🤖 Unitree Gives Robot a BRAIN — Now It’s Acting HUMAN! + More AI & Robotics News 🚀 | AI Revolution - video Dailymotion

Ai Revolution

Unitree just upgraded its robot with an advanced brain, making it behave more like a human! Explore this breakthrough and catch the latest updates in AI and robotics shaping the future. From smart machines to cutting-edge tech — the revolution is here! 🌟🤖 #Unitree #RobotBrain #AIRevolution #Robotics #ArtificialIntelligence #TechNews #FutureTech #SmartRobots #Innovation #MachineLearning #AI

Transcript

00:00In just the past few days, the robotics world has exploded with news that sounds completely unreal.

00:08Except it's all completely real.

00:11Unitree's G1 Humanoid just got a major upgrade with a system called Ammo

00:15that lets it clean your house, pick up toys, open the fridge, and even load the dishwasher

00:20like it's been living with you for years.

00:23Their B2 robot dog now got redesigned for firefighting

00:27with a water cannon that can blast foam nearly 200 feet.

00:31Lenovo stepped into the game with its first humanoid, Légion No One,

00:36which performed Tai Chi live on stage and answered real-time business questions.

00:42And Beijing just announced it'll be hosting the World Humanoid Robot Sports Games this August

00:48inside actual Olympic venues with robots competing in track, gymnastics, and even soccer.

00:55So let's talk about it.

00:56All right, by now, pretty much everyone in the robotics space knows the Unitree G1.

01:02But what really flipped the narrative wasn't the hardware.

01:05It was ammo.

01:07And not just some marketing buzzword.

01:08Ammo or adaptive motion optimization is probably the most advanced real-time,

01:14whole-body control system we've seen on a consumer-level humanoid.

01:19It's what turned G1 from an impressive machine into a robot that genuinely understands how to move like a living thing.

01:26Most robots struggle with complex motion because humanoid bodies are hard to control.

01:32You've got 29 degrees of freedom, nonlinear physics, contact dynamics.

01:36All of that makes it extremely difficult to balance flexibility with stability.

01:41Older methods relied on rigid control systems or motion capture data that didn't really translate well into dynamic environments.

01:49They'd train robots to copy how people move, but not how people adjust in real time.

01:55That's where ammo changes everything.

01:58AMO is built differently.

02:00It combines reinforcement learning and trajectory optimization in a way that lets the robot not only learn how to move,

02:06but how to adapt on the fly.

02:08First, the system runs millions of motion tests in simulation using sim-to-real learning

02:13so the robot can fail over and over without breaking anything.

02:18It learns how to pick things up off the ground, reach high shelves, twist its torso, crouch low,

02:23or even stretch sideways without losing balance.

02:26Then, those lessons get translated into real-world behavior that works.

02:30Not just in theory, but in your kitchen.

02:32So, when you see G1 bend down to pick up a toy or carefully adjust its balance to slide a bottle onto a high shelf,

02:41that's AMO in action.

02:42It's controlling the robot's entire body, legs, torso, waist, everything,

02:48based on an internal plan, not just isolated joint movements.

02:51And it does this in real time, responding to changes in its environment and adapting to unpredictable inputs.

02:58You can even throw it into a teleoperation mode using a VR headset,

03:02and it'll track your movements like a shadow.

03:04But what's crazy is that once you let go, it doesn't freeze.

03:09It keeps going.

03:10It understands the goal and keeps executing,

03:13smoothly transitioning from human guidance to autonomous action.

03:17Under the hood, AMO uses what's called a hybrid motion synthesis pipeline.

03:22That means it blends human-like arm movements from motion capture data

03:26with sampled torso commands to generate new kinds of whole-body actions that weren't even in the original data sets.

03:32The robot isn't just imitating, it's generalizing.

03:35Whether it's yaw, pitch, roll, or height control,

03:40AMO gives the G1 way more flexibility than previous systems.

03:43For instance, it doesn't rely solely on waist motors to tilt the upper body.

03:48Instead, it shifts its legs, bends the knees, and uses full-body posture to reach angles that older robots couldn't even attempt.

03:56There's also a ton of work behind the scenes to make sure this control works out of distribution,

04:01which basically means G1 can handle commands in situations it never saw during training.

