I really enjoyed this conversation with Russ. Here's the outline: 0:00 - Introduction 4:29 - Passive dynamic walking 9:40 - Animal movement 13:34 - Control vs Dynamics 15:49 - Bipedal walking 20:56 - Running barefoot 33:01 - Think rigorously with machine learning 44:05 - DARPA Robotics Challenge 1:07:14 - When will a robot become UFC champion 1:18:32 - Black Mirror Robot Dog 1:34:01 - Robot control 1:47:00 - Simulating robots 2:00:33 - Home robotics 2:03:40 - Soft robotics 2:07:25 - Underactuated robotics 2:20:42 - Touch 2:28:55 - Book recommendations 2:40:08 - Advice to young people 2:44:20 - Meaning of life
Few years back, I use to be that kid knowing nothing about what you guys are talking listening for pure enjoyment,now i am listening to the conversation again as a graduate student doing soft robotics having read Russ’s work, it’s just beautiful. Thanks for being a huge inspiration.I am so grateful.
Timing couldn't have been better, I was stuck in something that's related to this field of study and this podcast gave me the motivation to push forward
This is one of my all time favorite episodes. I have rewatched this video over 4-5 times now. I love it, great teacher talking about an interesting topic
great point Lex about editing out the non-pedalogical bits and ramblings in lectures to get the best most clear and concise pedagogical experience for students
dakrontu I thank Apple for my phone and GE for my dishwasher. At what point will I feel compelled to say “Thank you” to the machine itself? It will be awkward at first.
Hugh Higley I can’t wait till we see the first generation of AGI, like the bridge between Narrow AI, and General Purpose AI...... Think we can get there by 2030? That is if we don’t have an AI winter next year. There’s only been one small breakthrough this year in AI so far that has allowed unsupervised learning at the speed of light. But I still think we need a major breakthrough in hardware, I think the software is out there for AGI.
I can listen to this podcast hours and hours with enormous excitement on my face.
4 ปีที่แล้ว +19
I have a feeling that if we want durable dishwasher-filling robots, teaching them to say "I'm sorry" and giving it an appropriate facial expression would improve their durability significantly in the real world.
33:01 think rigorously with ml 40:45 simplifying assumptions 1:36:46 contact issues. Rigid body assumptions 1:56:00 rare event simulation 2:17:35 underactuated is the way of life, man 2:18:35 robotic manipulation course 2:40:08 advice to young people. Learn math, build things
RE: segmenting the topics of your podcasts - I often find this to be a very useful feature and I thank you much for the time you spend to do it. Two reasons: (1) sometimes I want to come back and give another listen to try to better understand, or just because I particularly enjoyed something, and thanks to you it’s already indexed (relevant to ‘how to read’ book convo), and (2) sometimes it’s the opposite and I’m not that into a topic and I know exactly how far ahead to skip to the next one.
Amazing podcast, what you mention at the end about the impact it can have on people to guide them through engineering and curiosity is entirely true. THANKS.
@Lex Fridman. Please do more interviews like this. I'm very tired of the culture wars and I just want to listen to serious roboticists talk about what they love.
Thanks, Lex! Since I found the podcast last year it's been my favorite thing on the internet. No joke! Keep up the great questions, curiosity, and positivity!!
fell asleep on the couch with this episode waking me up at 4 AM. Now stoned (in US and not Russia), listening during a sunrise involving Venus, inspired to start today. I salute you Lex.
@Lex Fridman A very entertaining lecture and many thought provoking ideas on the intersection and differentiation between the human body and current robot body design approaches. Love the idea of having squishy robots that morph to their environment rather then inefficiently & ineffectively choosing route based on pure avoidance principles.
At 1:58:50ish - You could over-ride the loops that control things, and just force bad outcomes.... like when the foot was in the wrong mode getting out of the car... don't try to handle every failure, just try to minimize harm over those failures.
