Skip to Content
Artificial intelligence

That Atlas robot might be doing parkour now, but it still can’t chase you up the stairs

October 12, 2018

What should we make of seeing Atlas, an advanced humanoid system developed by Boston Dynamics, leap over logs and bound up steps, Parkour-style? Watch the full video here.

You might remember the hilarious video showing several humanoid bots, including a few Atlas machines, toppling over, collapsing, and generally freaking out during a big robotic challenge held a few years ago.

So what’s going on? Has Atlas suddenly developed new ninja-like agility? Not exactly.

Navigating the real world remains a huge challenge for legged robots. Just balancing on two legs, in fact, takes extraordinary dynamic sensing and control. The accident-prone robots really struggled when trying to do something besides balancing, like opening a door.

The video does suggest that the latest version of Atlas has very impressive dynamic balancing capabilities. Without this, the robot would be unable to cope with the slightest slippage that naturally occurs when you put your foot onto a surface at high speed. This technology is the secret behind Boston Dynamics’ remarkable robot menagerie.

Marc Raibert, the founder and CEO of Boston Dymanics, says this version of Atlas uses some novel sensing capabilities to judge each leap. "To perform these tasks Atlas uses its vision system to align itself and measure distance on the approach to the obstacles," he says. 

Raibert adds that the new systen also features some new hardware. He says it is significantly more advanced that the bot shown performing a backflip last year.

Impressive, but it's still a pretty controlled setup. Atlas also still requires a fair amount of human control. So, as cool as this new video is, it’s worth remembering that in the unfamiliar and unpredictable real world, a robot like Atlas is probably still more likely to trip and fall than run after anyone.

And if you’re really afraid, well, you can always just shut the door. (Unless this one is after you, of course).

Updated with comment from Marc Raibert, CEO of Boston Dynamics.

Deep Dive

Artificial intelligence

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

What’s next for generative video

OpenAI's Sora has raised the bar for AI moviemaking. Here are four things to bear in mind as we wrap our heads around what's coming.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.