Robotic Achievement

[thriftyandboringinohio]

For years now I've bantered with friends about my own twist on the Turing Test. https://en.wikipedia.org/wiki/Turing_test
In my twist you hand a robot a basket of laundry
If the robot can separate the pieces, turn them right-side-out, and fold them, it passes the Laundry Test.
Academics at Imperial College London have apparently decided to tackle that problem, and have gotten the last few steps done.
Behold




When you think about the problem in detail, it's pretty tricky.

[DSpencer]

Re: Robotic Achievement

Very interesting. Are you aware the Yaskawa Motoman company is headquartered in Miamisburg, Ohio?

For whatever reason I've always found folding laundry to be one of the most boring tasks in my life. I would love a robot that could do it for me. I'm guessing this isn't quite cost effective yet.

[dcarrigg]

Re: Robotic Achievement

Very interesting. Are you aware the Yaskawa Motoman company is headquartered in Miamisburg, Ohio?

For whatever reason I've always found folding laundry to be one of the most boring tasks in my life. I would love a robot that could do it for me. I'm guessing this isn't quite cost effective yet.

Around these parts for $30/mo, basically the same price of trash pickup, you can leave the whole family's laundry weekly and have it picked up and returned washed, dried, pressed, and folded same-day.

It's even cheaper if you bother to drive it down to them yourself. Of course, the price goes up considerably if you have a few things in the mix you want dry-cleaned (dress shits/dresses/wool). But then you're also saving on all the water and gas and everything to run the washer/dryer, not to mention the time and annoyance of it all.

If you really hate doing it, there are some cheap ways around it. No robots required.

But I'd advise you to stay away from cleenly or the any of the apps or any cleaners that pays for advertising. Just find the local immigrant-owned laund-o-rama, and odds are damn good they'll offer the service, offer it for less than half the price, and do a better job too.

For $400 something per year versus the $200 something per year you'd spend running the washer/dryer and buying detergent etc. anyways, it's not a terrible chunk of expense to add into your life.

[akkibaba]

Re: Robotic Achievement

Cool. Dishwashing Robot takes priority over folding clothes.

[DSpencer]

Re: Robotic Achievement

Around these parts for $30/mo, basically the same price of trash pickup, you can leave the whole family's laundry weekly and have it picked up and returned washed, dried, pressed, and folded same-day.

It's even cheaper if you bother to drive it down to them yourself. Of course, the price goes up considerably if you have a few things in the mix you want dry-cleaned (dress shits/dresses/wool). But then you're also saving on all the water and gas and everything to run the washer/dryer, not to mention the time and annoyance of it all.

If you really hate doing it, there are some cheap ways around it. No robots required.

But I'd advise you to stay away from cleenly or the any of the apps or any cleaners that pays for advertising. Just find the local immigrant-owned laund-o-rama, and odds are damn good they'll offer the service, offer it for less than half the price, and do a better job too.

For $400 something per year versus the $200 something per year you'd spend running the washer/dryer and buying detergent etc. anyways, it's not a terrible chunk of expense to add into your life.

You tell me this after I already got married?Just kidding of course. I have heard of those services but I'm not sure they are offered in my area. That is much cheaper than I expected. I figured it would be closer to hotel rates where you're almost better off buying new clothes.

[DSpencer]

Re: Robotic Achievement

Cool. Dishwashing Robot takes priority over folding clothes.

I always wonder what people in the distant past would think if they could hear us now. We already have machines to wash our clothes and dishes but the hassle of putting them in and out is still too much to bear. Of course, I'm never convinced that my dishwasher works well enough to not pre-wash everything. I often wonder if it's even worth having. I know people that swear their dishwasher can handle everything with no pre-washing, but I remain skeptical.

[thriftyandboringinohio]

Re: Robotic Achievement

I find problems like folding laundry and washing dishes fascinating from a technical point of view.
They are so simple for a real human, any small child can do them.
But to automate them with machinery and AI is nearly unsolvable.
They are good landmarks for how technology is advancing (or not).

I recall many years ago a senator (perhaps William Proxmire?) wanted to prove a new weapon was a waste of money.
It was a new heat seeking anti-armor weapon.
The senator has his staff sneak out onto the test range before the big demonstration with tin cans and charcoal, lighting little burners around the range.
The test failed spectacularly, the missiles targeted the charcoal burners rather than tanks or trucks.
The senator made a big show of it all during the hearings that followed.
The defense contractor complained that the trick was not fair and the program should proceed.

The senator disagreed. "All of our potential enemies have access to tin cans and charcoal".

The program was cancelled.

[Techdread]

Re: Robotic Achievement

I always wonder what people in the distant past would think if they could hear us now.

Distant people in the past used to use slaves or women to do these tasks.

We already have machines to wash our clothes and dishes but the hassle of putting them in and out is still too much to bear. Of course, I'm never convinced that my dishwasher works well enough to not pre-wash everything. I often wonder if it's even worth having. I know people that swear their dishwasher can handle everything with no pre-washing, but I remain skeptical.

