SPARKY 1 Humanoid Robots New AI Hand Is a GAME CHANGER (10000 SENSORS)

Diamond Robotics has introduced a new robot, Sparky-1, with advanced hand-like features that can mimic complex human movements. The robot’s hands, called DM Hand-1 and DM Hand-2, have multiple joints and degrees of freedom, allowing them to perform delicate tasks like cutting tofu with a potato chip without breaking it. The hands also have a unique sensor system in the fingertips, enabling them to detect light forces and respond quickly. Additionally, the robot’s artificial intelligence uses visual, tactile, linguistic, and action-based inputs to predict and execute precise movements, making it capable of learning and refining its skills in real time.

I’m Ann News, and Diamond Robotics just unveiled a world’s first breakthrough in humanoid dexterity. Plus, Tesla just upgraded its Optimus Robot. But first, imagine if humanoid robots could finally touch and feel the same way that humans could, or with even greater precision. And what could they do as a result with this kind of dexterity? And this is exactly what Diamond Robotics’ latest tech upgrade is addressing, so here’s the full three-part breakdown of everything you need to know about the Sparky-1 humanoid robot and its latest dexterous tech. Starting with the robot’s DM Hand-1 manipulators, which have a biomimetic design featuring 11 joints with 6 active degrees of freedom, and these mimic the most complex human hand movements, like grasping, gripping, and manipulating.

But what really sets it apart is the fact that it features the world’s first millimeter-level vision-based tactile sensor system, which is integrated into the robot’s fingertips, and each fingertip houses over 10,000 sensing units, capable of detecting forces as light as 10 grams at a speed of 120 hertz, and this allows for ultra-fast, ultra-precise responses, enabling feats like tracking the subtle rebound of an inflating balloon, or sensing the tug on a tissue that’s being pulled from its grip. In fact, the robot is so precise that it can even cut tofu with a potato chip without breaking the chip, or it can hold a plastic cup without crinkling it.

But there’s another even newer hand called the DM Hand-2, but this one has a total of 20 degrees of freedom instead, with 15 active degrees of freedom. Plus, the DM Hand-2 combines a parallel linkage transmission structure with its hybrid force position control. And all this is to say that the robot basically understands its own skeletal structure in its hand, and it uses force control for exact positioning and adjustments to control itself ultra-gently. And when it comes to strength, the hand weighs less than 1kg but supports an 18kg payload, and all the sensitivity equips the Sparky-1 with a human-level tactile system which lets it perceive object textures, recognize different materials, and adapt to new scenarios.

But the Sparky-1’s artificial intelligence is even more impressive than its hands, because at its core it features a multimodal VTLA manipulation model. VTLA stands for Vision Tactile Language Action, and this just means that the AI integrates visual, tactile, linguistic, and action-based inputs to predict and execute precise movements. In fact, this closed-loop system maps perception directly to control, enabling the robot to reason, generalize tasks, and operate seamlessly in unpredictable environments. And with previous models featuring processing speeds of 275 tera-operations per second, it means the robot is capable of continuously learning and refining its skills in real time.

But before we get into the robot specs and breakthrough wearable training technology first, a message from today’s sponsor. As AI revolutionizes every sector and transforms the global landscape, are you equipping yourself with the skills to stay competitive? For instance, in today’s job market, AI proficiency is no longer just an advantage. It’s a necessity, because whether you aim to excel in your current position or safeguard your career for the future, mastering artificial intelligence will distinguish you from the rest. That’s why growth school is providing an exclusive three-hour interactive AI training session, crafted to empower professionals like you to master over 25 cutting-edge AI tools, and for a limited time, it’s entirely free for the first 1000 AI news viewers.

Be it marketing, sales, HR, design, product, or even finance, not limited to tech experts, AI skills can optimize your processes, sharpen decision-making, and boost efficiency. Plus, everyone who signs up will gain access to $500 worth of additional resources as a bonus. So seize this opportunity to join the top 1% by building critical AI expertise and click the link below to join Growth School’s WhatsApp Group for exclusive updates. And when it comes to the robot specs, previous videos showed that it can achieve a 220-centimeter vertical reach height, plus it features whole-body motion control with 14 degrees of freedom in its arms, and between 22 to 40 degrees of freedom in its hands depending on which hand is used, plus the robot features 16-bit joint angle feedback with 800-plus hertz sampling frequency for ultra-smooth, low-latency motion capture that rivals human motion.

And on top of this, the robot even uses panoptic segmentation and visual navigation for environmental perception, which allows the robot to interact with its surroundings with an acute level of awareness. And here’s the twist, because it features this portable, low-cost remote operation system for end-to-end data collection to accelerate the robot’s learning, meaning that anyone can strap this thing on and teach the robot simply by doing tasks themselves. And the robot can translate this directly into training data, and the company says they’re conducting a presale right now, but there’s no word on how much the robot will cost just yet.

So tell us in the comments below how much you’d be willing to pay for a robot like this. But what about legged robot locomotion, as Tesla just upgraded its Optimus and released a video showcasing its results, saying accurate actuators accelerate automation. And this just means that the company has improved its actuators to achieve more precision. And users on Twitter are already saying that the Optimus no longer walks like it has a load in its pants. But when it comes to generative AI, Google just upgraded its AI studio to allow developers to build apps directly in AI studio using its newest Gemini model without needing an API key.

And it’s all for free. Plus, the platform provides a ton of sample apps that users can remix and customize and it has a sharing feature to collaborate with friends. But the star of the launch is definitely Gemini 2.5 flash, which now is in preview on the Gemini API in AI studio and vertex AI. And importantly, it builds on Gemini 2.0 flash for a boost in hybrid reasoning performance, while keeping costs and latency low. In fact, developers can toggle the thinking setting to on or off in order to balance their budgets for quality, cost and speed.

But even without thinking 2.5 flash is just as fast as 2.0 flash, but with even better results. And some people on Reddit are highlighting its performance to cost ratio, putting Gemini’s models at the Pareto frontier of price and performance. Plus 2.5 flash also powers features like canvas, which is an interactive tool for refining documents and code. But tell us in the comments, which is better flash 2.5 or these two new models that open AI just released, which are its 03 and 04 mini and these are its smartest models yet. Both can agentically use tools like web search, Python analysis, visual reasoning and image generation to solve complex problems and return thoughtful answers in under a minute.

And open AI’s 03 is the most advanced reasoning model yet in the areas of coding, math, science, and visual perception. And it sets new benchmarks in code forces, SWE bench and MMMU performing strongly in visual tasks like analyzing charts and graphics. And external evaluation show 03 makes 20% fewer errors than 01 on complex tasks, particularly in programming, business and creative ideation. And the smaller 04 mini offers fast cost efficient reasoning and it beats 03 mini and non stem and data science, supporting high throughput use. And both models improve instruction following and conversational naturalness.

And when it comes to visual reasoning, 03 and 04 mini integrate images into their reasoning process. And they solve their problems by interpreting whiteboard photos, diagrams or sketches, even if they’re low quality and they manipulate these images by rotating and zooming them during reasoning to achieve top accuracy and multimodal benchmarks. And on top of their agentic use of chat GPT tools, they can even use custom API functions and they can reason about when and how to use them. And in terms of efficiency, 03 and 04 mini are smarter and often cheaper than 01 and 03 mini respectively.

But tell us what you think in the comments below and check out this video if you want to know more about the latest in AI news. [tr:trw].