04:06You can tell it to stretch further, crouch lower, or rotate more than it was trained to,

04:11and it still finds a way to do it.

04:13One test had it picking up baskets from both sides at floor level,

04:17then walking forward and placing them on a shelf at eye level.

04:21That kind of full-body coordination, from crouching to twisting to reaching to placing,

04:26used to be science fiction.

04:28AMO makes it routine.

04:30In fact, they ran detailed evaluations comparing AMO to older control strategies.

04:35Across the board, YOW pitch, roll, base height tracking, it outperformed everything else.

04:41Even when pushed into untrained territory, AMO showed minimal tracking error.

04:46That means the robot wasn't just guessing, it was adapting with precision.

04:51In trash throwing tasks, it smoothly twisted its torso 90 degrees and nailed the throw.

04:56In another setup, it picked up a paper bag by aligning its torso,

05:01maneuvering its hand through the loop,

05:02and then lifting and placing it without losing grip or posture.

05:06And the whole system worked with both teleoperation

05:09and full autonomy, depending on the task.

05:12What's even more impressive is how AMO builds that behavior.

05:16During training, they use a two-stage learning framework.

05:19A teacher policy gets access to all the ideal data.

05:22Then a student policy learns from that in a more restricted setup,

05:28basically mimicking what the robot would experience in the real world.

05:31This approach lets the final system perform reliably without needing perfect conditions.

05:37So yeah, this isn't your average walking bot.

05:40The G1 with AMO isn't just reacting.

05:42It's planning, adjusting, and executing complex tasks that most robots still fumble.

05:48It stretches, crouches, twists, walks, balances,

05:51all while coordinating 29 joints across its entire body.

05:54And with AMO driving it, you don't get that weird mechanical stiffness you usually see.

06:00You get motion that feels smooth, deliberate, and honestly kind of human.

06:05This is where humanoid robotics is heading.

06:07Not just mobility, but full-body dexterity with the kind of responsiveness

06:11that makes real-world deployment finally possible.

06:14But Unitree didn't stop there.

06:17They've now got a four-legged firefighting robot dog,

06:21a modified version of their B2 robot,

06:23and it's actually made for emergency scenarios.

06:25This firefighting version comes with a water or foam cannon

06:29that can shoot up to 60 meters away at 40 liters per second.

06:34That's intense.

06:34Its joints are upgraded by 170% over the regular model,

06:40so it can climb obstacles up to 15 inches

06:43and handle steep stairs at a 45-degree angle,

06:47perfect for broken buildings and chaotic scenes.

06:51And it's not just muscle.

06:53The B2 firefighting dog is loaded with tech, live video,

06:57LiDAR sensors, and communication gear for sending updates to human teams.

07:01It has a built-in cooling sprinkler system to survive the heat,

07:04plus swappable batteries that won't mess with its waterproofing.

07:08This thing is meant to go where humans can't.

07:11Collapsed buildings, toxic zones, zero visibility areas.

07:15It can map surroundings, locate fires, carry modules,

07:18even support rescue operations.

07:20Now, let's jump from the real world into the corporate world.

07:24Lenovo just entered the humanoid robot race,

07:28and they're not playing around.

07:30At their Tech World 2025 event in Shanghai,

07:33they unveiled Le Xiong No 1,

07:36calling it a silicon employee.

07:39This wasn't just a press statement.

07:41They put it on stage and had it perform a tai chi routine.

07:45Yeah, not just walking or standing.

07:48Actual, slow, balanced martial arts, live.

07:52During the Q&A,

07:53the robot was pulling data from Lenovo's systems in real time

07:57and answering questions like a trained rep.

08:00Under the hood,

08:01this robot runs on three core intelligent frameworks.

08:05It can understand and communicate across devices naturally,

08:09access both public and private knowledge across ecosystems,

08:12and perform advanced tasks with autonomy.

08:16Everything's layered over a secure design,

08:18and it's all running on Lenovo's hybrid architecture,

08:21device, edge, cloud, and network.

08:24This means data collection, processing, and AI model training

08:27all happen seamlessly across platforms.

08:30But Lenovo's not just showing off.

08:32They're planning real-world use cases.