A wonderful human conversation about robotics. Thank you. My amateur comments: 1. A great Japanese SF short story about a robot bar girl: “Bokko-Chan” by Shinichi Hoshi. 2. A great Chinese SF short story about the loneliness of a future world full of robots. “Goodnight Melancholy” by Xia Jia 3. A remembered ending to a story I wrote: An aging man sits in front of the fireplace with his lifelong companion, a talking robot dog, and the man says, “You robots have it easy, because you never die.” The robot dog says, “It’s true we never die, but we do not have it easy, because you see-we never forget.” 4. When we replace human companions with robots or VR worlds, we might lose something. In the Dirt World, we humans always have some, even if slight, risk, that other humans in our presence might kill us. This gives human interactions a stronger flavor of life. Thank you. William L. Ramseyer
The problems of underactuated robotics and contact physics are specifically the problems that an AGI will need to solve before it's truly intelligent in our real world. I really wish AGI and robotics were more of a priority for our civilization. The current investment is so small in the scope of our actual resources. Seeing Russ's talks are very inspirational in real life and here on your podcast, they truly show how far we have to go.
im trying my best to understand barefoot running and have been slowly doing more for the last 3 years, ive taken advice from many it seems but between Russ Tedrake and Tony Riddle in the uk i think I'm understanding things much better, really fucking amazing podcast
Such a bummer Anki closed. They were doing great things toward getting a robot in every home. We used Cozmo all the time to help get kids interested in learning Robotics and Coding.
I'm a little surprised that Tedrake didn't discuss the solution to the problem at about 1:40:00 of including a model for mechanical deformation of the material. That's the way it was introduced in statics and mechanics of materials. Maybe that kind of modeling is intractable, but weird to not mention it at all.
Evolutionarily speaking, homo sapiens and even early hominids such as homo erectus started walking upright due to change in environmental conditions, as well as behaviourally advantageous. When the Sahara dried up and less trees/territories where available for the large pop. size, hominids started walking over open terrain, were it helped seeing further, over long grass from a higher altitude viewpoint. As Lex said, having freer hands for weapons and social activities (carrying food).
Nah, the early ancestors started walking by wading in the streams in the areas around their homelands among the trees. The water helped them gain an upright position with buoyancy, and protected them from cat predators. Think that over.
Plottwist, he's actually one of the avatar people (Na'vi) placed into a human body, just like they did in the movie but the other way around. And as a special customer service they kept his right hand in his pre-transmit configuration.
Ugh- 1:54:50 - Nerf the world.... use unbreakable plates, and add detectors that can tell you if you've pushed them too hard, etc. We nerf the world for our kids, we can nerf it for robotics to learn as well. As your parents about this very question.... they will have a lot of wisdom to share.
"We" are shaping the new sub-set of values around AI/Robots, I am concerned about the intellectual emphasis on "feeling" re the robot and the emotions it could invoke on us the "serviced" in that moment. However, our world include our "left behind by automation" neighbour who has an equal vote thankfully. Automation/robots are just tools. Not so, only if you get your values re other humans muddled-up. Well done and thank you for an absorbing channel.
Yes. And for people looking for advice on starting barefoot running: 1) Make sure you've had a Tetanus shot at least in the last ten years. 2) Start by walking barefoot, anywhere you can, but certainly along the path you plan to run. It can take awhile for your feet to get used anything rougher or harder than grass or dirt. If you'll be on sidewalks or roads, it might be a couple/few weeks for your soles to toughen up.
Seems to me that nature combines control and physics as part of its overall solutions to problems. For example, insects bump into things, which partly solves the problem of how to avoid obstacles, ie don't bother, or don't try hard, it's not worth it. So the physical structure of an insect, the nature of the world within which it functions, and the control provided by its simple brain and sensors, are all parts of the solution. The environment limits what is actually needed. The physical structure limits the need for too much computational complexity. The simple brain uses simple clues to recognise situations (eg finding a mate) and it does not matter if it gets things wrong some of the time as long as doing so is not fatal. So there is an optimisation that has evolved that encompasses all of environment, physical structure, sensors, and the limited control capacity of the brain. Sophisticated brains consume more energy, so a solution that minimises brain energy consumption is important. Maybe this is also why humans prefer to avoid thinking by jumping on bandwagons, ie rely on other people having thought something out already.
