00:00We also have been working on physically embodied agents for a long time.
00:05We call them robots.
00:06And the AIs that they need are physical AIs.
00:09We have some big announcements here.
00:11I'm going to just walk through a few of them.
00:14110 robots here.
00:16Almost every single company in the world, I can't think of one,
00:20that are building robots is working with NVIDIA.
00:23We have three computers.
00:24The training computer, the synthetic data generation and simulation computer.
00:28And, of course, the robotics computer that sits inside the robot itself.
00:32We have all the software stacks necessary to do so.
00:35The AI models to help you.
00:38And all of this is integrated into ecosystems around the world
00:42and all of our partners from Siemens to Cadence, incredible partners everywhere.
00:48And today we're announcing a whole bunch of new partners.
00:52As you know, we've been working on self-driving cars for a long time.
00:55The chat GPT moment of self-driving cars has arrived.
00:58We now know we could successfully, autonomously drive cars.
01:02And today we are announcing four new partners for NVIDIA's robo-taxi-ready platform.
01:12BYD, Hyundai, Nissan, Geely.
01:18All together, 18 million cars built each year.
01:22Joining our partners from before, Mercedes, Toyota, GM.
01:29The number of robo-taxi-ready cars in the future are going to be incredible.
01:35And we're announcing also a big partnership with Uber.
01:39Multiple cities.
01:40We're going to be deploying and connecting these robo-taxi-ready vehicles into their network.
01:46And so a whole bunch of new cars.
01:49We have ABB, Universal Robotics, KUKA.
01:54So many robotics companies here.
01:56And we're working with them to implement our physical AI models integrated into simulation systems
02:02so that we could deploy these robots into manufacturing lines all over.
02:07We have Caterpillar here.
02:08We even have T-Mobile here.
02:11And the reason for that is in the future, that radio tower used to be a radio tower
02:17is going to be an NVIDIA aerial AI RAM.
02:20And so this is going to be a robotics radio tower.
02:23Meaning it can reason about the traffic, figure out how to adjust its beam forming
02:29so that it could save as much energy as possible
02:31and increase the amount of fidelity as possible.
02:35There are so many humanoid robots here.
02:38But one of my favorites, one of my favorites is a Disney robot.
02:44You know what?
02:45Tell you what.
02:46Let me just show you some of the videos.
02:48Let's look at that first.
02:57The first global rollout of physical AI at scale is here.
03:02Autonomous vehicles.
03:04And with NVIDIA Alpamayo, vehicles now have reasoning,
03:08helping them operate safely and intelligently across scenarios.
03:14We ask the car to narrate its actions.
03:16I'm changing lanes to the right to follow my route.
03:21Explain its thinking as it makes decisions.
03:24There's a double parked vehicle in my lane.
03:27I'm going around it.
03:30And follow instructions.
03:32Hey Mercedes, can you speed up?
03:35Sure, I'll speed up.
03:41This is the age of physical AI and robotics.
03:45Around the world, developers are building robots of every kind.
03:49But the real world is massively diverse, unpredictable, full of edge cases.
03:56Real world data will never be enough to train for every scenario.
03:59We need data generated from AI and simulation.
04:04For robots, compute is data.
04:08Developers pre-train world foundation models on internet scale video and human demonstrations.
04:13And evaluate the model's performance to prepare them for post-training.
04:19Using classical and neural simulation, they generate massive amounts of synthetic data and train policies at scale.
04:29To accelerate developers, NVIDIA build open-source Isaac Lab for robot training and evaluation and simulation.
04:37Newton for extensible and GPU-accelerated differentiable physics simulation.
04:43Cosmos World Models for neural simulation.
04:47And Groot Open Robotics Foundation Models for robot reasoning and action generation.
04:53With enough compute, developers everywhere are closing the physical AI data gap.
05:01Paritas AI trains their operating room assistant robot in NVIDIA Isaac Lab, multiplying their data with NVIDIA Cosmos World Models.
05:10Skilled AI uses Isaac Lab and Cosmos to generate post-training data for their skilled AI brain.
05:17They use reinforcement learning to harden the model across thousands of variations.
05:24Humanoid uses Isaac Lab to train whole body control and manipulation policies.
05:30Hexagon Robotics uses Isaac Lab for training and data generation.
