đ Dein Abo hilft uns: https://tublo.eu/abonnieren â Source: Nvidia âĄïž Mehr in unserem Automagazin: www.tuningblog.eu
Nvidia prĂ€sentiert mit Isaac Groot N1.5 eine Plattform, die Robotik und KI vereint. Mit dem Jetson Thor Prozessor, realitĂ€tsnaher Simulation ĂŒber Newton und digitalen Zwillingen entstehen lernfĂ€hige Roboter, die sich in bestehende Industrieprozesse integrieren lassen.
Dank Cosmos und Groot Dreams lernen Roboter aus wenigen menschlichen Demonstrationen durch KI-generierte Bewegungsdaten â ein echter Durchbruch angesichts globaler ArbeitskrĂ€ftemĂ€ngel. Die Plattform ist offen zugĂ€nglich und bereits tausendfach heruntergeladen.
#Roboter, #Nvidia, #KI, #Robotik, #Simulation, #Zukunft, #Technologie #tuningblog - das Magazin fĂŒr Auto-Tuning und MobilitĂ€t!
00:00Nvidia treibt mit seiner neuesten Plattform fĂŒr Robotik, Isaac Groot N1.5, die nĂ€chste Generation humanoider Roboter und industrieller Automatisierung entscheidend voran.
00:11Die Grundlage dafĂŒr ist ein neu entwickelter Computer namens Jetson Thor, der speziell fĂŒr den Einsatz in Robotern konzipiert wurde.
00:19Er ist leistungsstark genug, um neuronale Netze, Sensorverarbeitung und Aktuatoren in Echtzeit zu steuern.
00:25Alles auf Basis des Nvidia-Betriebssystems namens Isaac.
00:28Damit Roboter effizient lernen können, nutzt Nvidia ein physikalisch-realistisches Simulationssystem auf Basis des neuen Newton-Engines.
00:37Diese erlaubt es, virtuelle Welten zu erschaffen, in denen Roboter ihre Aufgaben trainieren. RealitÀtsnah und in Echtzeit.
00:46Ein weiterer Meilenstein ist Cosmos, ein generatives KI-Modell, das aus wenigen menschlichen Demonstrationen mithilfe von synthetischen Daten ganze Handlungsketten erzeugt.
00:56In Kombination mit dem Groot-Dream-System werden daraus dreidimensionale Bewegungsdaten gewonnen, mit denen Roboter neue FÀhigkeiten lernen können.
01:05Die so erzeugten Daten ersetzen tausende Stunden manueller Demonstrationen.
01:09Dank digitaler Zwillinge von Fabriken und Produktionslinien, die mit Omniverse erstellt werden, können ganze Industrieanlagen inklusive Robotik prÀzise geplant, getestet und optimiert werden, noch bevor sie real gebaut werden.
01:24Isaac Groot ist dabei offen zugĂ€nglich, bereits tausende Male heruntergeladen und gilt als SchlĂŒsseltechnologie in einer Zeit globaler ArbeitskrĂ€ftemangel.
01:33Mit dieser Plattform ebnet Nvidia den Weg fĂŒr humanoide Roboter, die sich in unsere bestehende Welt nahtlos einfĂŒgen.
01:41In einem Markt, der bald ein Potenzial von mehreren Billionen Euro erreichen dĂŒrfte.
01:46Okay, let's talk about robots.
02:05So agent AIs, agentic AIs, AI agents, a lot of different ways to say it, agents are essentially digital robots.
02:16The reason for that is because a robot perceives, understands, and plans.
02:21And that's essentially what agents do.
02:23But we would like to build also physical robots.
02:26And these physical robots, first, it starts with the ability to learn to be a robot.
02:32The ability to learn to be a robot can't be done in the physical world productively.
02:38You have to create a virtual world where the robot can learn how to be a good robot.
02:43That virtual world has to obey the laws of physics.
02:47Most physics engines don't have the ability to, with fidelity, deal with rigid and soft body simulation.
