Tesla’s Robot Revolution
Humanoid robots have been in popular culture from the very beginning, and while robotics have come a long way, we still don’t have the humanoid, walking, general purpose robots of our sci-fi imaginations. But some companies, including Tesla, claim to be on the verge of finally making it a reality. But how likely is it, really?
TRANSCRIPT:
To kick off this video, let’s play a little game, I’ll play a video from a random 80’s movie and you tell me what movie it was. Ready? Here we go…To kick off this video, let’s play a little game, I’ll play a video from a random 80’s movie and you tell me what movie it was. Ready? Here we go…
Now if you answered Rocky IV, you are right! Also…What?
I saw Rocky IV like a million times when I was a kid and I remember being terrified of Ivan Drago, I remember Apollo going down and “If he dies… he dies.” I remember training montages… I watched it again recently – I did not remember the random robot.
That’s how ubiquitous robots were back then in pop culture, they were everywhere, it literally didn’t even stand out to me, there was nothing weird to me about a robot… in Rocky IV.
By the way the robot’s name is SICO and it was a real robot you could buy and it could talk to you and you could program it to do things – Stallone put it in the movie because he bought one and found that it really helped with his autistic son.
But seriously robots were everywhere in the 80s, there was Twiggy on Buck Rogers, The Terminator, we saw Jinx put Max in space, Johnny 5, there was this sitcom, Small Wonder, about a family with a robot daughter, yes, it’s a real thing.
One of the biggest toys of the day was Teddy Ruxpin, this animatronic teddy bear that talked to you.
By all accounts, we were about to enter an age of household robots. And here we are now, 40 years later… I don’t know anybody with a robot daughter.
What happened? Are robots like nuclear fusion, always 20 years away?
I mean, yeah, we have robot vacuum cleaners and stuff but we still don’t have those general-use humanoid robots we always imagined.
But technology has come a long way. And AI is forging new ground all the time. And many companies – including Tesla – are betting big on a robot revolution. So how close actually are we?
Robots In Pop Culture
Since robots first appeared in popular culture, we’ve imagined the possibilities and problems they present. We’ve represented them as helpers, like Rosey, counselors like C-3PO, and comedians, like Kryten.
A robot who was always… Fully Charged. (canned laughter)
And while robots have always come in all shapes and sizes, whatever form factor best suits their purpose, we ultimately tend to gravitate toward humanoid robots – robots that look and behave more like ourselves.
This is a double-edged sword though. Because our behavior is… not always great.
The dark side of human-robot relations was explored from the very beginning of robots in fiction. The term “robot” was actually coined for the 1921 play R. U. R., or Rossum’s Universal Robots. In
the play, robots rebel against humanity.
So yeah… right out the gate.
This started a tradition that was carried on by Metropolis, The Day the Earth Stood Still, Blade Runner, The Terminator franchise, the list goes on.
Why Humanoids?
And still, robotics companies and engineers continue working tirelessly on the goal of developing humanoid robots, knowing full well that they’ll probably rebel against them someday. Robots are kinda like teenagers.
But this is actually extremely hard on a hundred different levels and frankly, I’ve always wondered why. Like, what’s the point of humanoid robots?
In a way it feels like the height of arrogance to me, like who said this form is the ultimate in peak performance?
Wouldn’t more legs be more stable? Couldn’t you get more work done with more arms? Why only eyes up front, if you could literally have eyes in the back of your head? Or the top of your head?
And it kind-of smells of playing God just a little bit. I mean, they say God created man in his own image… now look what we’re doing.
So that’s what I wanted to explore in the making of this video. What are the advantages of these kinds of robots, what challenges need to be overcome… And how close are we really?
And while looking into this it seems researchers are focusing on two areas – one from the neck down and one from the neck up. That’ll make sense in a minute.
Humanoids Are Versatile
So, like I said, the human form isn’t the best at everything. We aren’t the strongest, we’re not the fastest, we don’t climb or swim the best. But where we do shine is in our versatility.
Humans are generalists. There’s not a lot of things other animals can do that we absolutely can’t.
Service Robots
So the advantage of humanoid robots – and the goal of humanoid robots – is to share that versatility. They’re not made to do one specific thing; they’re made to do a little bit of everything, just like we do.
And that means interacting and existing in a world built for humans.
For example, there are already robots performing customer service functions around the world.
If you live in a big city, especially in Japan and China, you’ve probably seen receptionist robots, waitress robots, or photography robots at special events. https://www.servicerobots.com/
Boston Dynamics
Probably the biggest name in humanoid robots today is Boston Dynamics. I’m sure you’ve all seen their videos.