05:36Foxconn fine-tunes group models in Isaac Lab.
05:40As does Noble Machines.
05:43Disney Research uses their Kamino Physics Simulator in Newton and Isaac Lab to train policies across their character robots in
05:52every universe.
06:10Disney Research uses their Kamino Physics Simulator in Newton and Isaac Lab to train policies across their character robots in
06:11every universe.
06:17Disney Research was the same, as evaluative model across their team Mae Lab instead of...!
06:40Disney Research uses their ownakboard logic lazy-jo swordskiffes�로 wiege Carp
07:01Untertitelung des ZDF für funk, 2017
07:21Untertitelung des ZDF für funk, 2017
07:48Untertitelung des ZDF für funk, 2017
07:51Using this Newton solver that runs on top of NVIDIA Warp that we jointly developed with Disney and with DeepMind
08:00that made it possible for you to be able to adapt to the physical world. Check that out.
08:08That's how smart you are.
08:11I'm a snowman, not a Snokelope.
08:17Could you imagine this? The future of Disneyland? All these robots, all these characters wandering around.
08:26You know, I have to admit though, I thought you were going to be taller.
08:30I've never seen such a short snowman, to be honest.
08:34Nope.
08:36Hey, tell you what. You want to help me out?
08:39Hooray!
08:41Okay. Usually, usually I close the keynote by telling you what I told you.
08:47We talked about inference inflection. We talked about the AI factory.
08:51We talked about the open claw agent revolution that's happening.
08:55And of course, we talked about physical AI and robotics.
08:59But tell you what. Why don't we get some friends to help us close it out?
09:03Of course!
09:04All right, play it.
09:06Come on.
09:08Terminating simulation.
09:16Hello?
09:21Anybody here?
09:47The keynote's over.
09:49All was said.
09:50Jensen mapped the road ahead.
09:52AI factory's coming alive.
09:55Agents learning how to drive.
09:57From open models to robots too.
09:59Now we'll break it all down for you.
10:05Compute exploded.
10:06Well, we saw from CNN's to open claw.
10:10Agents working across the land.
10:12But they need the power to meet demand.
10:14So we solved the problem.
10:15It was brilliant.
10:17We multiplied compute by 40 million.
10:27Well, once upon an AI time trend was the paradigm.
10:34Sure, it taught the models how.
10:37But inference runs the whole world now.
10:39Vera shows us who's the boss at 35 times less the cost.
10:43Blackwell makes the token sing.
10:45NVIDIA, the inference king.
10:51Yeah, our factories once took years.
10:54Vendors pulling racks and gears.
10:56Built up slowly piece by piece.
10:58No clear way to scale this piece.
11:01DSX and Dynamo know what to do.
11:05Turning power into revenue.
11:14Agents used to wait and see.
11:16Now act autonomously.
11:18But if they ever try to stray.
11:20Safe cause block and say no way.
11:22Nemo claws there to guard the course.
11:26And yes, my friends.
11:30It's open source.
11:40Cars that think and droids that run.
11:42This ain't the movies.
11:43It's all begun.
11:44Alka Mayo calls the shots.
11:46It's a GPT moment for the bots from sim to streets.
11:50Now watch them drive.
11:52Blow your hands up.
11:56For physical AI.
12:10Industrial age.
12:11Build what came before.
12:13Now we build for AI.
12:14Even more.
12:15Vera Rubin plus Grok.
12:16Make the inference splash.
12:17Put them together.
12:18Now it's raining cash.
12:19We build new architecture every year.
12:21Cause claws keep yelling more tokens here.
12:23The AI stacks for all to make.
12:26So let us all eat five.
12:27Lay a cake.
12:28The moment's bright.
12:29The path is clear.
12:30Cause open models led us here.
12:32When data's missing, there's no dispute.
12:34We just generate more with compute.
12:37Robots learning without flaw.
12:39Fueling the force.
12:40Scaling laws.
12:41The future's here.
12:42Won't you come and see?
12:43Welcome all to GTC.
12:58All right.
13:01Have a great GTC.
13:06Wave.
13:08Thank you, everybody.
13:11See you, love.
13:19Für mehr Videos.
13:21Einfach abonnieren.
Kommentare