02:54And so we partner with DeepMind, Google DeepMind, and Disney Research to build Newton, the world's most advanced physics engine.
03:06It's going to be open sourced in July.
03:09It's incredible what it can do.
03:13It's completely GPU accelerated.
03:16It's differentiable so you could learn from experience.
03:20It is incredibly high fidelity.
03:23It's super real-time.
03:26And so we could use that Newton engine.
03:30And it's integrated into Mujoco.
03:32It's integrated into NVIDIA's Isaac Sim.
03:34So irrespective of the simulation environment and framework you use.
03:39And so with that, we can bring these robots to life.
03:43Thank you.
03:50Can you imagine one of those little ones or a few of them running around the house, chasing your dogs, driving them crazy?
04:00And so did you see what was happening?
04:03It wasn't an animation.
04:04It was a simulation.
04:07And he was slipping and sliding in the sand and the dirt.
04:12All of it was simulated.
04:13And so in the future, we'll take the AI models that we train and we put it into that robot in simulation and let it learn how to be a great robot.
04:31Well, we're working on several things to help the robotics industry.
04:36Now, you know that we've been working in autonomous systems for some time.
04:39Our self-driving car basically has three systems.
04:42There's the system for creating the AI model.
04:45And that's GB200, GB300.
04:48It's going to be used for that, training the AI model.
04:51Then you have Omniverse for simulating the AI model.
04:55And then when you're done with that AI model, you put that model, the AI, into the self-driving car.
05:02Okay?
05:02This year, we're deploying Mercedes around the world, our self-driving car stack, end-to-end stack.
05:07But we create all of this.
05:10And the way we go to market is exactly the same that we work everywhere else.
05:13We create the entire stack.
05:16We open the entire stack.
05:18And for our partners, they use whatever they want to use.
05:21They could use our computer and not our library.
05:24They could use our computer, our library, and also our runtime.
05:26However much you would like to use, it's up to you.
05:29Because there's a lot of different engineering teams and different engineering styles and different engineering capabilities.
05:34We want to make sure that we provide our technology in a way that makes it as easy as possible for everybody to integrate NVIDIA's technology.
05:42You know, like I said, I love it if you buy everything from me, but just please buy something from me.
05:52Very practical.
05:56And so we're doing exactly the same thing in robotic systems, just like cars.
06:02And so this is our Isaac Groot platform.
06:04The simulation is exactly the same.
06:06It's omniverse.
06:07The compute, the training system is the same.
06:09When you're done with the model, you put it inside this Isaac Groot platform.
06:14The Isaac Groot platform starts with a brand new computer called Jetson Thor.
06:18This has just started in production.
06:20It is an incredible processor.
06:22Basically, a robotic processor.
06:24It goes to self-driving cars, and it goes into a human or robotic system.
06:28On top is an operating system we call NVIDIA Isaac.
06:33The NVIDIA Isaac operating system is the runtime.
06:35It does all of the neural network processing, the sensor processing, pipelines, all of it, and deliver actuator results.
06:42And then on top of it, pre-trained models that we created with an amazing robotics team that are pre-training these models.
06:53And all the tools necessary in creating this we make available, including the model.
06:57And so today we're announcing that Isaac Groot N1.5 is now open-sourced, and it's open to the world to use.
07:05It's been downloaded 6,000 times already, and the popularity and the likes and the appreciation from the community is incredible.
07:15And so that's creating the model.
07:18We also opened the way we created the model.
07:21The biggest challenge in robotics is, well, the biggest challenge in AI overall is what is your data strategy?
07:28And your data strategy has to be, that's where a great deal of research and a great deal of technology goes into.
07:34In the case of robotics, human demonstration, just like we demonstrate to our children or a coach demonstrates to an athlete,
07:43you demonstrate using teleoperations.
07:46You demonstrate to the robot how to perform the task, and the robot can generalize from that demonstration,
07:54because AI can generalize, and we have technology for generalization.
07:57You can generalize from that one demonstration other techniques, okay?
08:01And so what if you want to teach this robot a whole bunch of skills?
