Xiaomi-Robotics-1

Every single thing that’s come out of China in the entire AI/Robotics space lately has been a massive net positive for the world

Open Source is a good thing, regardless of where it comes from

There tends to be a lot of low-key (and some not-so-low-key) anti-Chinese xenophobia on HN, so that's likely a big part of it too.

We destroyed our planet to get there. Great achievement.

> You're telling me we've finally got robots that can do our laundry?

Did you watch it actually fold that shirt? Super slow and bad at it.

I am of course impressed by the model and the robot. Great stuff,

But the downside is that won't we all just turn into the obese, stupid humans at the start of the WALL-E film?

Mindfulness teaches us that we can enjoy the washing up, the laundry, planting seeds, tending the garden. It connects us to something. Without this will we end up with a rise in mental health issues, because for one thing people will just end up on social media in the time they save.

Just a thought.

The pessimism in the comments here is entirely the reflection of the times we're living in.

Technology got weaponized against the average citizen.

It has become a synonym for unemployment, enshittification, mass surveillance and social erosion.

If this video appeared in 60's or 70's, people would be in awe imagining the future where robots do everything.

A subset of the people on HN hate AI and yell "slop" at everything they see. It's even worse if you go to other social websites.

Just ignore these people. If they don't use AI, they likely won't remain practicing in our field for much longer. They will naturally self-select themselves out of relevancy.

I was Claude coding and got hit by the 3 AM reset bug. I went on reddit and saw hundreds of other engineers hit by the same. We are different people than the AI haters, and there has never been something so powerful it's kept me up past 3 AM every single night working beyond every fiber of my body telling me to sleep. I feel empowered as if I command an army. And there are millions of us. The sheer might of all of this will win out over negativity.

I've built multiple successful businesses in the last year. If you're doing this with any sort of interest in self-employment you don't need your employer anymore.

I have never been happier. This is the best future, and I feel lucky to be living it. It's what I dreamed about when I was a kid, and it feels magical. As far as I can tell, the future has no limits.

SciFi novel "The Mote in God's Eye." has aliens with three hands.

One "gripping hand" kinda coming out the gut, and two dextrous hands for detail work. Led to abuse of English language with: On the one hand... On the other hand... On the gripping hand...

My thoughts exactly, but to train it you'd need a lot of data with someone manually operating it. With just two hands it's enough to have a lot of footage of humans doing it. Maybe down the line the robots will be able to self-improve or someone will come up with ways to optimize this..

Anybody who worked on a car knows an extra hand would always be helpful.

Yeah, folding sheets requires 4 hands. With a large distance between each pair.

I have been slopfolding since before it was cool

Isn't that the story of <any> technology? When the coffee machines first came to be, Barista was definitely a better option. However, through the evolution, in many cases today machine does better job than barista.

Folding clothes is already a compromise between leaving them unfolded and hanging them

saw the way it folded the clothes and thought "better than me" :/

I suspect you might be jumbling up your smalls with your origami.

How privileged was your upbringing?

When AI is a the forefront of scientific breakthroughs and all human endeavor is painfully minuscule compared to AI, we'll still be calling it slop as we all hypocritically generate more and more slop.

I wonder, what will you do with that free time you will crave?

I think this perspective is now demonstrably false. It was basically the defacto position of most roboticists until a few months back. But we've already seen the cost of humanoid robots fall precipitously (even without the kind of mass production that genuinely useful home and real world intelligence would bring).

Which seems more likely, that cheap humanoids will continue to improve in dexterity, battery life, strength etc. Leveraging the vast investment in research into real world task training. Or - that we'll redesign millions of everyday objects to fit the limitations of current task specific robots?

If you answer the latter, I've got a whole bunch of mid century quirky labour saving inventions in the back room I'd like to sell you.

I can't believe someone shares my point of view. We have already many non humanoid robots to do boring stuff (underwater maintenance, stuff like that). For some reasons people tend to think that for robots to do stuff close to humans, we need human like robots.