ROFL, I've been aging in place for the last 20 years. What have you all been doing? Hopefully living your lives, happily and productively. Myself, I'm an Underactuated Humanist.
Have you noticed how Lex internalizes the rhythm and cadence of the guests? It even seems like he's a completely different person. It's fascinating. I wonder if it has something to do with this: th-cam.com/video/FDhlOovaGrI/w-d-xo.html (TED Talk - Your brain on communication - Uri Hasson).
AI Superpowers: China, Silicon Valley, and the New World Order ~ Kai-Fu Lee Sapiens: A Brief History of Humankind ~ Yuval Noah Harari The Black Swan: The Impact of the Highly Improbable (Second Edition) ~ Nassim Nicholas Taleb How to Read a Book ~ Mortimer J Adler I, Asimov: A Memoir - Isaac Asimov by Isaac Asimov
I really enjoyed this conversation with Russ. Here's the outline:
0:00 - Introduction
4:29 - Passive dynamic walking
9:40 - Animal movement
13:34 - Control vs Dynamics
15:49 - Bipedal walking
20:56 - Running barefoot
33:01 - Think rigorously with machine learning
44:05 - DARPA Robotics Challenge
1:07:14 - When will a robot become UFC champion
1:18:32 - Black Mirror Robot Dog
1:34:01 - Robot control
1:47:00 - Simulating robots
2:00:33 - Home robotics
2:03:40 - Soft robotics
2:07:25 - Underactuated robotics
2:20:42 - Touch
2:28:55 - Book recommendations
2:40:08 - Advice to young people
2:44:20 - Meaning of life
🗽✨🍃
Nice mug
[[[][u[upset[[[u hippo 👆 u[[yiuuuuu up u
Great channel, great guests, great interviews with great minds
Lex, please read manga Phoenix by Osamu Tezuka. It will blow your mind.
Few years back, I use to be that kid knowing nothing about what you guys are talking listening for pure enjoyment,now i am listening to the conversation again as a graduate student doing soft robotics having read Russ’s work, it’s just beautiful. Thanks for being a huge inspiration.I am so grateful.
Thanks for interviewing Russ. He is one of my favorite professors at MIT. Amazing person and great teacher.
Timing couldn't have been better, I was stuck in something that's related to this field of study and this podcast gave me the motivation to push forward
Cool :)
What kind of topic were you working on?
@@msergio0293 Trajectory optimization for bipeds to generate different types of gaits 🙃
@@hubble-02 My old sprinting coach approves of this line of research :D
You are not alone :D
Likewise!
This is one of my all time favorite episodes. I have rewatched this video over 4-5 times now. I love it, great teacher talking about an interesting topic
great point Lex about editing out the non-pedalogical bits and ramblings in lectures to get the best most clear and concise pedagogical experience for students
Thanks Lex for bringing engineers and scientists to the spotlight. We need more folks like you in the media ecosystem.
When a robot sees my need and offers a steady hand I will be grateful.
To the robot? Or to the developers?
dakrontu
I thank Apple for my phone and GE for my dishwasher.
At what point will I feel compelled to say “Thank you” to the machine itself? It will be awkward at first.
Hugh Higley I can’t wait till we see the first generation of AGI, like the bridge between Narrow AI, and General Purpose AI...... Think we can get there by 2030? That is if we don’t have an AI winter next year. There’s only been one small breakthrough this year in AI so far that has allowed unsupervised learning at the speed of light. But I still think we need a major breakthrough in hardware, I think the software is out there for AGI.
@@hughhigley6188 That'll be the point when you have been fooled.
I can listen to this podcast hours and hours with enormous excitement on my face.
I have a feeling that if we want durable dishwasher-filling robots, teaching them to say "I'm sorry" and giving it an appropriate facial expression would improve their durability significantly in the real world.
