In this guest post, Greg Shirakyan, a member of the Microsoft Robotics group, discusses why household robots, something that everyone seems to want, are nowhere to be found―and what's needed to be done to get them into our homes.
What Is a Household Robot?
When presented with this question, most people don't ask for clarification. They intuitively know what the question means―even children do. And then, after naming one or two examples that invariably include a robot tidying up a house one way or another, people often stumble.
Children, with their less constrained imagination, dream up Transformers-inspired big brother robots, or ones that would do their homework and play games. My kids were interested in a robot with a magic wand, but let's discuss synthetic Harry Potter a bit later.
First, let's focus on the notion of a household robot. In virtually all cases, people look at their daily lives, their familiar environment, and then try imagining how a robot would fit in. No wonder the robot comes up to be either an unrealistically advanced humanoid or a floor-roving disk about a foot across.
Homework first (no pun intended)
To talk about a household robot, let's first define the term. For the lack of a better definition, we'll go with a device that can intelligently move parts or all of itself around a home and do useful stuff.
The goal of this definition is not to be academically acceptable, but to help us zoom in on a specific set of qualities without implying implementation. The term “intelligently" in this context merely means not being continuously controlled by a person, and not getting stuck. Why do we want it to move at all? Because it's got to have at least some degree of freedom, reach places and interact with the physical world―otherwise it would not be called a robot.
And the more mobile it is, the more "robotic" it feels. Somehow, we humans view autonomously moving objects as "alive" regardless of how intelligent they are. For instance, many Roomba owners give their vacuums a name. But I do not know a single PC owner who bothers to name their desktop, even though a computer is much "smarter" than the randomly roaming vacuum cleaner.
If we visited a department store today and looked for products that fit the above description, we'd end up with devices that can be categorized in two main groups:
- Toys designed to appear intelligent, though usually are quite simple inside.
- Floor cleaners (and maybe lawn mowers); again simple, single-purpose devices.
That's pretty much it. Get it? Robots have been doing a lot of "useful stuff" on other planets for decades. In manufacturing, versatile industrial robots have been a norm long before the PC was born. And the military embraced robots earlier then you might think. So why are they so rare in our homes?
Let's explore some of the reasons, and maybe that will help us get a glimpse on what to expect going forward.
Reason #1: Discriminating Environment
Our homes, appliances, and tools are optimized to be used by people (hardly a surprise) and people only, i.e. if you had a cat with a human IQ of 120, what "useful stuff" could the poor puss do in an unmodified modern apartment? We can't expect it to be able to whip up an omelet or do the dishes, for example (though some cats can flush). The best we can hope for is a friend and companion that we must take care of.
Forget cats. Our homes are tricky places for our own children. How would you like eating with a 2-foot long fork from a dinner table that's taller than you are? Welcome to the world of a 3-year-old. But we love our kids and therefore enhance our living space to make them happy; we give them special chairs, special eating utensils, special toilet seats―you get the idea.
Any robot looking to inhabit a human dwelling and be "useful" in the utilitarian sense of that word faces similar challenges. It would have to feature at least some of the physical and mental characteristics of a grown person in order to perform even mildly complicated tasks.
Just imagine the level of dexterity and contextual awareness a robot would need in order to make a pizza in your kitchen. It would have to find, evaluate, and pre-process raw ingredients using tools and appliances designed for human hands, eyes, ears, noses, and even human psyche. Furniture, lids, peelers, wraps, labels, the sink, dishes, pans, stove … This makes me smile. Not going to happen. What's ironic about this example is that a robot is destined to be a far better cook than the vast majority of humans.
Let's drill down on this one. Think about it―taking raw ingredients and turning them into a descent stew is not rocket science, after all, even a complicated recipe is just an algorithm with a few dozen standard steps. A machine can do this. We just have to design an appropriate environment. A robot-friendly kitchen must have different tools, and different infrastructure for accessing water, fridge, stove, pantry, drain, trash, etc.
A humanoid robot like the ones we see today in laboratories would not be the best choice to operate such a kitchen; in fact it would be awkward for this kind of robot. An "integrated" robot would make more sense, something perhaps resembling some of the semi-stationary industrial machines used today in manufacturing, with manipulators capable of reaching any item in "robot-accessible" storage, a variety of "end of arm" tools that one minute can be whisking eggs, another flipping a pancake.
And then we can expect to have a conversation like this:
Me: Hi Kitchy, we are having six guests today. We'll need a nice three course meal. Indian. Not very spicy, ready to be served by 6:15. A salad too. Make sure it's not very heavy and at least one dish is gluten free.
Robot: Any non-vegetarian dishes?
Me: Sure, add one.
Robot: I am going to make … (lists the menu)
Me: Sounds good, and squeeze some fresh orange juice. Bye.
I can't wait.