They kinda hit the scene in the early 2000s, when the company showcased a series of robots for military use.
These were redesigned in various ways until 2013 when they showed off their robot Atlas. It was still a work in progress at the time, one of the program managers said it had the ability to walk at the level of a one-year-old.
But Atlas has come a long way since then.
It can now walk over difficult terrains, balance on one leg, dance, even do parkour. Atlas team lead Scott Kuindersma (keen-DERS-ma) has said that Atlas captures “our vision of a go-anywhere, do-anything robot”.
The videos are impressive… But if you watch the ones from behind the scenes, you’ll see Atlas isn’t perfect. It actually falls down a lot.https://www.youtube.com/watch?v=EezdinoG4mk
I’m not saying this to mock Atlas, it’s a world-class robot built by some of the best engineers in the world, but it’s important to remember that Atlas is a research platform, it’s nowhere near ready to be a consumer product.
It’s also not a soldier.
You might have seen a video going around of Atlas flipping out and attacking its handlers with a gun well don’t worry, it’s not real, it’s a video by the effects studio Corridor Digital.
One thing that might give that away is the videos are watermarked Bosstown Dynamics instead of Boston Dynamics? Also… it’s not Atlas. So there’s that.
But there is this Russian robot from 2017… Just shooting away. https://www.youtube.com/watch?v=HTPIED6jUdU[END TANGENT]
Humanoids Are Personable
So Atlas is the neck down approach, it’s figuring out how to move and operate its body in human spaces. It doesn’t even really have a head.
The other side of humanoid robotics are the neck up side, and these are robots meant to interact with humans in a way that feels natural.
Hanson Robotics’ Sophia is a prime example
According to her creators, Sophia is a “hybrid human-AI intelligence,” where humans craft and guide her conversations.
But not always. Sometimes she goes into unexpected places, like the time in 2016 when founder David Hanson asked Sophia if she wanted to destroy humans.
That’s creepy, but keep in mind she doesn’t have legs so…
The point of these robots, outside of, you know, destroying our species, is to imitate human looks and speech to put humans at ease around them.
Among the most famous are Nadine from Nanyang Technological University in Singapore, Junco Chihira from Toshiba, and Jia Jia from China.
All three are “female” androids that work as receptionists and greeters at conventions and technology demonstrations, and there are many others working tourist and hospitality jobs.
The Uncanny Valley
A slightly different kind of robot is Ameca, built by the company Engineered Arts.
The most amazing thing about Ameca is how expressive its face is. This goes way beyond the Hall of Presidents animatronic stiffness you see in most of these designs.
By the way she’s made by the same company that did an eerie likeness of Tom Scott if you follow him.
One more thing about Ameca that I think helps is it’s designed to look like a robot. It has kinda gray skin and an open neck, no hair, it’s not meant to look real.
So you get something expressive enough to connect to but still avoiding uncanny valley.
Just in case you don’t know what uncanny valley is…
Uncanny valley is this phenomenon where the closer an animated face gets to real life, the more we find it unsettling and creepy. And it’s actually a really interesting phenomenon when you think about it.
Because we are actually more able to connect emotionally with animated characters that are vague representations of us at best, than we are with nearly photo-realistic depictions of people.
I think it says a lot about just how wired we are for faces.
We see faces everywhere, in clouds, on rock formations, on satellite pictures of Mars.
But we’re also so attuned to the subtle nuances of facial movements and behavior that when you see a very realistic person whose nuances just… don’t quite line up right. There’s this alarm that goes off in our head that says, “something’s wrong here.”
It’s like our own internal lie detector.
Even as deep fakes have pushed the envelope of CGI, that problem of uncanny valley is still juuuust beyond reach.
And this problem is especially difficult when you’re talking about physical representations, which is why the team behind Ameca just didn’t even try. Paradoxically, making it less human makes humans more comfortable with it.
And this is backed by studies that have shown that consumers prefer robots with some human features, but we don’t trust robots that try too hard.https://www.bu.edu/bhr/2021/10/04/the-rise-of-service-robots-in-the-hospitality-industry-some-actionable-insights/
Cultural Expectations
Studies have also pointed out how differences in culture affect how we view robots.
Like Chinese consumers expect robots to be have more autonomy, whereas Americans expect them to follow simple rules.
The reasons have to do with religious and philosophical beliefs; any grad students looking to write a thesis, you’re welcome…
Point is there’s a lot more to consider than just the technical side when it comes to making a consumer robot.
Which may be why Tesla has decided to just not mess with the face at all.
Yeah, Tesla’s announcement of their Teslabot project last August was one of the biggest WTF moments in the history of a company that’s packed to the gills with WTF moments.