33:01 think rigorously with ml
40:45 simplifying assumptions
1:36:46 contact issues. Rigid body assumptions
1:56:00 rare event simulation
2:17:35 underactuated is the way of life, man
2:18:35 robotic manipulation course
2:40:08 advice to young people. Learn math, build things
I love listening to Lex and Russ while playing misc missions in Ghost of Tsushima! I feel like I'm learning while doing something fun.
Give some insight on the game
It’s an RPG and you play as a samurai during the mongol invasion. Easy to play, very fun.
I listen while playing Monster Hunter World it's pretty comfy to listen to while grinding the game on mute.
RE: segmenting the topics of your podcasts - I often find this to be a very useful feature and I thank you much for the time you spend to do it. Two reasons: (1) sometimes I want to come back and give another listen to try to better understand, or just because I particularly enjoyed something, and thanks to you it’s already indexed (relevant to ‘how to read’ book convo), and (2) sometimes it’s the opposite and I’m not that into a topic and I know exactly how far ahead to skip to the next one.
Amazing podcast, what you mention at the end about the impact it can have on people to guide them through engineering and curiosity is entirely true. THANKS.
Thank you. These podcast help me understand what my son, a PhD in Robotics, does and thinks about.
@Lex Fridman. Please do more interviews like this. I'm very tired of the culture wars and I just want to listen to serious roboticists talk about what they love.
I love the moment when you ask them about the meaning of life. Also, grats for the 500k subscribers!
Thanks so much for doing these Lex. Great conversation
Thanks, Lex! Since I found the podcast last year it's been my favorite thing on the internet. No joke! Keep up the great questions, curiosity, and positivity!!
3 sponsors?? Wow lex getting big
Awesome interview LEX. Thanks.
1:27:35 "The smartphone is already the robot we should be afraid of"
I enjoyed this podcast a full lengthy video of under standing thank you both and god bless you for your wisdom
Ive been waiting so long for this, this is amazing Lex.
fell asleep on the couch with this episode waking me up at 4 AM. Now stoned (in US and not Russia), listening during a sunrise involving Venus, inspired to start today. I salute you Lex.
I got to say, I love your podcasts Lex!
Fantastic interview Lex, thanks.
@Lex Fridman A very entertaining lecture and many thought provoking ideas on the intersection and differentiation between the human body and current robot body design approaches. Love the idea of having squishy robots that morph to their environment rather then inefficiently & ineffectively choosing route based on pure avoidance principles.
Another great podcast from you. Thanks for your efforts.
At 1:58:50ish - You could over-ride the loops that control things, and just force bad outcomes.... like when the foot was in the wrong mode getting out of the car... don't try to handle every failure, just try to minimize harm over those failures.
Love your content Lex. Have you considered getting Daniel Dines from UiPath on for an interview? I think that'd be very interesting
Congrats on 500k subscribers Lex! Love the guests that you bring on
A wonderful human conversation about robotics. Thank you. My amateur comments:
1. A great Japanese SF short story about a robot bar girl: “Bokko-Chan” by Shinichi Hoshi.
2. A great Chinese SF short story about the loneliness of a future world full of robots. “Goodnight Melancholy” by Xia Jia
3. A remembered ending to a story I wrote: An aging man sits in front of the fireplace with his lifelong companion, a talking robot dog, and the man says, “You robots have it easy, because you never die.” The robot dog says, “It’s true we never die, but we do not have it easy, because you see-we never forget.”
4. When we replace human companions with robots or VR worlds, we might lose something. In the Dirt World, we humans always have some, even if slight, risk, that other humans in our presence might kill us. This gives human interactions a stronger flavor of life.
Thank you. William L. Ramseyer
Thank you, great episode
THIS IS AMAZING! Thank you Lex.
The problems of underactuated robotics and contact physics are specifically the problems that an AGI will need to solve before it's truly intelligent in our real world. I really wish AGI and robotics were more of a priority for our civilization. The current investment is so small in the scope of our actual resources. Seeing Russ's talks are very inspirational in real life and here on your podcast, they truly show how far we have to go.
