is internet scale context, which means that it combines egocentric and third-person video to learn natural behaviors. But while Genesis is working on dexterous tasks, Boston Dynamics just released this brand new footage of its Atlas humanoid performing new acrobatics. And this humanoid is designed for car manufacturing inside of Hyundai’s plants.

And in this video, it is not wearing its four finger dexterous hands. But as for some stats, this will weigh 198 pounds or 90 kilograms with 56 degrees of freedom and stand 6.2 feet tall. And it only features two unique actuators to minimize the cost and complexity. And each limb can be swapped in less than five minutes. And to further push the envelope of humanoid robotics, high from kinetics AI just upgraded not only its waist but its shoulders as before it wasn’t able to use them. But now it’s using them during its tasks that are being demonstrated in this latest video of the robot folding a shirt.

And this robot has 115 degrees of freedom throughout its body, with 36 degrees of freedom in each hand. And of those 22 of the joints are motor driven or active, while 14 are compliant or passive, which allows the hand to move more naturally and mold around these objects like cloth. Next, we see this demo of the robot negotiating a panda and an apple and just watch the way that its hand molds around each of these deformable objects. And this is made possible partially because of its 18,000 tactile sensing points, which cover 80% of its body.

And the engineers of the Kai robot were previously working on x pangs robot. So they may have brought some of that technology over to Kai. And as for power, this robot uses a semi solid state battery with 1.7 kilowatts of power and three hours of endurance while managing up to 20 kilograms or 48 and a half pounds with a dual arm payload. And it uses custom actuators with its tactile resolution being sensitive enough to detect as little as 0.1 newtons. But make sure to comment down below whether or not you think that opening up this robot’s garment and revealing its internal components could prove that this isn’t just a human in a suit, as some of you have been suggesting.

And finally, this is gestures 19 joint 10 degree of freedom robotic hand. Currently, it’s on a Kickstarter for fundraising. But this is a modular robot hand that can dynamically manipulate up to one kilogram or 2.2 pounds, or statically lift up to four kilograms or 8.8 pounds. And the hand itself only weighs just under one pound and features silicon fingertips with a desktop app for automatic calibration and for control modes being programmable in both Python and C sharp software developer kits. But as for its speed, its fingers can move at 200 degrees per second, whereas its wrist can move at 100 degrees per second with a control loop speed of 100 Hertz.

But make sure to comment down below whether you would prefer a prebuilt robot, or if you’d rather build your own with modular parts. Anyways, like and subscribe, check out this video here for more of the latest in AI and robotics news, and thanks for watching. [tr:trw].