ChatGPT-Powered "AGI Robot" STUNS The Entire Industry

Matthew Berman
1 Mar 202409:15

Summary

TLDRThe video discusses the latest advancements in humanoid robotics, highlighting Figure, a company creating AI-powered robots that recently raised $675 million from major tech firms. It shows footage of a Figure robot making coffee and fixing errors during the process. Other clips demonstrate affordable open-source robot options costing just $200-$600, using imitation learning and reinforcement learning to train robots. Additional emerging technologies covered include remote-controlled mining robots, unique rolling cage drones combining ground and aerial capabilities, and Amazon's fleet of warehouse robots designed to operate alongside people.

Takeaways

  • 😲A humanoid robot named Figure raised $675M from top tech companies like Amazon, Microsoft, Nvidia, Intel and OpenAI
  • 😎Figure's robot can make coffee - it picks up a capsule, puts it in the machine and presses start
  • 👍The video shows the Figure robot self-correcting after failing to properly insert a coffee pod
  • 💸Robots don't need to be expensive - a $200 robot arm was trained to pick objects using a camera and AI
  • 😀Bipedal robots are learning agile skills like playing football using deep reinforcement learning
  • 🤖A teleoperated robot setup that costs ~$600 can be trained to perform tasks by recording videos
  • 🚘Tesla's Optimus robot is looking more natural and able to walk around smoothly
  • 🕹Video game controllers can enable remote control of robots to perform physical tasks
  • 🤩A spherical drone robot can roll around then take off and fly
  • 📦Amazon is testing digit robots in warehouses designed for humans without rebuilding

Outlines

00:00

😊 Funding and development of humanoid robots by Figure and other companies

The paragraph discusses the recent $675 million funding round raised by Figure, a company developing humanoid robots powered by AI. Major investors include Amazon, Microsoft, Nvidia, and OpenAI. The funding will support Figure's efforts to build advanced humanoid robots. A video shows one of Figure's robots making coffee, demonstrating its capabilities, though the task is relatively simple. The paragraph highlights the impressive self-correcting AI that enables the robot to improve its coffee-making abilities.

05:00

😁 Advances in robotic technology through AI and deep learning

This paragraph showcases several examples of robots learning new skills through reinforcement learning and other AI techniques. This includes a bipedal robot learning to play soccer, a $200 robot arm learning to pick up objects, and Tesla's Optimus humanoid robot improving its walking ability. Additional examples demonstrate how robots can be controlled remotely by humans through video game interfaces and drones inside protective cages. The paragraph emphasizes how AI is enabling rapid advances in robotic capabilities without requiring extensive specialized hardware.

Mindmap

Highlights

Figure raised $675M from Amazon, Microsoft, OpenAI, Nvidia for humanoid robots

Robots taking simple actions like making coffee, but using AI for self-correction

