Homepage AI ‘Reconstructing the real world’: New AI clones your living room...

‘Reconstructing the real world’: New AI clones your living room to teach robots how to walk

AI robot taking selfie with mother and daughter in living room
Shutterstock

When you point a smartphone camera at your living room, you probably just expect a quick photo. Most people never consider where those visual details end up. But behind the scenes, those everyday snapshots are quietly feeding a massive new technology.

For years, language programs have completely dominated the artificial intelligence conversation. They simply guess the next word in a sentence to write essays or answer complex questions.

Now, developers are moving beyond text. They are building complex digital replicas of physical spaces, known as world models.

These tools use visual data to simulate gravity, space, and motion. According to Wired, companies are using everything from videos shot by freelance workers to old data from games like Pokémon Go to build them.

As the chief technology officer at Niantic Spatial, Brian McClendon sees huge potential in this shift. He told Wired, “We’re very much about reconstructing the real world.”

Training the machines

The main goal is to give robots a safe place to practice. Instead of testing a machine in a real kitchen, engineers can run millions of tests in a digital clone.

Oliver Cameron, chief executive of the AI lab Odyssey, explained the shift to Wired.

He noted that while text models predict words, these new systems predict physical outcomes. “With a world model, you give it a visual state of the world, an image, and it will then predict the next visual state of the world,” Cameron said.

He added, “They’re simulations, but they look and feel real.”

Beyond home robots

Soon, this same technology could guide driverless cars through difficult weather. Wayve is one startup designing these models specifically for vehicles, using dashcam footage to recreate dangerous roads.

At Wayve, Jamie Shotton serves as chief scientist. He stressed that these digital systems must comprehend cause and effect before hitting the streets.

“Importantly, a world model must be able to react to the actions that we take in the world, and so it must understand physics, cause and effect, and how other objects and people change their actions in response to our own,” Shotton told Wired.

Major investors are certainly noticing the financial potential. Wired reported that tech giant AMD recently bought the startup World Labs for more than $8 billion.

Ads by MGDK