Figure 03 Humanoid Robot Demos VULCAN 1.0: Break A Leg (AI NEWS)

And the reason why this stands head and shoulders above a human performing at a workplace is because when a human becomes injured, they can’t keep working. But these robots will continue and adapt. You can see it hobbling here. And this is all part of a larger system, which I’m about to show you in this next video from the CEO of Salesforce. This is also the figure three. And it’s sorting packages here. This is one X speed, no tell operation completely autonomous. And it’s flipping over these packages. But this is all part of the larger system, which is the Helix zero two.

And what you’re seeing here is system zero, which is responsible for foundational whole body control operating at 1000 Hertz. And this ensures this human like movement, stability and balance. And it allows for the robot to make contact with these packages and coordinate across this entire body. Whereas when you see these packages being thrown in, this is system one working, this is reactive behavioral control. And this system thinks fast and translates perception into action operates at 200 Hertz, and it translates high level goals into immediate movements, like the grasping of these packages, then this integrates with system two, which is responsible for high level reasoning, or slow planning.

And this is taking care of long term goals, language and context. So it interprets its scenes, understands natural language instructions and sequences its behaviors to achieve its goals. And overall, we’re seeing these robots not only operating at a level as fast and competent as unskilled human workers, but also now becoming superhuman in the fact that they’re able to continue working, despite breaking down. But another robot is also serving in a different regard. This is more for entertainment. This is the AGI bot x two, playing ping pong autonomously, no tell operation. All it’s doing is integrating its real time visual perception and motion intelligence to respond to the ball’s speed, spin and trajectory.

And what you’re seeing here is the next generation of embodied AI systems, taking on different real world scenarios here. This is just for entertainment. But you could imagine some other applications that this robot could possibly serve towards. For instance, they also have this robot riding a bicycle as well as operating a balance board. So you can see these humanoids are indeed taking on these adaptive performances. And in some cases, like the figure three even carrying on operation when they’ve already been handicapped by parts breaking down. But the x two isn’t the only robot from AGI bot as the g two robots are also being deployed now on production lines in the real world.

And this is another example of embodied AI serving towards consumer electronics precision manufacturing, where here you can see these g two humanoids autonomously picking up tablets, navigating these factory floors, and placing the devices into their testing fixtures with millimeter level accuracy, which frankly exceeds that of a human. And while they may not be as fast as humans are currently, they’re certainly catching up quickly. So how fast will it be before these not only operate faster than humans cost less than humans, but exceed humans when pieces of their hardware break down mid task.

And so far, these g two robots are operating at a success rate of 99% producing 3000 units per shift. And each cycle here is lasting between 19 to 20 seconds, but they’re operating in this factory 24 seven autonomously with very minimal human intervention. But the real question is, would you rather buy your own robot or build your own robot? Because craft just released its open source hand, which is tendon driven, and costs under $600 to build yourself. And it features these links for strength where it can bend in ways that a human probably wouldn’t be able to.

And it uses these soft rolling contact joints to absorb impact driven by 15 motors in the forearm and it can perform all 33 grasp types. But we still need to see a bit more in the real world. So far, we have some simulation demos of its dexterity. But one thing that does look promising is that because this robot hand can bend in ways that a human hand cannot, it lends to this robot probably being more durable than many of the robots that are already on the market. But tell us down below how much you’d be willing to pay for an entire humanoid robot kit that you build yourself versus one that you purchase pre built.

But we aren’t just seeing AI breakthroughs in the real world. Because overworld just released waypoint 1.5, which is bringing AI generated 3d worlds to both Mac and Windows. So this update allows for real time generation of these 3d worlds, the software runs locally on both Mac and Windows for the first time ever, offering two tiers. The first is 720p at 60 frames per second for high end machines. And the second is 360p for broader hardware like Nvidia RTX cards and eventually Apple Silicon. And the model is half the size of the previous versions, despite being trained on roughly 100 times more data.

And this delivers improved visual quality and efficiency over waypoint 1.0 and 1.1, where users can actually install it locally using the biome runtime environment, or they can stream it in browser at overworld.stream. And while it still looks a bit rough around the edges, it does point to a future where soon, we may have completely synthetic experiences that are generated in real time, custom to our preferences, thus antiquating the traditional video game or entertainment experience. And last but not least, unitry just broke a world record, having its robot run 10 meters per second.

And this robot weighs approximately 62 kilograms. But the question is, why do robots need to run this fast? Who will they be chasing after? Anyways, like and subscribe. Check out this video here for more of the latest in AI and robotics news. And thanks for watching. Bye. [tr:trw].