$200 robot arm trained using imitation learning to pick objects

Inexpensive robotics becoming reality using open source hardware/software

Robots learning complex skills like football using deep reinforcement learning

Tesla's Optimus robot looking more natural, could compete in robotics

Humans can remotely control robots using video game interfaces

Creative robot designs like drones inside metal balls

Amazon deploying humanoid warehouse robots designed for human environments

2024 predicted as the year of robot takeover based on increasing examples

AI and reinforcement learning key to robot capabilities without extra hardware

Open source software enabling affordable consumer robotics

Tesla's humanoid Optimus robot rapidly improving, could compete

Remote robot control through video games can solve labor shortages

Innovative robot designs like flying drones within metal balls

Transcripts

play00:14

so what you just saw there is a humanoid

play00:16

robot from a company called figure. and

play00:19

they just raised a ton of money from top

play00:22

companies throughout Silicon Valley

play00:24

Amazon Microsoft open AI Nvidia all of

play00:28

these huge Tech compan companies are

play00:30

putting money into figure let me talk

play00:33

about why so first here's the Inc

play00:35

article Bezos Nvidia joins $675 million

play00:39

funding round for startup making

play00:41

humanoid AI powered robots now figure

play00:43

raised from Microsoft Amazon Nvidia who

play00:46

added nearly $300 million and open aai

play00:50

and Intel both committed money as well

play00:53

now I've been talking a lot about robots

play00:54

lately I've shown off the 1 X Robot in a

play00:57

previous video but now we have another

play00:59

humanoid robot and remember one of my 10

play01:02

predictions for 2024 is that robots are

play01:05

going to take

play01:15

over hey figure one can you make me a

play01:18

cup of

play01:28

coffee

play01:59

thanks

play02:25

now let me break down this video a

play02:27

little bit a lot of people are

play02:28

criticizing it because they basically

play02:30

take the easiest to make coffee machine

play02:31

put it right in front of the robot put

play02:33

the cup in the position to actually get

play02:35

the coffee and there the robot is

play02:37

actually picking up the capsule putting

play02:39

it in the coffee maker pushing it down

play02:42

and then pushing start I still think

play02:44

this is super impressive and not only

play02:46

that the robot itself looks really

play02:48

futuristic I love it but yes of all the

play02:51

demos that we've seen of robots this is

play02:54

one of the more simple ones but they are

play02:57

using Ai and it is self-correcting AI

play03:00

which let's take a look at that so here

play03:02

in these videos what we're seeing is the

play03:03

robot places one of the pods in the

play03:05

curig doesn't get it right and then

play03:08

fixes it itself so that part is pretty

play03:11

darn impressive so they must be doing

play03:13

something right to get so much funding

play03:14

from such top tier investors but that

play03:18

robot looks very expensive it's not

play03:20

actually out yet and when it does launch

play03:22

it probably is going to come with a

play03:23

hefty price tag but robots don't need to

play03:26

come with a hefty price tag if you

play03:28

remember the Aloha robot that entire

play03:31

thing was built for basically I think it

play03:33

was around

play03:34

$30,000 and it used completely

play03:36

open-source hardware and software if you

play03:39

read the GitHub page you can buy all of

play03:41

the hardware they provide you with all

play03:42

of the software and you can put that

play03:44

whole thing together but it can get even

play03:46

cheaper here is Alexander kooch And he

play03:48

says early results from my AI training

play03:50

runs I've trained my $200 robot arm on a

play03:53

simple picking task using imitation

play03:55

learning it has learned to control the

play03:57

robot arm using only camera imag and

play04:00

Joint States now when you combine this

play04:03

with the work that Dr jiman and Nvidia

play04:06

are doing with their robotics

play04:08

Innovations the potential for really

play04:11

inexpensive robotics in everybody's

play04:13

household we can see can become reality

play04:15

sooner rather than later let's take a

play04:17

look at this

play04:28

video

play04:36

and extending on this we have another

play04:39

example of a robot in an office setting

play04:42

and this looks very familiar to the

play04:44

Aloha robot except this is called

play04:46

Universal manipulation interface in the

play04:48

wild robot teaching without in the wild

play04:51

robots what that basically means is

play04:53

anybody can teach the robot how to do

play04:55

something by simply taking a video of it

play04:57

and using that as training data very

play05:00

similar to the work Dr Jim fan is doing

play05:02

with mimic genen let's take a look at

play05:04

this

play05:23

video so check this out this is a

play05:25

bipedal robot learning agile football

play05:27

skills with deep reinforcement learning

play05:29

now that's the key principle here all of

play05:32

the robot advancements that we're seeing

play05:33

how they are becoming so good yet not

play05:36

needing a ton of extra Hardware is

play05:38

because of reinforcement learning let's

play05:40

watch this

play05:58

video

play06:08

now back to Alexander kooch he posted

play06:10

another video just a week ago and he

play06:13

said added more arms the entire setup is

play06:15

now about $600 so he is able to actually

play06:18

teach this robot using teleoperation and

play06:22

again look this looks like a Raspberry

play06:24

Pi all of this looks to be very open-

play06:26

Source

play06:28

Hardware

play06:39

but don't count Tesla out of the game

play06:41

yet Optimus is looking incredible Elon

play06:44

Musk just posted this a week ago look at

play06:46

this robot walking around the

play06:57

lab it is looking more natural every

play07:01

week let's say you don't want to use AI

play07:03

you have a more complex task that humans

play07:06

can do pretty easily well check this out

play07:08

now we can have remote control of a

play07:10

robot using video games and here it says

play07:13

this is how you solve the mining labor

play07:15

crisis millions of gen zers playing

play07:17

video games all day can finally put that

play07:18

into good use give them a joystick and

play07:20

$100,000 base salary can move Rock from

play07:22

the comfort of their home technology is

play07:28

amazing

play07:40

so very very cool not putting any humans

play07:42

at risk and another incredible robot in

play07:45

a form factor that I have never seen

play07:47

before let's take a look at this video

play07:49

it is basically a drone inside of a

play07:53

metal ball cage type thing and it's

play07:56

actually able to roll the ball and then

play07:58

well let's take a

play08:19

look absolutely incredible so the ball

play08:22

is rolling on the ground rolling up the

play08:24

rocks and then all of a sudden it takes

play08:26

off it's so cool and last of course

play08:30

Amazon the biggest Warehouse owner in

play08:33

the world has been deploying humanoid

play08:35

robots in their warehouses let's take a

play08:38

look at what those look like Amazon has

play08:40

already started testing digit in its

play08:42

Seattle robotics research and

play08:43

development facility potentially adding

play08:45

to the company's Fleet of more than

play08:47

750,000 robots the fact that it can

play08:49

enter places that were designed for

play08:51

humans um and solve those challenges

play08:54

that exist without having to rebuild uh

play08:56

the environment for Automation and

play08:59

doesn't quit and it doesn't quit so 2024

play09:02

is the year of robots I am more

play09:03

confident in my prediction than ever I'm

play09:06

going to keep collecting all of these

play09:07

amazing examples of robots so I can show

play09:09

you if you liked this video please

play09:11

consider giving a like And subscribe and

play09:13

I'll see you in the next one