1X shows wild month-on-month acceleration in humanoid autonomy

A month ago, we were impressed by these robots being able to pick things up, put them in the right spot, open doors and charge themselves. But new video released hours ago makes it clear that autonomous humanoid work is starting to accelerate like mad.

Norwegian robotics company 1X is OpenAI’s other bet in the humanoid game – the bigger recent headline being its investment in, and collaboration with American company Figure.

1X’s Eve robots don’t have legs; they roll around on wheeled platforms. They don’t have the extraordinary hands that companies like Sanctuary AI, Figure and Tesla are making, settling instead for stumpy-looking claw grippers. And they’re pretty underwhelming to look at, too – their smiley faces, frankly, look silly, and as we’ve noted before, they’re draped in little tracksuity arrangements that make them appear like they’re late for their luge race.

But they’re beginning to learn at a truly dizzying rate. First, take a quick scrub through this video published on February 9, 2024.

All Neural Networks. All Autonomous. All 1X speed | 1X Studio

Now, take a look at 1X’s March update video, less than six weeks later.

End-to-End Autonomy | March 2024 Update | 1X Studio

The sophistication of the tasks these bots are now handling – fully autonomously, all working from the same behavioral AI model – has taken a giant leap forward in this short space of time. Let’s recap what this video shows:

  • Multi-step task handling, as it picks items out of a shopping bag, decides where to put them, and opens and closes the cupboard door after putting one away
  • Using soft tools, in this case the use of a cloth to wipe up a spill
  • Deft handling of a physically tricky task without thumbs, as it gets all four tabs of a cardboard box folded down together
  • Fully autonomous shirt folding, requiring the robot to delicately handle deformable surfaces

The latter, of course, is a rather cheeky inclusion, probably referencing the embarrassing gaffe Tesla made showing off its Optimus bot doing the same job – but later admitting the video was showing the robot being teleoperated by a human.

But Optimus is built on a learning-capable AI model as well; we’re certain that the next Tesla Bot update will be every bit as stunning as what we’re starting to see happening at every other AI-driven humanoid company, all at once.

This feels like the GPT-3 moment for humanoids – the point at which they’re suddenly popping up onto everyone’s radar. And they’re taking their first wobbly steps into the workforce, notably Agility’s Digit at Amazon, Figure’s 01 at BMW, and Apptronik’s Apollo at Mercedes-Benz.

Figure Status Update – OpenAI Speech-to-Speech Reasoning

Videos like these, as well as the video above, in which Figure’s 01 starts to speak and reason as it works, and the video below, showing Sanctuary’s Phoenix beginning to display a staggering speed and fluidity of movement – these are making me feel that same stomach-lurching sense of acceleration that’ll be familiar to anyone who’s tried to keep up with AI and robotics over the last year or two.

There are plenty of naysayers that see little of significance in what’s happening in humanoid robotics, but what I’m seeing is an entire industry taking its first baby steps. These AI models are just beginning to learn how to interact with the world using physical bodies, they’re putting together the foundational building blocks of useful work.

They’re not being programmed; they’re figuring these things out more or less for themselves, with their human helpers guiding and helping. And all of these next-gen humanoids are incredibly young, most have only been operating for a year or two. Take a look at these videos again and ask yourself how quickly a human baby could pick them up.

And remember – these are older models, running on chips and hardware that have probably been superseded in the last 36 hours. The acceleration in this field is already staggering to me, and we ain’t seen nothin’ yet.

Source: 1X

Latest articles


Related articles

Leave a reply

Please enter your comment!
Please enter your name here