Almost six months ago, Figura announced that its shiny silver humanoid robot had landed its first job at a BMW manufacturing plant. He’s been training for the occasion and a new video shows just how far this all-rounder has come.
There are a lot of these AI-powered humanoids in development right now, with all kinds of different hardware and learning approaches, but they’re all focused on the same goal: take any physical task a human can do and do the job. cheaper and more consistently.
The more they learn, the more robots these companies can sell, potentially decoupling economic growth from population and freeing up a labor pool that is limited only by resources. It sounds like a wild fever dream, but the hardware looks surprisingly capable, and while the AI models that train and control these robots are certainly still in their toddler stages of development, they certainly seem to be advancing at an impressive pace.
You can really see how these machines will differ from traditional industrial robots when they start talking back – in Figure’s case, using OpenAI’s voice engine and language model:
Image Status Update – OpenAI Speech-to-Speech Reasoning
A true general-purpose humanoid is still a long way off, but the companies building them want to get them performing simple, useful tasks in real-world customer operations as soon as possible. It’s probably more to collect data and educate customers on how to work with these new-level AI workers than as an actual source of revenue, but either way, manufacturing seems to be a key industry for early adoption of humanoid robots.
For example, Apptronik made a deal with Mercedes to test its Apollo robot. Even the segment’s venerable granddaddy, the Boston Dynamics Atlas robot, has been trained for automotive manufacturing tasks.
Of course, Tesla has its own massive car manufacturing plants where it can train, test, and get initial value out of its Optimus robot — though character creator Brett Adcock seems like that in-house approach might be too easy for Tesla’s robot. , tweeting, “Having an external customer to fire us if we fail will be a competitive advantage for Figure.”
Today we’ll see how the Figure robot does with training in the fully autonomous “BMW Full Use Case” video below:
Image Status Update – BMW Full Use Case
It’s definitely not the most amazing humanoid robot video we’ve ever seen, but there are plenty of positives; 01 is learning to grasp large, complex shapes with rubber-tipped fingers. It’s learning to move with them without bumping into things – a skill I don’t know I’d encourage in my own children at this stage.
It’s no small feat to be able to place a large plate on top of the jig, with the plate blocking the two locating pins it needs to fit through the holes.
Lifting two smaller pieces with one hand is a little less impressive, as any person would do it two at a time. But he correctly identifies that he has them wrong, orients them and aligns them with the preparation and sets them on the right path. And when he notices that they don’t fit there correctly, he taps them into place with the back of his hand and corrects his own mistakes.
On the other hand… Well, sure, it’s by the hour, not by the job, as one cheeky YouTube commenter points out. To Figure’s credit, though, the company clearly isn’t speeding up its shots, and these kinds of moves will only get faster and smoother over time.
Image Status Update – Real World Task
That might not seem like a huge improvement over the company’s first real-world workload demo from February, as noted above. Both could be categorized as pick-up-and-put-away tasks, so this update feels more evolutionary than revolutionary.
But the 01 is now definitely taking on much more complex and interesting shapes that require more complex grips. And it’s not just placing a box on a conveyor belt, it’s lining up complex parts into precise fixtures.
Although the video is labeled “BMW Group, Plant Spartanburg,” it appears to have been filmed in the Figures’ own building on a test facility. Assuming that’s the case, we’re not sure when the 01 is likely to start seeing action on BMW’s production floor.
Figure
It will be fascinating to see how these robots perform in the real world. They will be slow, cumbersome and limited at first, but as their AI behavior models evolve, their movement and skill acquisition rate will speed up.
But in order to take over the world, be manufactured by the millions, and start making huge societal changes, they will have to demonstrate a clear economic advantage over today’s flesh-and-blood workers. That’s where it gets really interesting!
Source: Image
