Ameca’s incredible expressions and interactions are part of a major leap for AI and robots. They can now see the world and react to it, which means they’re starting to replace human workers. Robot muscle and AI’s creating immense wealth and removing millions of jobs, but the upsides are stunning. I’ll also explain all this. People often mistake this for a real dolphin – but it’s a robot designed to set dolphins free, by replacing them in marine parks.
Thousands of dolphins are locked up in parks, but closing the parks would cut off the millions they raise for conservation. They’re working on a new version to perform entirely by itself – and that could be the point where they can impress the crowds and replace the dolphins. And to help reduce the plastic that ends up in the ocean, recycling can now be sorted by robots like this.
It can see different types of plastic, pick them up and throw them into bins. The things we order online are also being handled by increasingly advanced robots. Stretch, from Boston Dynamics, can work alongside its friend Spot. It can deal with large volumes of boxes, and of course, it can dance. A lot of the food we eat is also moved by robots, or made by them. And humanoid robots are entering warehouses. Starting in back rooms, once they’re considered safe, they’ll work alongside humans. And it’s hard to compete with robots that can work 24 hours a day, without pay – particularly at Amazon, which is infamous for pushing its workers to their limits.
“My workday feels like a nine-hour intense workout, every day, and they track our every move.” Amazon workers were twice as likely to be seriously injured as workers at other companies. One thing that’s preventing robots from taking over entirely, according to an amazon executive, is that humans are good at quickly recognizing and sorting products. And robots are rapidly developing this kind of skill. Digit sees the world through lidar and depth sensors. Most of its energy isn’t used for movement, it’s used for computing. The company expects its robots to go on to help you around the home. Robot hands are becoming impressively dextrous.
They can use tweezers, scissors and hold tricky objects. And robot bodies are starting to capture more natural human movement. Ameca is being used to test and develop AI, so it’s going to get smarter. And robots are rapidly learning new skills through simulations. Here’s a simple example, where an AI learns to jump over an increasingly tall barrier. It also learned to fight, and to master an obstacle course. And robots have learned incredible dexterity from simulations, including rubik’s cube skills. We create thousands of different simulated environments. This means like thousands of years of experience that this neural network has had in simulation.
Everytime the algorithm has gotten good at the task, we make the task harder. This ability to generalise to new environments feels like a very core piece of intelligence. Here, AI solves an incredible 55×55 cube. This machine learned to play table tennis in just 90 minutes, returning 98% of balls: And some AI’s have outsmarted their creators – like this spider which was asked to minimise the time its feet touched the ground. When it reported that it had leant to move without its feet ever touching the ground, its creators were shocked to find that it had turned itself over.
when AI’s learned to master hide and seek, one of them found that it could use a ramp to jump outside the game walls. The rapid progress of AI is giving robots incredible skills. And it could enable new machines like this bird-like evtol, designed to land in difficult terrain in Africa, carrying medical supplies. It’s an ambitious project, but the team has some character. It was designed to blend in with the surrounding landscape. Unlike most other drones which don’t actually fly around, they’re just so ugly, the earth repels them from the ground. Some robots can already walk and fly. And a flying humanoid robot is in development. Designed to help search for survivors in disaster zones. Creative new designs keep emerging, like this one that can accelerate rapidly like a car, walk a dog, or stand up like a human. It can carry things pretty much anywhere. Of course, carrying humans at high speed requires a higher level of safety.
You can always think of tesla as, the world’s biggest robot company, um, or, semi-sentient robot company. We’re effectively creating the most advanced, practical AI. It would be tempting to write it off as hype, but they’ve created some incredible technology. The AI behind autopilot will also power the Tesla bot. What I find kind of fascinating about this, is that we are effectively building a synthetic animal, from the ground up. It moves around, senses the environment, and acts autonomously and intelligently.