Have you seen Nassim Taleb’s work (Incerto) ? I think you would like him ! He does work on uncertainty !
im trying my best to understand barefoot running and have been slowly doing more for the last 3 years, ive taken advice from many it seems but between Russ Tedrake and Tony Riddle
in the uk i think I'm understanding things much better, really fucking amazing podcast
I wish I was as honest and as real as any of these guys
Great conversation!
2:39:08 - lol, its literally me. I'm listening this podcast in Russia right now)
I was looking 4 this :))!
Stunning interview
Amazing content! Your guests are remarkable. Keep up the great work!
Such a bummer Anki closed. They were doing great things toward getting a robot in every home. We used Cozmo all the time to help get kids interested in learning Robotics and Coding.
Great Talk. Environments and environmental niches do co-evolve with the species they support. See Stuart Kauffman and complexity theory generally.
I'm a little surprised that Tedrake didn't discuss the solution to the problem at about 1:40:00 of including a model for mechanical deformation of the material. That's the way it was introduced in statics and mechanics of materials. Maybe that kind of modeling is intractable, but weird to not mention it at all.
‘Instead of taking the forces of the world and replacing them, we should let the forces push us around and go with the flow’ 🧘
2:39:28 Lex saluted me personally >
Half a million subs. Congrats!
Best coffee mug 2020
This podcast and The Nine Club are up there with my favs. Would be cool to see robots skateboarding and AI discussing politics.
Evolutionarily speaking, homo sapiens and even early hominids such as homo erectus started walking upright due to change in environmental conditions, as well as behaviourally advantageous. When the Sahara dried up and less trees/territories where available for the large pop. size, hominids started walking over open terrain, were it helped seeing further, over long grass from a higher altitude viewpoint. As Lex said, having freer hands for weapons and social activities (carrying food).
Nah, the early ancestors started walking by wading in the streams in the areas around their homelands among the trees. The water helped them gain an upright position with buoyancy, and protected them from cat predators. Think that over.
When do we decide that we know and control deep learning sufficiently, before or after it designs our bridges?
Thanks for the content. A podcast with Rachel Thomas would be nice too.
I even listen to the ads, just out of respect.
i caught myself doing the same. :D
Absolutely.
❤️
Thanks Lex 🤖
Perfect timing! Thank you.
Plottwist, he's actually one of the avatar people (Na'vi) placed into a human body, just like they did in the movie but the other way around. And as a special customer service they kept his right hand in his pre-transmit configuration.
Sweet mug Lex!
fascinating ! Thank you !
I'm in Illinois but I still appreciate the solute.
take a look at Raven Protocol it is a decentralized and distributed deep-learning training protocol and very undervalued.
Ugh- 1:54:50 - Nerf the world.... use unbreakable plates, and add detectors that can tell you if you've pushed them too hard, etc.
We nerf the world for our kids, we can nerf it for robotics to learn as well.
As your parents about this very question.... they will have a lot of wisdom to share.
Beefcake 3blue1Brown in the house 😭🙌🏾
Lex : Pain.
Lex's mind : *Don't smile, don't smile don't smile*
Lex : ( ͡° ͜ʖ ͡°)
Lex's mind : *Fuck*
Any chance that ROS was used in the Atlas development? Oh and for the UFC robot is chainsaw allowed?
Art is the skillfully representation of reality, so not photography but now robotic kinematics... #truth in opinion
"We" are shaping the new sub-set of values around AI/Robots, I am concerned about the intellectual emphasis on "feeling" re the robot and the emotions it could invoke on us the "serviced" in that moment. However, our world include our "left behind by automation" neighbour who has an equal vote thankfully. Automation/robots are just tools. Not so, only if you get your values re other humans muddled-up. Well done and thank you for an absorbing channel.
The imagined question “Had a few, have we?” made me laugh when the robot fell out if the car.