Yeah, we were skeptical about dish washers too, until my wife done some research and decided to use one. They have vastly improved and can wash burnt pans no problem.

[dcarrigg]

Re: Robotic Achievement

I find problems like folding laundry and washing dishes fascinating from a technical point of view.
They are so simple for a real human, any small child can do them.
But to automate them with machinery and AI is nearly unsolvable.
They are good landmarks for how technology is advancing (or not).

I totally agree. And it's not just what humans can do. My dog at 6 months old could lead me back home from a mile away in just a few minutes with no mistakes over varied and complicated terrain, navigating sidewalks, street crossings, bike paths, and woods, and using the bathroom only in permissible areas, knowing which path to walk on, with nearly no repetition. Soon, by 7 or 8 months, he could generalize commands. "Go Inside" meant go to the gate and wait for me to open it. It began to work everywhere. New gates at new parks he'd never seen before...didn't matter. This wasn't just rote memorization, it was the application of a concept...the order "go to gate" worked just as well for chain link gates, glass sliding doors, wooden gates on hinges, whatever. Put a whole new material gate in front of him, and he'll just figure it out. Stairs were tougher. It took a lot of coaxing for downstairs to work. But that was like a switch flipped. And all of the sudden it was a two second race down and up. Easy.

Better still was watching him develop ways of communicating back to you. He figured out early on that I like my sodastream and I like to sleep with a bottle of seltzer on the windowsill next to my bed. Well, he figured out he could come up and clank the cap on the window with his nose to wake me up if he really had to go to the bathroom in the morning. Thankfully, by about 11 months, he could hold it a lot longer. But back then, you'd only get 6 hours or so. So he just made his own alarm. I mean, compared to a 6-8 month person, this sucker could point squirrels, find his way home from a ways away, respond to and even give commands of his own. And he's a dog. They're dumb as bricks. But goddamn they all look like Einstein next to a Roomba.

There's an old story a couple towns down about an old timer who ran out of liquor and decided he'd ride his horse down to the inn/bar on the coast. He got in there, maybe got served one or two, but then got shut off for being too far gone. Well, he demanded and wouldn't take no for an answer, so they called the cops down. Amidst the commotion, he hopped on the horse, and ordered it home. Passed out on it on the way. But the horse crossed US-1, then led the police through the woods on foot pursuit, before getting the old timer home and waking him up. Police picked him up there, and he didn't know how he'd arrived. The original 'self-driving car.' No LIDAR required.

[lakedaemonian]

Re: Robotic Achievement

I totally agree. And it's not just what humans can do. My dog at 6 months old could lead me back home from a mile away in just a few minutes with no mistakes over varied and complicated terrain, navigating sidewalks, street crossings, bike paths, and woods, and using the bathroom only in permissible areas, knowing which path to walk on, with nearly no repetition. Soon, by 7 or 8 months, he could generalize commands. "Go Inside" meant go to the gate and wait for me to open it. It began to work everywhere. New gates at new parks he'd never seen before...didn't matter. This wasn't just rote memorization, it was the application of a concept...the order "go to gate" worked just as well for chain link gates, glass sliding doors, wooden gates on hinges, whatever. Put a whole new material gate in front of him, and he'll just figure it out. Stairs were tougher. It took a lot of coaxing for downstairs to work. But that was like a switch flipped. And all of the sudden it was a two second race down and up. Easy.

Better still was watching him develop ways of communicating back to you. He figured out early on that I like my sodastream and I like to sleep with a bottle of seltzer on the windowsill next to my bed. Well, he figured out he could come up and clank the cap on the window with his nose to wake me up if he really had to go to the bathroom in the morning. Thankfully, by about 11 months, he could hold it a lot longer. But back then, you'd only get 6 hours or so. So he just made his own alarm. I mean, compared to a 6-8 month person, this sucker could point squirrels, find his way home from a ways away, respond to and even give commands of his own. And he's a dog. They're dumb as bricks. But goddamn they all look like Einstein next to a Roomba.

There's an old story a couple towns down about an old timer who ran out of liquor and decided he'd ride his horse down to the inn/bar on the coast. He got in there, maybe got served one or two, but then got shut off for being too far gone. Well, he demanded and wouldn't take no for an answer, so they called the cops down. Amidst the commotion, he hopped on the horse, and ordered it home. Passed out on it on the way. But the horse crossed US-1, then led the police through the woods on foot pursuit, before getting the old timer home and waking him up. Police picked him up there, and he didn't know how he'd arrived. The original 'self-driving car.' No LIDAR required.

Agreed.

But what about when Roombas all learn from each other in an IOT 24/7 fat pipes world?

The way the driverless car relentless improvement was explained to me(after my concerns about deconfliction in wad space as opposed to the far simpler 3D airspace deconfliction and congestion problem) is that twice a day every vehicle in the network learns from each other.