We are building the synthetic visual cortex. So the processing starts when light hits out artificial retina and we are going to process this information with neural networks” The cars also work together. Here different cars driving the same route combine their data to build a more detailed image of the environment. Cars also shared ten thousand clips of wind and snow, to learn to identify things from all angles, but also to remember that they’re still there, even if they’re covered up. The cars have a big advantage over us. While humans focus on a small area at once – a problem exploited by magicians You were focused on your hand, that’s why you were distracted. While you were watching this I couldn’t quite get your watch off, it was difficult.
Yet you had something inside your front pocket, do you remember what it was? Money? Check your pocket, see if it’s still there. Is it still there? You’re human you’re not slow” AI can see everything in its field of view at all times, and pick out what’s important. Tesla also uses an impressive simulation, to train its AI. Notice the road is cracked and patched up. They create unusual situations, like this couple and their dog running on the road. Musk believes we all live in a simulation. Simulation theory shows that if the sims continue to improve, even at a slow pace, eventually they’ll become indistinguishable from reality. And there will be many of them, so the chances that we’re living in the one reality is very small.
Virtual characters are getting spookily realistic. Is this the real Keanu Reeves? The choices we make, the worlds we build, also confront us with questions, about why we want to choose this over that. Or is this him? It was important for me to ask people, how do we know what is real? You can probably tell, but it’s getting harder. In just 35 years, we’ve gone from this to this. Who knows what we’ll have in a thousand years. And what would reality mean, when a world we can build feels as real as our own? And Tesla’s building an impressive matrix.
Its cars have been trained on 300 million images. In just one training project, 10 billion labels were applied to 2 million clips, using 20,000 CPU cores. And Tesla has built its own incredibly powerful training matrix. It’s designed to be the world’s fastest AI training machine and the most powerful computer. This new chip is more powerful than most computers, and there are 25 of them in this AI training tile. I can’t believe I’m holding nine petaflops out here. They’re connecting 120 tiles in one computer – 3000 chips in total. Straight after announcing this, Tesla revealed plans to build the Tesla bot. Neural nets, recognizing the world, understanding how to navigate through the world. Uh, it kind of makes sense to put that onto a humanoid form.
It’s intended to be friendly, of course, um, and navigate through a world built for humans and eliminate dangerous, repetitive, and boring tasks. Most of the one million warehouse jobs in the US, and millions more in other sectors. Musk is straightforward about the impact of this. What happens when there is, er, you know, no shortage of labor, um. This is why I think long term that there will have to be universal basic income” The robots have a screen for a face, which could show information or expressions.
It’s powered by the same computer used for autonomous driving, the same cameras – with two sets of eyes – and will learn via simulation, in their supercomputer. The robots will generate incredible wealth. It obviously has profound implications for the economy because, given that the economy at its foundational level is labor, I mean, is there an actual limit to the economy? Maybe not. Robot workers have already made Musk the world’s richest man, and he could be the first trillionaire.
It’s made him a target, in a country where many struggles to pay the rent and half a million are homeless. Please don’t call the manager on me, Senator Karen. She struck first, obviously. Yeah, she did. She called me a freeloader, and a drifter who doesn’t pay taxes basically, and I’m literally paying the most tax that any individual in history has ever paid this year, ever, uh and she doesn’t pay taxes, basically at all, and her salary is paid for by the taxpayer. If you could die by irony she would be, she would be dead. Whatever you think of all this, the wealth gap is growing. So what happens when billionaires start building humanoid robot workforces? Countries with higher wealth gaps have more crime, more health problems, and lower levels of satisfaction and happiness. They also have lower economic growth when money goes to luxuries like super yachts instead of workers.
Researchers are experimenting with a solution that could help everyone, but first, let’s have a look at personal robots. It could be actually a very good companion. It could develop, like a personality over time that is, that is like, unique, and the, suttle perfections of the personality of the robot, could actually make an incredible, buddy basically. In that way. Like R2D2 or like C3PO sort of thing you know. Where are you taking these prisoners? These are prisoners? Yes, where are you taking them? I am taking them, to imprison them, in prison. He is taking us to. Quiet! Can you spot the Boston dynamics robot used in this star wars TV series? Humanoid robots could help with a big global problem that might surprise you. The birth rate has been declining for decades.