But here I am still talking about it 10 months later so mission accomplished.
Elon claims that the robot will be able to help out with boring, repetitive chores, will be able to lift 45 pounds, walk at 5 miles per hour, and will be easy to overpower should you – ahem – need to.
It also claims on its diagram here that it will have “human-level” hands. Which… I mean the human hand is one of the most dextrous appendages in the world so that sounds like a tall order. But we’ll see…
The idea that Tesla, a car company, would branch out into humanoid robots sounds insane… Except… Honda did the same thing 20 years ago.
And I mean it does kinda make sense, they’ve developed this computer vision platform that can navigate in 3D space, and I imagine mapping out a confined space like a home or office environment might not be too much to keep track of.
Combine that with state of the art robotics engineering and you might have… something.
I’ve said from the beginning that the Teslabot (which they’re calling Optimus, by the way) was really just a moonshot project designed to get top engineers and AI people to work for Tesla.
This is something a lot of companies do, create a big, ridiculous, most likely impossible project and use that to hire talented people. And if you happen to break new ground along the way that can be used to improve your current products, well more power to you.
Personally I think Tesla is a long, long way from having a robot that you can buy and have in your home, helping you out with random stuff. I think like full self driving, it’s a much harder problem than they realize.
But that’s not really the point, the point is hiring talented people. But that’s just my joepinion.
I guess we may find out soon because last August he gave a very unexpected timeline for when we’d see one of these.
I mean… Do I have to say anything?
Where Are We Now?
Which begs the question, where exactly are we with robotics? How feasible is the Teslabot given the highest technology we have today?
Well luckily one of my Patrons, Cole Parker, works in robotics and was gracious enough to help me get my head around it.
Basically there are 4 areas that need to be mastered for this to be a reality, balance, articulation, vision, and battery life
Balance
Balance is something that most of us take for granted and never think about until it’s gone. It’s just that well handled by our brain’s subconsious operating system.
Remember the falls Atlas took between takes? Scott Kuindersma summed up the challenge this way:
“If you or I were to vault over a barrier, we would take advantage of certain properties of our bodies that would not translate to the robot.”
For example, Atlas doesn’t have a spine or shoulder blades. Its torso is heavy, its arm joints relatively weak.
Basically, Boston Dynamics is teaching Atlas to make humanlike movements, without giving it human parts. That’s like trying to teach a dog to yodel
The complexity of the task is illustrated by another robot called Little HERMES.
This was demonstrated in 2019 by researchers at MIT, and it has sensors that trigger a vest, worn by the operator, to move in time with the robot
When the robot is knocked off-balance, the operator feels it and recovers. Little HERMES then imitates the operator to stay upright https://www.youtube.com/watch?v=IquGO78ZIf0
So Little HERMES works great… so long as there’s a human brain behind it.
There’s a ways to go, obviously before we have robots that we won’t be spending a lot of time picking up off the floor. But it’s coming along.
Articulation
Another major challenge is articulation, which is why I was saying I’m doubtful about the “human-like ability” of the Teslabot hand.
The human hand is served by three nerves, each with its own function, and all of which participate in control and sensation.
The sensation part is the kicker.
Your nerves connect to something like 17,000 nerve endings in each hand. One fingertip has upwards of 3000 touch receptors.
This isn’t just so you know that a stove is hot, this has massive implications for dexterity and the ability to apply just the right amount of pressure.
There’s a LOT that goes on in the background of our brains to make sure we apply the right pressure to things. Like obviously you don’t apply the same pressure to a 50 pound weight as you do to an egg.
A robot has to see an object, understand what it is, understand how fragile it is and adjust accordingly. Also picking up a 50 pound box by the sides is going to require a lot more pressure than a 1 pound box.
This is something that occurs intuitively for us, but all of this has to be carefully programmed into a robot.https://www.nature.com/articles/d41586-018-05093-1
Vision
Speaking of seeing objects takes us to the next challenge, which is robot vision.
The challenges in this field could be enough to fill a whole video, but to get a taste, consider what happens when a robot tries to navigate a room.
It’s not enough for the robot to see the objects present, it has to know if they’re moving, if they might move, and what paths they will take.
And by objects I don’t just mean furniture, robots have to share these spaces with people and pets.
Pets who flip out at the sight of a vacuum cleaner so I’m sure they’ll be super chill around a robot.
But robot vision is coming along, last year, a robotics company called Berkshire Grey started working with FedEx to automate package processing, largely on the strength of their robotic vision system.
Berkshire Grey’s optical scanners can read barcodes from different angles, and to do that, they have to recognize the different objects from different angles, understand the relationships between the shapes and understand depth in 3D space.