We do not. I want a swarm of insects-like robots to fight mosquitos. A snake one that fixes my pipe. A spider-like robot that cleans surfaces. Etc.

I will suggest mix of both: some everyday tools are already built for being used with hands and legs (e.g opening a door). But indeed adding some different grip types, additional arms, etc, could be helpful.

Exactly what I came here to say. A robot doesn't have to look like a human. Just a few arms that does the job neatly is more than enough.

When the page says "embodiment-free (UMI)" it means they've used a standardised gripper and camera arrangement [1], while aiming to be agnostic to the type of robot arm that gripper is attached to.

[1] https://umi-gripper.github.io/

I’d imagine the training is hardware specific

Yes, it's truly remarkable how small modern VLAs are.

Even Tesla's FSD AI stack is small by modern LLM standards. Once the hardware is there to enable real time 100B inference on endpoint SBCs? We'd see gains in AI robotics from that alone.

Why cross-entropy follows a curve is fairly trivial to understand:

https://www.youtube.com/watch?v=GlYgs6v2YfU

In general, motion planning multi-axis systems is NP-Hard. Have a great day =3

I'd like to see a serious HN discussion thread about WAIC first.

I watched it on mute and was still satisfying, so I guess it's not the music.

They're trying to show their investors a butler robot that can do all types of household chores.

Folding fabric has traditionally been very hard to solve with robots; the ideal places to grab the fabric might be hidden, the fabric patterns are repetitive, and the material flops around, tangles in the grippers, and brings other pieces of laundry with it.

Successfully sorting and folding a basket of laundry shows they've got a control system capable of performing some quite difficult tasks.

As a parent with 3 kids and both my wife and I working, folding clothes is the most tedious household chore. It's physically and mentally exhausting (unlike doing the dishes, which is only physically exhausting as I can wash piles of dishes while zoning out). And not to mention take so much time, a week's worth of unfolded pile of clothes for our entire household can take me 1-2 hours to fold uninterrupted (2-3 hours if I do it while watching Netflix or something).

Sorting and folding laundry is one of the most tedious chores in our household. Do you have kids?

Funny you mentioned H&M. They are boycotted on China due to their public stance against slave labour:

https://www.bbc.com/news/business-57691415

It is important to remember that China is not only about open weight LLMs, fastest cars and flying taxis. There is still a lot of very serious human right abuse going on in this country.

I think its a ubiquitous task which requires some amount of precision dexterity, so its an easy to digest example.

A lot of smaller chores look trivial when considered in isolation, but have disproportionate time and mental cost because of how they fit in the flow of life. Sometimes the effect is similar to interrupting an engineer for a couple of minutes.

I have a pile of washed clothes that moves from the bed to the chair for a good week before it finally goes into the closet.

Come back when you have a family and you run a washing machine full of clothes per day.

It's meant for war. Today it folds your clothes, tomorrow it plants mines and shoots at you on a battlefield.

From my skim of their paper, they have gotten robodojo 14% versus previous sota 9%. So its an improvement, but doesnt seem like a reliable model yet, which is fine. I also notice the model becomes worse with more data on the printer refilling task, so thats either a bad run or its a data point against the robustness. The transfer learning results are meaningful. 75% on new complex tasks from <10 hours of demo is really good. I believe this is a 2x improvement over the SOTA. 10 trials per task is standard for these papers but is pretty weak statistically. Memory tasks its weaker at but they have acknowledged it, and I cant see why this arch would have problems w that in the future.

Well, couldn't it have been trained on the 2-second fold trick? Besides, it might require more precise or quick movements than a regular fold.

Automated ironing sounds like massive fire hazard even for showing off video.

You could get wrinkle free shirts and a plastic folding board / helper, and you are 90% there for 1% of the price ;-)

Hand held steamer > ironing

The video also tells you that there is an uncut video at the bottom of the page. Did you take a look at that?

The uncut video is at the bottom of the page.