He has changed. The machines are still unbuilt, but the money is flowing: companies and investors have invested $6.1 billion in humanoid robots in 2025 alone, four times the amount invested in 2024.
what happened A revolution in which machines have learned to interact with the world.
Imagine you want to install a pair of robot arms in your home to do just one thing: fold clothes. How will it learn to do this? You can start by writing the rules. Before tearing, test the fabric to determine how much damage it can withstand. Identify the shirt collar. Move the gripper to the left arm, lift it, and turn it inward the exact same distance. Repeat for the right arm. If the shirt is rolled, turn the plan accordingly. If the sleeve is twisted, correct it. Very quickly the number of rules explodes, but a thorough calculation of them can produce reliable results. This was the original skill of robotics: anticipating every possibility and encoding it in advance.
Around 2015, Cutting Edge began to do things differently: create a digital simulation of robotic arms and clothes, and give it a reward signal every time the program successfully folded and a ding every time it failed. Thus, it gets better by trying all kinds of techniques through trial and error, with millions of iterations—that’s how AI is good at playing games.
The arrival of ChatGPT in 2022 triggered the current boom. Trained on vast amounts of text, large language models learn to predict which word should come next in a sentence, not through trial and error. Similar models tailored to robotics were soon able to absorb images, sensor readings, and the position of the robot’s joints and predict the machine’s next action, issuing dozens of motor commands every second.
This conceptual shift—to rely on AI models that use large amounts of data—works whether a helper robot talks to people, moves through an environment, or even performs complex tasks. And it was combined with other ideas about how to accomplish this new way of learning, such as deploying robots, even if they are not yet perfect, to learn from the environment in which they are meant to work. Today, Silicon Valley roboticists are once again dreaming big. Here is how it happened.
Jebo
Jebo
An animated social robot talked long before the age of LL.M.
An MIT robotics researcher named Cynthia Breazeale introduced an armless, legless, faceless robot called Jibo to the world in 2014. It actually looked like a lamp. Brazell’s goal was to create a social robot for families, and the idea raised $3.7 million in a crowdsourced funding campaign. Initial pre-orders cost $749.
An early jebo could introduce himself and dance to entertain the children, but that was about it. The vision was always for it to become a sort of embodied assistant that could handle everything from scheduling and emails to telling stories. It gained many devoted users, but the company eventually shut down in 2019.
Courtesy of MIT Media Lab
In the past, one thing Jebo really needed was better language skills. It competed with Apple’s Siri and Amazon’s Alexa, and at the time all those technologies relied on heavy scripting. Broadly speaking, when you speak to them, the software will translate your speech into text, analyze what you want, and create a response extracted from pre-approved snippets. Those pieces could be catchy, but they were also repetitive and just plain boring.–Straight robotic. This was especially a challenge for a robot that should be social and family-oriented.