Imagine spending your weekends free of chores. Forget about the laundry, the dishes, and tidying up. It’s all been taken care of. And the one responsible is an amalgamation of motors, wires, and computer chips: a robot.
That’s at least part of the promise of humanoid robotics. And Silicon Valley companies are lining up to get in on the action. Amazon (AMZN), Google (GOOG, GOOGL), Microsoft (MSFT), Nvidia (NVDA), Tesla (TSLA), and now, according to Bloomberg, Meta (META) and Apple (AAPL), are each funding or developing humanoid robotics systems that could lead to a world where robots walk among us.
There are already a number of robots in the world today, whether they’re moving objects in warehouses, delivering food to your table at restaurants, or, like Amazon’s Astro bot, rolling around your home like a motorized sentry.
Humanoid robots, however, will be able to perform more specialized tasks using hands that can grab and grip without crushing objects and interact more seamlessly in a world that’s built for us. They could be deployed across factories to address labor shortages, assist in caring for the elderly, and perform tasks in environments that would put human lives at risk.
Optimus, also known as Tesla Bot, is a general-purpose robotic humanoid under development by Tesla, Inc., Hong Kong, China. (Bob Henry/UCG/Universal Images Group via Getty Images) ·UCG via Getty Images
“We have such a huge need for labor, for taking care of our aging and disabled populations. It really is such a huge cost to provide that care, and we can’t provide it to everybody,” explained University of Michigan robotics professor Chad Jenkins.
“Taking care of our aging and disabled populations will provide such a huge benefit to society through improving quality of life, improving our productivity, and robots provide a great way to be able to do that,” he added.
But don’t start planning for care-free weekends just yet. There are still plenty of complex problems to solve before humanoid robots are as ubiquitous as our smartphones. And getting there could take decades.
Humanoid robots like Tesla’s Optimus, Agility AI’s Digit, and Figure AI’s 02 are bipedal machines that can interact with their environment in a manner similar to humans. That means picking up and moving objects with artificial hands, or in some cases claws, and manipulating them in ways that less sophisticated robots simply can’t.
It’s not just how robots interact with objects and the world around them, though. They’ll also have to be able to see objects and recognize what they can do with them, a concept referred to as affordance.
“Affordance basically says, whenever I see my mug, I know I can grab that handle, and then I can take a sip out of it or pour out of it,” explained Jenkins.
A Robotic technology server moves to deliver a meal at the Robot Cafe, the food hub featuring robots that serves customers in Nairobi, Kenya, Aug. 22, 2024. REUTERS/Monicah Mwangi ·REUTERS / Reuters
“That will bring together the hand with the computer vision with all the reasoning that our robots can do. And so that’s a big step forward … for a robot to know when and where it can do things and what’s going to result, and then what it can do after that,” he added.
Getting robots to understand when and how they can use or interact with objects, whether that’s a mug, a door, or a piece of heavy equipment, requires an incredible amount of training. There are currently a few ways to “teach” robots. The first is through mimicking human movements via teleoperation.
To do this, a person wears specialized equipment and performs a set task. The equipment, a glove or headset and suit, collects data on the person’s movements and then transfers that information to the robot, effectively showing it how it should move, explained Pulkit Agrawal, associate professor of electrical engineering and computer science at MIT.
The second way to train robots is via simulations. During his keynote at CES 2025 in January, Nvidia CEO Jensen Huang showed off his company’s approach to teaching robots via simulations called Nvidia Cosmos.
According to Agrawal, researchers use simulated teaching to create a variety of scenarios in a digital world and then translate that knowledge to a robot so it understands how to behave in the real world. The benefit of this approach is that it’s fast and doesn’t require putting a pricey robot prototype in danger of falling over or breaking something.
Nvidia CEO Jensen Huang delivers a keynote address at the Consumer Electronics Show (CES) 2025, showcasing the company’s latest innovations in Las Vegas, Nevada, USA, on Jan. 6, 2025. (Artur Widak/Anadolu via Getty Images) ·Anadolu via Getty Images
“But that also has its own set of challenges,” Agrawal said. “For example, the simulation is different from the real world. You have to bridge that gap. And some things are still hard to simulate like clothes and things which are deformable, things which are soft.”
A third approach fuses these two paths, teaching robots via both real-world human movements and simulations, but that’s still a work in progress.
“We’re trying to get the best of both of those worlds. And that’s still research that we’re exploring,” Jenkins said.
Outside of training robots how to interact with the world, researchers also have to ensure that humanoid robots are not only safe to use in settings like the home but can also operate for long periods of time.
“If a humanoid robot falls down at home, one thing is the cost to repair, the other is the safety of the humans in the home,” explained Ye Zhao, an assistant professor at Georgia Tech’s George Woodruff School of Mechanical Engineering and a member of the Institute for Robotics and Intelligent Machines.
“The reliability of humanoid robots still needs a lot of work,” he added.
At the moment, humanoid robots cost tens of thousands of dollars and don’t last long before they need to be recharged. They also overheat, preventing them from moving for extended periods of time.
“There isn’t a humanoid hardware that you can buy and say it’s reliable. That simply doesn’t exist today,” Agrawal said.
“People are claiming they’re going to have humanoid robots which are $15,000, $20,000. But if you actually go and look up the costs [it’s] more like $50,000, $60,000. They still heat up. They’re not reliable. They’ll last for half an hour to an hour.”
That doesn’t mean the technology won’t get there eventually, he said. Just don’t expect to see a humanoid robot for sale in your local Best Buy anytime soon. Until then, the closest we’ll likely get to a practical, in-home robot is the trusty Roomba.
Sign up for Yahoo Finance’s Week in Tech newsletter. ·yahoofinance
Email Daniel Howley at dhowley@yahoofinance.com. Follow him on Twitter at @DanielHowley.