Edit. Alright I’ll try to shut up now
Fleet learning in operation:
- My owner kicked me...
- My owner kicked me...
- My owner kicked me...
- "Distribute solution: Remove owners legs"
Since you ask about js that much, could we have Ryan Dahl at your show ? 👏
I can't take my eyes off that awesome steampunk skull cup.
That's one funky looking cup lex
Where is he from?? That R sound is so odd I can’t tell which accent it’s from
Getting in and out of a car can be difficult? Try doing it when you’re old 😂!
"UK Version is better.." SHOTS FIRED!!!
Aditya Prakash but........clearly it is!
There're already plenty of robot that would win UFC. Every Robot Wars contestant for one.
1:07 Excited for a robot UFC Champion!!!
Barefoot ?
Barefoot leads to no fallen arches, no in-grown toenails, no varicose veins ..... among other stunning benefits.
Yes. And for people looking for advice on starting barefoot running: 1) Make sure you've had a Tetanus shot at least in the last ten years. 2) Start by walking barefoot, anywhere you can, but certainly along the path you plan to run. It can take awhile for your feet to get used anything rougher or harder than grass or dirt. If you'll be on sidewalks or roads, it might be a couple/few weeks for your soles to toughen up.
ขอบคุณครับ
Can anybody point me to some source on "control theory [...] active experiment design" (1:55:07)?
12:20 here is who gets credit for noticing the fish thing: en.wikipedia.org/wiki/Viktor_Schauberger
At 01:00:00 it seems like deep learning to me, only manual mode lol
Seems to me that nature combines control and physics as part of its overall solutions to problems. For example, insects bump into things, which partly solves the problem of how to avoid obstacles, ie don't bother, or don't try hard, it's not worth it. So the physical structure of an insect, the nature of the world within which it functions, and the control provided by its simple brain and sensors, are all parts of the solution. The environment limits what is actually needed. The physical structure limits the need for too much computational complexity. The simple brain uses simple clues to recognise situations (eg finding a mate) and it does not matter if it gets things wrong some of the time as long as doing so is not fatal. So there is an optimisation that has evolved that encompasses all of environment, physical structure, sensors, and the limited control capacity of the brain. Sophisticated brains consume more energy, so a solution that minimises brain energy consumption is important. Maybe this is also why humans prefer to avoid thinking by jumping on bandwagons, ie rely on other people having thought something out already.
Anybody else wonder whether Lex preordered Cyberpunk 2077.
love it
My best friend tripped and spilled vodka and orange soda all over my AI you can love an AI but it's good to love people more
ROFL, I've been aging in place for the last 20 years.
What have you all been doing? Hopefully living your lives, happily and productively.
Myself, I'm an Underactuated Humanist.
2:39:00 - i bet a surprising amount are both those people
nice cup
Go with the Flow. There really is no choice, based on personal experience.
Have you noticed how Lex internalizes the rhythm and cadence of the guests? It even seems like he's a completely different person. It's fascinating. I wonder if it has something to do with this:
th-cam.com/video/FDhlOovaGrI/w-d-xo.html (TED Talk - Your brain on communication - Uri Hasson).
Like before watching..♡
how many people actually buy magic spoon?
I am the kid from India watching this and smoking some weed.
hey kiddo, your brain is your best friend. don't mess with it...
"Is that art or?.. science?"
Phi= Unity
This guy is an actual NPC, in a Russian reversal kinda way, where in, Soviet Russia NPC is You
AI Superpowers: China, Silicon Valley, and the New World Order ~ Kai-Fu Lee
Sapiens: A Brief History of Humankind ~ Yuval Noah Harari
The Black Swan: The Impact of the Highly Improbable (Second Edition) ~ Nassim Nicholas Taleb
How to Read a Book ~ Mortimer J Adler
I, Asimov: A Memoir - Isaac Asimov by Isaac Asimov
Thanks
So, who is the kid from russia that was watching this while smoking a joint and enjoying him/herself ?