08:34You'll be seeing this robot in elder care

08:36and health care environments soon.

08:39And in August,

08:40they're bringing it to compete

08:41in the Beijing Humanoid Robot Sports Games,

08:45which, yes, is an actual thing now.

08:47The event is going down at Beijing's two biggest Olympic venues,

08:51the Bird's Nest and the Ice Ribbon,

08:54from August 15th to 17th.

08:56This isn't a gimmick, either.

08:58There will be 11 actual human sports

09:00recreated by humanoid robots.

09:02Track and field, football, gymnastics, and more.

09:06Robotics experts and sports professionals

09:08teamed up to create events

09:10that mimic human movement as closely as possible.

09:13And the goal is clear.

09:14Refine mechanical structure

09:16and push motion algorithms further

09:18through real performance under pressure.

09:21This is the second major robot sports event in China this year.

09:25The first was a humanoid half-marathon in April.

09:2820 robots lined up to run over two hours straight,

09:31and one of them, Kyan Kung Ultra,

09:33finished first after nearly two hours and 40 minutes.

09:36That wasn't just about endurance.

09:38That was a massive benchmark for stability,

09:41safety, and real-world readiness in complex environments.

09:44The whole event became a testing ground

09:46for robots from different companies,

09:48validating their ability to operate outside the lab.

09:52Now, while all this is happening in public arenas,

09:54let's zoom back into the research lab.

09:56There's another robot making waves.

09:59This one's called Atom,

10:00developed by a team at P&D Botics in China.

10:03What sets Atom apart is the way it walks.

10:05It doesn't just move.

10:06It walks like a human,

10:08adapting its stride, balance, and pace on the fly

10:12across uneven terrain.

10:14This is thanks to its proprietary reinforcement learning system

10:17combined with imitation learning.

10:19Atom has been in development since mid-2023,

10:22and they've upgraded nearly every part of it since.

10:25The legs and feet are reinforced for durability,

10:28and the actuators are modular,

10:30so it can handle all sorts of unpredictable environments.

10:32It runs on a full-stack system powered by an Intel i7 chip

10:37with real-time controls and full-body motion architecture.

10:40It's got 25 force-controlled QDD actuators with legs

10:45delivering up to 360 newton meters of torque.

10:49The arms have five degrees of freedom,

10:51the waist has three,

10:52and the robot stands 1.6 meters tall,

10:55weighing 60 kilograms.

10:57What's really clever is how it learns.

10:59It's been trained with NVIDIA's Isaac Gym,

11:02using deep reinforcement learning at scale.

11:05Then, they used motion capture

11:07to feed in extremely precise human movements,

11:10adapted the data to Atom's body,

11:12and fine-tuned everything.

11:14It doesn't rely on vision for now.

11:16It's mostly focused on blind locomotion.

11:18But even without vision modules,

11:21it's already adapting dynamically

11:22to whatever you throw at it.

11:24We're talking full-on simulation to reality transition,

11:27something many robots still struggle with.

11:30Meanwhile, we're seeing the broader robotics industry

11:33coming together for massive exhibitions.

11:36The World Robot Conference is also happening

11:38in Beijing this August,

11:40just before the robot sports games.

11:42This year's event will have more than 200 companies

11:44participating and around 100 new product launches.

11:48It's going to be co-hosted by major international groups

11:51like the World Federation of Engineering Organizations

11:54and EU Robotics.

11:56And yes, humanoid robots are the main focus this time,

11:59not just as novelty,

12:00but as scalable solutions for everything

12:04from rescue missions to healthcare.

12:07So if you're looking at all this

12:08and wondering where it's headed,

12:10the answer is pretty clear.

12:13We're moving beyond the demo phase,

12:15not just futuristic,

12:17not just experimental,

12:19useful.

12:20Let me know in the comments

12:21if you had a robot like G1 or Atom,

12:25what's the first thing you'd have it do in your house?

12:27And yeah, maybe we'll cover Phantom's next upgrade

12:30once that $100 million lands.

12:32Thanks for watching,

12:33and I'll catch you in the next one.

🤖 Unitree Gives Robot a BRAIN — Now It’s Acting HUMAN! + More AI & Robotics News 🚀 | AI Revolution

Category

Transcript

Recommended