It Cooks, It Cleans! When Will Robots Be Doing Our Chores?
19:24 Share

With artificial intelligence, creating an ethical foundation isn't just the right thing to do, is crucial to success. Join IBM of the break to hear why from federal binet eris IBM consult into global leader for trustworthy ai.

I want a future with no housework, no doing dishes or laundry, no making beds or tiding up, a future where I can relax while technology does all the work. Some of us have been waiting since one thousand nine hundred and sixty two for this kind of robot to make IT into our homes. Like the one we saw on T. V. Is jane Jackson overwhelmed with the estimated ty?

It's just that the housework .

gets me down. Rosie cook SHE clean, and he still finds time to play ball with l. roy. Rosy is the ideal made.

Despite the machine made promised by the seton's cartoon, dynamic domestic robots are currently not a widespread fixture and homes only the rumba, the autonomous floor cleaner, has been a real commercial success. According to irobot, approximately twelve million rumbles have been sold since two thousand, one for charly camp. The invention and success of the ruber was a revelation that was inspiring.

But that was happening at that. This is now, this is just the first one, and now is going to be this explosion of robots doing useful things in homes.

Charly camp has been making robots his whole career and taught robotics at George attack for sixteen years. He started out doing his P. H. G with rod Brooks, who was the founder and former chief technology officer at eye robot. Ee robot makes the robot.

Twenty years later, I would have expected robots and homes be doing more things.

Currently, charly camp is the chief technology officer at hello robot, which is piloting a home aid robot called stretch three that does everything from picking things up off the floor to folding laundry. Stretch three is primarily for research, education and for corporate R N. D. But IT is available to the public on the website for around twenty .

five thousand dollars. But what has kept me going is I have seen, in research context, great potential for the technology to benefit people in their everyday lives.

From the wall street journal, this is the future of everything. I'm charger garden burk. Today we're asking how far are we from having our own rosy, the robots in our homes stay with us.

How do you start to lay the foundation for responsible ai in your organization? Here's feature bonus. IBM consultants, global leader for trust for the .

A I IT starts with asking the question, what is the kind of relationship that we ultimately want to have with A I? The purpose of A I is not meant to human being that has meant to augment human intelligence. Students, you have a glimmer in your eye about how you are thinking you might win A U.

A. I. Then asking the questions like what would be required in order to earn people's trust in such a model.

Charlie camp has been building robots in teaching other people to build robots for over two decades. Seven years ago, he co founded hello robot in pursuit of a robot that could help people in the home. I started by asking him what actually makes very, truly helpful domestic robot.

so examples of things that you might want a robot to do, which I think are in the near term possible, pick up toys maybe from the floor, but the mini box. Go retrieve an object. Get a drink for you.

Find a displays. Object taking care pets, taking care plants for people, everybody, people. There are task which just seem so easy, like, how could I be hard for robot to do that? But something of dishes, dishes are actually surprisingly hard to do. And my expectation is that is going to be a task that happens farther out.

Why don't we have that robot yet? I mean, tesla and amazon are already using humanoid robots in their manufacturing, robots that have human shaped spaces, arms, torsos, their A I powered. So IT feels like the jump to domestic robots. Maybe even humanoid domestic robots shouldn't be that bigger. I like, why do we have domestic robots?

The first thing to realize is that homes are one of the most chAllenging places for robots to be doing things. And one of the biggest reasons for that is because people are in homes and not just professional workers who could be in a factory working beside a humanoid. You can have children, all the adults, people disabilities.

There could be a baby walking around. We love our pets. We don't want our pets to to get heard. And those agents in our environment, they are not professionals, they're not trained. They're hard to predict. Robots have historically done well in structured environments where there isn't a lot of very ability where IT is very predictable.

So in order to get this this dream robot that will do my laundry and wash your dishes, can you break down what are .

all the capabilities .

that such a robot would need to have?