What one learns, they all learn.

Exponetial machine learning as opposed to evolutionary biological learning.

Still a way to go, but that's how I'm viewing it now.

Give it 10 years, just half the time since the internet went mainstream 20 years ago.

[dcarrigg]

Re: Robotic Achievement

Agreed.

But what about when Roombas all learn from each other in an IOT 24/7 fat pipes world?

The way the driverless car relentless improvement was explained to me(after my concerns about deconfliction in wad space as opposed to the far simpler 3D airspace deconfliction and congestion problem) is that twice a day every vehicle in the network learns from each other.

What one learns, they all learn.

Exponetial machine learning as opposed to evolutionary biological learning.

Still a way to go, but that's how I'm viewing it now.

Give it 10 years, just half the time since the internet went mainstream 20 years ago.

The question is really "How much can roombas practically learn from each other?" Do they pick up bad habits/information? I mean, they're limited by their physical abilities, which are pretty pathetic. But they're also limited in that they cannot and will never be able to see a completely new situation, size it up, and just deal with it. There will have to be lots of failure before the first success...not unlike evolutionary biology.

The problem with machine learning is that it's all probabilities based on known knowns and unknowns. Those pesky unknown unknowns I'd trust a bio-mouse to figure out before the best algorithm you've got going in a robo-mouse...even in the year 2100.

[thriftyandboringinohio]

Re: Robotic Achievement

Agreed.

But what about when Roombas all learn from each other in an IOT 24/7 fat pipes world?

The way the driverless car relentless improvement was explained to me(after my concerns about deconfliction in wad space as opposed to the far simpler 3D airspace deconfliction and congestion problem) is that twice a day every vehicle in the network learns from each other.

What one learns, they all learn.

Exponetial machine learning as opposed to evolutionary biological learning.

Still a way to go, but that's how I'm viewing it now.

Give it 10 years, just half the time since the internet went mainstream 20 years ago.

Thanks lake.
It is a very promising approach.
I worry that such an information channel, like every other digital channel, will immediately get attacked by crooks and vandals.
The imagination runs wild when you think of malicious actors controlling self-driving cars with hourly updates.
From annoying unwanted stops at retail outlets to mass carnage on highways "just for the lulz".

[lakedaemonian]

Re: Robotic Achievement

The question is really "How much can roombas practically learn from each other?" Do they pick up bad habits/information? I mean, they're limited by their physical abilities, which are pretty pathetic. But they're also limited in that they cannot and will never be able to see a completely new situation, size it up, and just deal with it. There will have to be lots of failure before the first success...not unlike evolutionary biology.

The problem with machine learning is that it's all probabilities based on known knowns and unknowns. Those pesky unknown unknowns I'd trust a bio-mouse to figure out before the best algorithm you've got going in a robo-mouse...even in the year 2100.

I certainly agree with the lots of mistakes will occur with robotic machine learning, but I do think the networked machine learning will inevitably allow far faster development thru shared learning.

I'm reminded of a post EJ made a few years back about the potentially unexpected disruptions to labour markets.

Not necessarily impacting high variability unskilled labour, but more low variability complex highly skilled labour.

Less ditch digger disruption, more ophthalmology surgeon disruption perhaps.

The presentation by a classmate running Google's autonomous car project left me both very impressed as well as slightly disappointed in progress.

A plastic bag floating in a micro wind eddy can cause confusion for a vehicle(disappointing), but a vehicle has been shown to be able to react with far greater speed in mitigating other vehicles on a dangerous intersecting course.

I think we are both farther head and behind than most of us think.

[lakedaemonian]

Re: Robotic Achievement

Thanks lake.
It is a very promising approach.
I worry that such an information channel, like every other digital channel, will immediately get attacked by crooks and vandals.
The imagination runs wild when you think of malicious actors controlling self-driving cars with hourly updates.
From annoying unwanted stops at retail outlets to mass carnage on highways "just for the lulz".

It's funny you mention that.

I just spent some time with an exceptionally capable network security subject matter expert.

We spent some time discussing the vastly overwhelming economic crime aspect of computer security.

I was far more interested in the far smaller, but potentially far more dangerous geopolitical/warfighting aspects of computer security.

Some of the items discussed were along the lines of what you mentioned, autonomous or human driven vehicles that are killed at traffic choke points based on GPS locations and time windows.

The good old news is that requires a fairly broad and deep skill set as well as considerable operational planning and management. Likely the domain of only a small but growing number of sovereign states, or possibly a few hacker networks(at this stage highly unlikely to shift from economic crime to geopolitical and associated unwanted attention).

Personally, what I'm excited to see is how technology can be applied to assist cities in better unclogging congestion.

Policing and ambulance fleet management has improved thru best maps and big data analysis, I would think unclogging mega cities of the future will be both incredibly interesting and challenging work bringing instantaneous and positive economic impact to the immediate community.

[thriftyandboringinohio]

Re: Robotic Achievement

Very cool, lake.