Dead or alive, you're coming with me!

A robot that makes your bed sounds like something I could get used to. And that in the week Jensen Huang, CEO of NVIDIA, did not take the warnings about autonomous AI all that seriously and asked the industry to “run as fast as you can”. It is no accident that he expects to sell twice as many chips next year as this year. That is partly because NVIDIA holds an enormous share of the development of robots and the systems that drive them.

And Figure (you know, the American start-up that suddenly put a humanoid up for sale last year, out of nowhere) showed this week how far it has got with that. The company presented Helix 2.5, the AI model behind its robots. It was tested in thirty homes in the San Francisco Bay Area, on jobs like making beds, folding towels and putting toys away. The model had never seen those homes. You can hardly rebuild every customer’s house first because the robot does not understand its job otherwise. It has to cope with a different bed, a full room and things lying where they do not belong.

According to Figure, 56% of the jobs were fully completed. Without training on human behavioural data that was 9%. These are their own test results. A half-tidied room did not count as a success. Do watch the video at Figure yourself.

At 56%, by the way, I would keep the cleaner on for a while. Or in my case: keep sacrificing your weekend.

Sci-fi

The combination of AI and robots has been haunting me for a while. Partly because of the sci-fi films that coloured my childhood. Nobody looks up any more at writing text, finding information, making pictures or building apps and websites. But meanwhile the development of robots is not standing still either. This is something quite different from those 20 running (and falling) Chinese robots on an athletics track.

Once a robot can work reliably in an unfamiliar environment, other uses suddenly come into view. Then it gets interesting for a hotel how many rooms one of these can handle. How often someone has to step in. What happens when something is lying on the floor that it does not recognise. A bar of soap on the shower floor, for instance. 😅

Or robots in hospitals. Collecting the laundry and asking every patient how they are doing at the same time. Or a robot behind the front desk at the office. Receiving clients and fetching coffee. Or RoboCop…!? 😏

Slow

Well; in the end it comes down to this: deploying a robot has to cost less work than supervising it. I spoke to several people about this video this week, and before long it was “It is slow though, I can do that much faster myself!”

But who cares!? As long as the house is tidy by the time I get home!