Let's look at a task that most people would consider to be simple. I told robot I would like a glass of water. Now, of course, that has to figure out what I mean by that.

Everything has to figure out where to go. So IT starts to go there. And then all the sudden a cat zones by IT has to respond reasonable.

I can't just run the cat over.

That's right. Yeah, it's like IT has to respond to this dynamic environment. But let's say, IT does IT handles at IT IT safely stops.

IT senses what appropriate, IT keeps moving along and IT gets to the general is entity is in the kitchen. And then IT realized weight. I have to open the cabinet.

Well, now IT has to figure out on how to do that, and then has to grab the object. And when grabbing the object, that's something that people think of as simple. But the details really matter.

Like one cup is different from another cup. A lot of cups are transparent. Some cups are glass. So, oh my god, I don't break that.

What does that manipulator look like? That's even grabbing the glass.

right? But that's a great question. Doesn't need to be a full human hand. Probably not. Maybe it's a simple gripper and maybe that's enough for this task, but IT has to be able to to figure out where to put its fingers on that object, squaze at appropriate pressure, lifted up, pull IT out without bumping in anything. And and now it's finally, it's got step one, I gotta ask. And each one of these task, there is an opportunity for failure, and there is an opportunity for bad failure. Let's say what it's trying to get the glass bumps and other glass and IT goes on to the .

floor or that a tragedy. How far are we on each of these abilities separately before we go, you know, putting them all together in one bot? Let's start with grips or manipulators.

One thing that has been an area of progress has been the robots hands. And one thing that has been more common now is for instead of the robot having this very rigid, solid hand that is carefully control, there are now where they call soft grippers, where it's flexible and IT can bump into things without causing damage, and IT can grasp things without that real detailed control are sort of an intelligence in the mechanics. And on the one of that was bio inspired, because if you look at the ways that animals grasp things, it's not just the the intelligence, if you will, it's also a matter of the way their hands are made or their cripples are made.

Okay, what about the technology that allows those robots to sense as how does the robot know where IT is within an environment and and also what's in the environment?

When the room bus first came out, one of their key innovations was that they had this giant, basically half of the robot was a touch sensor. And so I could go, and if IT hit an object, IT could turn around and keep moving. And that that actually worked well as a robusta teach more recently.

Like if you see robots now, they often will have a three d sensor. So they have a sensor which is not unlike our two eyes. We know how far way things are. And that sort of three d sensing technology that can enable robots to Better detect obstacles, detect a hazard algorithm. Ally, it's interesting. I think the way robots are succeeding today is is a mix of older approaches, which are sort of older A I using planning and and some sorts of probabilistic models sometimes, and newer AI, where you have neural networks and deep learning that can enhance the performance, especially perception.

interesting. I wonder if that A I too is helping us make a leap in how we teach robots .

to do things in research right now, are learning from demonstration. So where you show the robot how to do something is is a very popular approach for teaching robots. And those methods right now are benefiting from a lot of the progress that has been made in other domains of artificial intelligence, like large language models.

One of the main chAllenges for robots is that or as these large language models have the benefit of just inconceivable amount s of writing that they're able to use, or if it's a general model for art in inconsiderable counts of artistic photos and artistic s works that they can learn from robots, there's a lotless information for robots to learn from, because just not common for people to take photos that look like what a robot would see as it's driving around at home, because it's really boring. We all take IT for grand, and why are we gonna capture lots of video of us, you know, oh, this is what the floor looked like as I was going to get a glass of water. And similarly, there's in that much data that the robots can benefit from for something like tactile sensing or or something like grasping an object. It's sort of falls into the category of what would have been called common sense.

So then is the way to teach robots spite, breaking down every single step in an action, and then teaching IT that way? Or are there other ways to take a robot?

There are a few ways that people are trying to tackle that. One is to use physics, space simulations to generate training data. So it's a virtual world.

