Hacker News Reader

Comment by ragebol | Hacker News Reader

Having worked on home robotics (in university, for RoboCup@Home so nothing commercial), home robots now finally feel really feasible. The breakthroughs from LLMs and other ML technologies could finally bring this dream to reality.

I'd gladly pay 10k for a robot to do all the chores at home.

View on HN

My take is that this is an opportunity to re-think what chores are, and what white goods are for

Like -- let's say you rent (as we all do, increasingly). Instead of buying different special purpose kitchen gadgets, white goods, and cleaning equipment to haul around with you, why not a general purpose robot that does the easy 80% of all chores?

Yes, it's an opportunity to simplify the home. You don't need labor saving devices if you don't need to save labor.

Hmm… like how smartphones replaced the stereo (badly, but still), Walkman, tape recorder, radio, camera, telephone, etc.

You're saying: why buy a dishwasher and a washing machine and a dryer and a vacuum robot etc if you could buy a home robot to do all of those chores by robot-hand?

The answer is going to be energy efficiency. Your dishwasher uses dramatically less power and water than you would use, if washing them by hand.

Unfortunately, I'm pretty sure most people only care about money efficiency: what choice is cheaper, now and over time.

That’ll still be the dishwasher, then.

> Your dishwasher uses dramatically less power and water than you would use, if washing them by hand.

Only if used optimally, or the handwashing is very wasteful. There is one study I know of, it was paid for by washing machine manufacturers and has quite a few issues (like not removing outliers).

And yes, I measured my usage of water and power.

(I’d still get a dishwasher if my kitchen had space for it)

I think 10k will be out of reach for awhile, but I think we'll start seeing humanoid home robots for ~100k in the next few years. I also think we'll probably see more subscription/leasing of such robots rather than outright ownership.

Why do people jump all the way to humanoid? That little box with an arm on it would not cost more than 10k, and I'd gladly take one for upstairs and one for downstairs WAY BEFORE I'd buy a 100kg walking steel frame to coexist with my toddlers.

The reason I jumped to humanoids is that I think that the fact that they'll be able to learn tasks by imitating us will make them a lot more versatile and cost-effective overall.

Isn’t current training paradigms requiring massive amounts of data? You aren’t going to get that easily from a handful of luxury robots.

Check out https://www.1x.tech/androids/neo

Their plan is to put droids in real homes and remote control them to create the kinds of training data that can’t be faked in a lab.

They explain how that works here

https://youtu.be/2ccPTpDq05A

Why handful? I would assume that domestic robots like that would sell at least as many units as luxury cars - hundreds of thousands a year. In terms of training, there are a lot of current approaches of learning from videos (of which we have a lot) and from synthetic data, which is getting easier to generate. And if they need more data, I would assume that offering a version of such a robot with telemetry at a large discount would be an easy sell for some. In particular, I'd expect there would be a lot of demand for assistants to elderly people.

I'm not sure I want that. I would probably trade some less versatile robots that are cheaper and do 90% of the work. I can't afford a million dollars - or even 100k dollars for a robot. Until you get into the $1000 range it is hard to see the value. (maybe $20k at the top end - car makers will hate this though as you are directly competing with cars as the only way most people can get that kind of money is to keep their current car longer)

I'm not sure that the cost to build a near-flawless, learn-anything-by-imitation humanoid would be less than the cost of a "laundromatic bot" that weights 5kg and picks up socks and runs on 4 wheels. It just goes against all logic to say that a more general, significantly more capable system would be cheaper than a more-specifically-designed system.

Obviously I'm not arguing that a humanoid will be cheaper than a laundromatic bot, but when I'm considering the full range of domestic tasks, I think it will start to make sense. Just for laundry, that would include: picking up items of all sizes, running the laundry, folding the clothes and then putting them in the closet. Other things I'd want help with include tidying (the defrag in the post), organizing my fridge&pantry and throwing away out of date items, cooking, vacuuming and mopping, and dozens of tasks that I'm sure would come to mind if I think for a bit longer.

Essentially, I want a full domestic helper like Rosey from the Jetsons. I think that technologically we're getting quite close to that, with R&D examples like Atlas[0], Optimus[1], Figure[2], Mentee[3] and Neo[4]. If you're sure you'd be happy with just a laundromatic bot, that's cool, but I think that any non-humanoid approaches to domestic work are pretty much a dead-end, as they are the robotic equivalent of a non-programmable computer.

[0] https://bostondynamics.com/atlas/ [1] https://www.tesla.com/AI (they apparently don't have a dedicated page) [2] https://www.figure.ai/ [3] https://www.menteebot.com/bot/ [4] https://www.1x.tech/androids/neo

Reply on news.ycombinator.com