I think the biggest problem people will face in 20 years is population collapse. It’s very easy to see what the world will look like in 20 years because humans have a 20-year boot sequence, so like you say, okay well, who was born last year? Okay now you know what the world will look like in 20 years, it’s that easy. I absolutely agree with that, the speed of population decreasing is going to speed up. Now you call it to collapse, I agree with that. Accelerating collapse. Accelerating collapse. The big problem is that there won’t be enough young people, enough workers. Japan’s already struggling with too few young people, to support its aging population. It’s former prime minister called for more rapid development of robots, to help overcome the problem. People in Japan are already dating AI and robots. That might be a symptom of isolation, but AI conversations are getting more interesting. GPT-3 learned to talk by reading hundreds of billions of words on the web.
What do you think is organic? And what do you think is artificial? Good question, I think everything that was ever made somehow stems from organic, even if it was processed to the point most of us like to call it artificial, how about you? I think that’s true, in times past, things felt more organic because people used handcrafting to make everything. Now everything is programmed and let’s face it, if we go way back in history, you couldn’t even make a wooden spoon, nevermind a skyscraper without a machine. The AI was initially thought too dangerous to release, in case it wrote huge volumes of fake news, but thousands of developers are now working on it. One man tailored the AI to sound like his fiance, who died ten year ago, and OpenAI removed his access. Robots are now learning to track our eyes and move theirs more naturally. Even for humans it’s a tricky skill, which can have a big impact on a first date. Hi, how are you? Good, Drew. Aleisha. Nice to meet you. Here their gaze is all over the place, and it was really awkward. She got on better with this guy, and their gaze was more calm and direct. Eye tracking has been used to bring characters to life in VR and in experimental robots like this. And now with Ameca, it’s eyes, face and its body all react to what it’s seeing. Look at the way it leans back.
Artificial muscles will create even more realistic movements. Just look at the range of motion in this hand. It’s incredibly strong – the weight is 7 kg, and here it lifts 26 kg. It’s powered by water pressure, with half as many muscles as a human hand, and sensors in each joint. There are 42 muscles in the human face. Once they’re recreated in a robot, AI can apply infinite expressions. At the moment, pre-programmed and remote-controlled robots get a lot of attention, like Boston dynamics dancing robots and this robot dog facing off with a cheetah at Sydney zoo. It’s an experiment to see if the robot could be used to control the Cheetahs if they got into a dangerous situation. But the real revolution is going on behind robots’ eyes. So, the lightsabers. Apart from a bit of fun, it shows two kinds of robots. Atlas is incredibly impressive, but its best moves are pre-programmed, and it’s largely independent. It has one life. With Tesla bot, they’re working towards a robot that can teach itself to perform many tasks. It will be part of a huge AI network. The robot itself is less important, and if the AI wanted to survive, it would be very hard to kill. Hello again.
AI doesn’t have to be evil to destroy humanity. If AI has a goal and humanity just happens to be in the way, it’ll destroy humanity as a matter of course, without even thinking about it, no hard feelings. It’s just like if we’re building a road, and an anthill happens to be in the way, we don’t hate ants, we’re just building a road, and so goodbye anthill. For now, it’s taking jobs. Up to half of all US jobs are expected to disappear over the next ten years I think long term that there will have to be universal basic income. A Stanford study of several UBI projects found some interesting results. People who received money regularly didn’t work less. They did spend more time in education, with higher school attendance. Their health improved, and rates of the disease dropped. Basic income also allows people to take risks, like this guy, who’s doing amazing things with robotics. Most of the world’s 40 million amputees can’t afford prosthetics – particularly children who grow out of them. To bring the cost down, these arms are 3D printed.