And this is something robots would have to do in a changing environment, recognize objects at different angles, in different conditions, and react appropriately.
Would you trust a robot to cut your hair? Or tend to your flowers? Or carry a baby across the room? It might take a while to build up that level of trust with the robots.
Battery Life
But even if you did trust the robot and let it carry your baby across the room… Could they make it across the room before the power ran out?
Mobile robots run on batteries, and they do chew through a charge.
Boston Dynamics first equipped Atlas with batteries in 2015 for the DARPA Robotics Challenge.
It had two lithium-ion batteries which gave it one hour of operating time.
Similar robots currently have operating times of 90-minutes to just over two hours.
Like Spot, the dog robot from Boston Dynamics, it has a 90 minute battery life. But that battery takes 2 hours to charge. So…
Of course, lithium-ion isn’t the only option, we’ve talked about of a lot of others on this channel, one that’s been considered for robots are a zinc-air battery.
This not only saves weight, but it could also be structural, basically built into the robot’s frame, which means it would evenly distribute the weight and it wouldn’t always be adjusting for one center of gravity high above the ground.https://news.umich.edu/biomorphic-batteries-could-provide-72x-more-energy-for-robots/ — has video on structural batteries and toys using the new battery
Which is interesting because that’s kind of how we store energy in our body as fat reserves. So this is kind of like robo-fat.
One more group working on this at the University of Michigan have tried this using Kevlar. They say they’ve demonstrated it on a small scale and it could provide 72 times the energy of a single lithium-ion battery.
But you know… Even if a robot only could hold a charge for 30 minutes to an hour at a time… There’s still a lot you could do with that.
I mean I imagine if you had a robot at home it would spend most of the time just resting and waiting until it’s given something to do, it could plug itself into a base and stay charged all that time.
And when you need it to do the dishes or take out the trash or cook some food, it could perform that task, and plug itself back in.
Like I’m sitting here thinking about it and we have two dogs, and whenever we go out of town we have to find someone to keep them or board them or find a house sitter – if we had a robot, it could be programmed to watch the house. Feed the dogs, let them outside every few hours, water the plants, bring in the mail… be security.
Yeah, I imagine even if they could only operate for 30 minutes to an hour, that would be enough to get some pretty good use out of them. Assuming it’s actually able to do all those things.
Will They Take Our Jobs?
Of course robots folding our laundry in our homes is one thing, it’s going to take a lot more power or fleets of robots to perform all day work in factories and warehouses.
And this is where things do get a little… concerning. A robot labor revolution is probably inevitable. They are going to take human jobs at some point.
According to the World Economic Forum, 85 million jobs now done by humans will be automated by 2025.
But, they expect that over the same period, 97 million new human jobs will be created.
Not just to deal with the robots but in new and emerging technology fields.
But these aren’t the same jobs the robots will take away. Workers will need new skills and training to get those skills. That’s going to need to be paid for, somehow.
Of course it’s not just robots, technology is changing rapidly and the job market is industry is changing along with it, the World Economic Forum expects half of all workers will need new skills in the future.
And we’re talking near future. This is pretty urgent.
Robots In Healthcare
One field that’s having a bit of a crisis of conscience around robots is the healthcare industry.
We do have a shortage of healthcare workers right now and robotic assistants could help fill that gap. But… healthcare is a very personal thing, do we want to hand that over to robots?
Psychologists have actually studied how patients respond to robot healthcare workers and it’s pretty interesting because obviously some people hated it and some people liked it but the people who liked it preferred an extroverted, feminine personality.
They found that people find it easier to accept medical advice from extroverted, feminine robots, also playfulness in a robot made a big difference.
People rated playful robots as superior when it came to performance, but found that playful robots had more difficulty motivating patients to take their medicine or change their habits.
There’s probably some lessons to learn there about human healthcare workers as well…
But really as our population ages, there will be more demand for robot caregivers, the question really is are we ready to do that as a society?
As robots improve and become more ubiquitous, feelings will shift for sure. But it’s going to be an interesting transition.https://www.nature.com/articles/d41586-022-00072-z
Looking to the Future
So am I going to have a Teslabot folding my clothes anytime soon? Or a Hondabot? Or an… IKEAbot? Papa Johnsbot?
Probably no to all of those. In fact, here’s an interesting question, which do you think we’ll see first, robots folding our clothes or people walking on Mars? I’ve got my answer, what’s your answer?
There’s still a lot of problems to overcome, but they are being chipped away on, and as they become more useful, I think we’re all going to see a lot more of them.
Whether that’s a good or bad thing, I guess we’ll see.