Comment by conductr
5 days ago
Given how incapable my robotic vacuums and lawnmowers have proven to be, even after several years of iterations, I’d almost prefer if it was all teleoperation and it would hopefully unlock a huge amount of additional tasks it could preform. This would essentially let me hire a human housekeeper at a global vs local wage which is very appealing.
Your robotic vacuums and lawnmowers are at a much lower price point that can't afford to run an LBM or computer vision to learn how to do a good job. Because those tasks are too specialized, it's unlikely anyone would pay $8000 for one that did a great job, but the technology is totally there.
The key is when the robot can use all the same tools you can, the generalist robot has a value high enough to pay for this technology.
I’ve paid over $3000 for a lawnmower and over $1000 for a vacuum before. So I’m already halfway there and with a fraction of the benefits as a general device.
I’d need more diligence but would certainly pay $8k on a mower that did a great job consistently. A service cost me $5000 per year so it has a reasonable break even if it’s built well. Robot mowers are pretty hands on, I’d almost rather just diy it.
I feel like this is blowing smoke a bit. If this better tech was possible it would be available, there’s always a market even if it’s at the high end. Do you have anything specific to point to? Any builders/makers that have done this? I feel like this stuff gets touted as “oh if they just had affordable lidar” or whatever then it would solve it, but IRL the variety of homes and yards is so large that it still doesnt generally perform well.
Of course there's a market, it's commercial, not residential: https://scytherobotics.com/
2 replies →
The robot has a two pronged gripper. I wonder what teleoperating that to do complex tasks would be like.
IMO, hard part is likely the lack of haptic feedback, not that it's just a simple gripper. Small and big feedback, so you don't break eggs (not accidentally at least), and also don't pick up a heavy load that makes the robot topple over and give the remote operator motion sickness in the process.
There's a surprising number of labour simulators sold as VR headset games:
https://skyhookgames.com/lawn-mowing-simulator-vr-is-out-now...
https://www.futurlab.co.uk/games/powerwash-simulator-vr
https://gg.deals/game/bartender-vr-simulator/
I’m not sure remote operations needs to be like a VR sim or joystick control of every movement. It might be something more similar to prompts. “Go to front door” “bring in packages” “open packages” “sort trash into pile” <click for item select>”is not trash” “take trash to bin”. Once the house is mapped and a good deal of these routines are learned they can be executed with some fuzzy input to the automation. In my mind this is still full teleoperation
It would be great if the operator could say things like, in above example, “you forgot to close the door, make sure you never leave this door open, even if you have to open it for a task and it’s not explicitly said you should always imply the door needs to be closed after the task. As a general rule, the door should never be open longer than 1 minute longer than it needs to be for task performance” or something similar and this trains the model over time.
I could sweep, mop, vacuum, etc with just my thumb and index finger if needed. But I agree, it’s a limiting design. Probably something they will want to grow beyond eventually. They’ve intentionally handicapped this version, but if they went with a human operated system then maybe they would have seen more potential and given it extra attention. Also, would be interesting if this thing could swap out its “hands” as different jobs may do better with different designs.
It's not a bad business idea, but has dystopian vibes. The human doesn't have to travel to the job site, they don't need to be paid a wage that allows them to exist in an expensive city, and they can watch N screens simultaneously, intervening only when needed. Maybe 1 OOM greater throughput per human-hour. The human teleoperator is also valuable non-public training data, which is part of the learning flywheel. That training data can be sold or kept as a private moat.
How is it any more dystopian than any other offshoring that exist primarily as labor arbitrage?
My in-laws have a full time live in housekeeper that costs $500 per month. And where she’s from, this is a huge opportunity that she went to “maid school” for and many consider excellent compensation. She’s able to send this money home and provides for many family members and has amassed a bit of a real estate empire back home. But, she is absentee. She lives away and rarely gets to visit only about 2-3 weeks a year. So I feel that’s quite sad. They obviously don’t live in the US, because this employment would cost many times more here and probably impossible to even get the proper visas.
Now if this maid was able to live in her village, with her family, and make the same income but perform her tasks through a robot then I think she’d see immediate value in that. So would consumers in America who would love to have housekeepers but can’t afford the local labor rates. Even if you can afford some here, you could get much more for the same budget. A lot of Americans pay this rate for 1-2 cleanings per month, that’s dumb money if you could spend it on something that would sweep up every crumb the day it hit the ground.
> make the same income…
Yeah, that's not the goal of these things.
2 replies →