You can get lots of data in a virtual world without putting people at risk. The usual chAllenge is that in the virtual world always differs a bit from the real world. Another thing is to have people Operate the robots themselves, and then learn from how the people do things. So person kind of inhabits the body of the robot, does things with that robot, and then the robot can learn from that. And and that has the advantage that is in the real world, but is also very time consuming.

Getting all of those things into one robot is a little bit chAllenging, but there are domestic robots in development, including stretch three more on those after the break.

Charly tell me about stretch three. What does that look like?

Stretch three is an attempt to find the sweet spot where it's big enough to do useful things in homes, but still light weight and small enough to be safe and effective and easy to use. The way we think about IT is that IT is robotic, and you take the human form, you deconstruct, and you put that back together. So first of all, stretch doesn't have legs.

Stretch as wheels, the wife of stretches sort of less than human hit with, so that I can navigate through those narrow channels we find in our homes. And there is an ARM that moves along the rail down so I can pick things up off the floor and up, so that I can pick things up off of high countertops, no ARM as a telescoping ARM. And that sort of a long gates out grabs something. And then the react, and that that's one of the reasons for the word, the named strategy.

So there are are other companies that are working towards domestic robots, one x technologies and attract ic. To some extent, those companies are focused on making human shaped robots. And I know stretch three is not humanoid, but tell me some prose to having domestic robots be humanoid.

Assuming a sufficiently capable body, then IT can theoretically do the things that people do, by the way, having a body that really has the cake village of human body is is not trivial. But if you had that, then you could say, look, now the problem is software pro would be there could be some benefit to having a body that's just like you. So when you're trying to teach at something, it's easier to do IT because it's it's .

what you're familiar with. And of course, I could climb stairs.

That's true. Stairs, stairs for legged robots, that is one of their big advantages is that they can traverse stairs.

Why not have your domestic robot be shaped like a human?

Because the the human form, although it's wonderful, is also very complex. So so on one side, you have lots of motors. And every motor is another thing that can go wrong, and it's another part of the system that can increase its cost. Another thing is that alleged human oye robot that is really modelled after people if IT has like blue screen of death, it's gonna fall. If IT runs out of batteries, unexpected, which sometimes happens with my devices.

it's going to fall and maybe on cat.

Ah exactly and that that's a series a serious consideration.

Well, let's go back to stretch three. So what has the threats three community managed to get this robot to do?

Everything from watering plants to cleaning surfaces, putting toys away, uh, retrieving particular objects, finding misplaced objects, playing games, help with laundry, checking on thing, sort of .

inspecting. Are most of those autonomous or combo.

There have been examples, words and autonomous, but is also the case that in some applications, people would like to directly control. The a clear examples of that is when IT empowers someone to do something they might not otherwise be able to do, such as if you have A A mobility .

impact OK time for your crayon ball, how soon do you think we might see, oh, got something like stretch three, really at scale in homes. And why do distribution.

how long IT takes depends in part on how much society wants IT and how much society invest in making this happen. I I think five years, it's very paul, you definitely could have robots. They are not unlike stretch, doing useful things in millions of homes within five years. But I I was like if IT doesn't happen in ten years, i'm going to be awfully I don't have any doubt that IT will happen, that there will be robots and homes helping us out and doing things we didn't even expect. But when IT actually takes off, time will tell.

The future of everything is a production of the wall street journal. This episode was produced by me charlock garden burg, mixing in sound design by Jessica fenton. Like the show, tell your friends and leave us a five star review on your favorite platform. Thanks for listening.

Earlier, we discuss what responsible A I looks like in practice. Here's fator point, and dear is from IBM consulting again on why that begins with data.

My favorite definition of the word date up. It's an architect of big man experience. A I is like a mirror that reflects our biases back towards us, but we have to be brave enough and introspective enough to look into the mirror.

Does this reflection actually line to my organization values? If IT allies betrays parents about why did you pick the data that you did? If IT doesn't align that when you know you need to change .

your entire approach, learn more about IBM artificial intelligence consulting services and IBM dot com slash consulting.