08:07How many different teleoperation people do you need?
08:11Well, it turns out to be a lot.
08:12And so what we decided to do was use AI to amplify the human demonstration systems.
08:21And so this is essentially going from real to real and using an AI to help us expand, amplify the amount of data
08:33that was collected during human demonstration to train an AI model.
08:37Let's take a look.
08:38Is that great?
08:39So in order for robotics to happen, you need AI.
08:45But in order to teach the AI, you need AI.
08:48And so this is really the great thing about the era of agents where we need a large amount of synthetic data generation,
08:58robotics, a large amount of synthetic data generation,
09:01and skill learning called fine-tuning, which is a lot of reinforcement learning,
09:05and enormous amount of compute.
09:08And so this is a whole era where the training of these AI, the development of these AI,
09:13as well as the running of the AI needs an enormous amount of compute.
09:17Well, as I mentioned earlier, the world has a severe shortage of labor.
09:21And the reason why humanoid robotics is so important is because it is the only form of robot
09:26that can be deployed almost anywhere brownfield.
09:30It doesn't have to be greenfield.
09:32It could fit into the world we created.
09:34It could do the task that we made for ourselves.
09:37We engineered the world for ourselves, and now we could create a robot that fit into that world to help us.
09:44Now, the amazing thing about human robotics is not just the fact that if it worked,
09:48it could be quite versatile.
09:50It is likely the only robot that is likely to work.
09:53And the reason for that is because technology needs scale.
09:58Most of the robotic systems we've had so far are too low volume.
10:03And those low volume systems will never achieve the technology scale to get the flywheel going far enough,
10:09fast enough, so that we're willing to dedicate enough technology into it to make it better.
10:16But human or robot, it is likely to be the next multi-trillion dollar industry.
10:21And the technology innovation is incredibly fast.
10:24And the consumption of computing and data centers, enormous.
10:27But this is one of those applications that needs three computers.
10:31One computer is an AI for learning.
10:33One computer is a simulation engine where the AI can learn how to be a robot in a virtual environment.
10:41And then also the deployment of it.
10:43Everything that moves will be robotic.
10:45As we put these robots into the factories, remember, the factories are also robotic.
10:53Today's factories are so incredibly complex.
10:58This is Delta's manufacturing line, and they're getting it ready for a robotic future.
11:04It is already robotics and software defined.
11:07And now in the future, there will be robots working in it.
11:09In order for us to create robots and design robots that operate as a fleet, as a team,
11:16working together in a factory that is also robotic, we have to give it Omniverse to learn how to work together.
11:26And that digital twin, you now have a digital twin of the robot.
11:29You have a digital twin of all of the equipment.
11:32You're going to have digital twin of the factory.
11:34Those nested digital twins are going to be part of what Omniverse is able to do.
11:40This is Delta's digital twin.
11:42This is WeWin's digital twin.
11:44Now, while you're looking at this, if you're not, if you look at it too closely, you think that it's, in fact, photographs.
11:53These are all digital twins.
11:55They're all simulations.
11:56They just look beautiful.
11:58The image just looks beautiful, but they're all digital twins.
12:01This is Pegatron's digital twin.
12:03This is Foxconn's digital twin.
12:09This is Gigabyte's digital twin.
12:11This is Qantas.
12:14This is Wishtron's.
12:18TSMC is building a digital twin of their next fab.
12:24As we speak, there are $5 trillion of plants being planned around the world.
12:32Over the next three years, $5 trillion of new plants.
12:36Because the world is reshaping, because re-industrialization moving around the world, new plants are being built everywhere.
12:44This is an enormous opportunity for us to make sure that they build it well and cost-effectively and on time.
12:50And so putting everything into a digital twin is really a great first step.
12:55And preparing it for a robotic future.
12:57In fact, building that $5 trillion doesn't include a new type of factory that we're building.
13:05And even our own factories, we put in a digital twin.
13:08This is the NVIDIA AI factory and a digital twin.
Schreibe den ersten Kommentar