Reason #2: Social Un-Intelligence
I am no expert in the rich and fragmented field of AI, but it is pretty clear that we have yet to overcome some serious hurdles in the human-machine interaction domain. There is a tremendous amount of context to human communication, and we are just beginning to make baby steps in modeling the mechanics of it. But why are we talking about intelligence? Do robots have to be intelligent at all? Aren't socially unaware robots useful?
Sure they are. We can call those Blue Collar Robots. Cleaning, cooking, maintenance, security, etc. They take on jobs that people must do, but hate doing, don't want to do, or simply can't. It's safe to bet on a steady increase in the number and variety of such machines. Low hanging fruit that include floor cleaning and mowing are already being tackled by a variety of vendors. A step up in complexity―security and surveillance may come next.
And then there is a White Collar Robot―one that helps us do things we love doing. I am most fascinated by this breed. Satisfying curiosity, being creative, socializing, studying, having fun―those sorts of activities are strictly human and therefore only human-centric robots can actively participate and enhance our experiences. We may change our environment to help blue collar bots excel at their jobs, but we can't modify our own nature, so a white collar robot has a difficult task of grasping the intricacies of human world.
And that's where the cutting edge of science is still not sharp enough. Cute toys and more toys seem to be the only available game in town. There are, of course many research projects, but widespread commercialization of technology has not started yet.
Note, a White Collar Robot alone is neither necessary nor sufficient to bring about a household robot boom. But it could greatly accelerate the process because Reason #1 does not apply here quite as strongly. We would readily forgive a few clumsy moves and lack of utility if those were traits of our delightful companion. We would also be open to modify the environment a bit to make our new family member happy, and that's exactly what the robot race needs to occupy our homes for good. So, an intelligent human-centric robot is quite important. This brings us to Reason #3.
Reason #3: A Reference Design, or Lack Thereof
We take the basic concept of a personal computer (screen + keyboard + "magic box with wires") for granted now, but it was a great discovery. It has proven to be incredibly versatile, and through custom software let us engage in fundamentally different activities: Have fun, do work, learn, and communicate. And the personal computer does all that using conceptually the same hardware configuration and open design, making it easier for new industries (i.e. PC software, peripheral devices) to appear and grow quickly.
There seems to be no similarly versatile household robot design yet. As mentioned before, it is more than likely that we'll see more of specialist bots in the near future, but it raises a question: Who is going to manage them? Eventually it is us, of course, but constantly reprogramming a legion of independent machines to fit our volatile calendar does not scale. We will need a "central command center" of some sort, or … a white collar concierge robot to act as a liaison between its less social hard working brethren and people.
What physical, sensorial, and communicational characteristics would such a device need in order to do "all things robotic" the same way a PC can do “all things digital"? How intelligent will it have to be? Will it need a screen? Manipulators? Will it even be a physical entity or a virtual one living on the home network, or somewhere in the cloud? And will we even call it a robot at that point?
No one knows for sure. The only way to find out is to take a few educated guesses and give it a try. That's exactly what we did at Microsoft with the MARK design (Mobile Autonomous Robot using Kinect). [Editor's Note: Others are exploring similar ideas.]
We believe the MARK bot [image, right] to be a great platform for imagining up household robots. Here is one example. Microsoft Robotics Developer Studio 4 supports MARK in the simulator, and RDS services will be available when our partners start selling the hardware.
You won't have to do soldering, debug firmware, or write low-level control code. All built-in devices, including Kinect, are supported by RDS and can be accessed from high level, C# code. As my colleague Gershon Parent put it, “a robot should be something you program, not something you solder."
When my 8-year-old daughter wished for a robot with a magic wand, and my 5-year-old empathetically agreed, I initially dismissed their silliness. Then I thought a bit more, defined a magic wand as a portable device that takes commands from its holder and changes the state of objects and realized it was me who was being silly.
We all have magic wands, of course. The fact that remote controls are relatively single-purpose is not a function of technological limits or even cost, but of practicality. Sure there are "universals" that can operate just about anything that supports a remote, and can even be programmed for new devices, or "learn," but let's face it: universal remotes are Swiss Army knives―jacks-of-all-trades, and masters of none. They may be cumbersome to set up and they work reasonably well only if all devices are close to each other. As soon as you have more than one room with remotely controllable electronics, a single controller becomes impractical.
A robot does not have a problem with misplacing its "remote" or forgetting which button does what. It can easily interact with hundreds of devices ranging from light switches to other robots in different locations over different protocols and interfaces. Infrared, radio, network, you get the picture. Robots managing robots on your behalf. How do you like that? And it's not even far-fetched.
Magic Wand Robot Edition© would make most real wizards envious. For it is a two-way wand. Not only can it "command" other objects, but "hear" them too. RFID technology is so cheap that it's not hard to imagine that most items in your house can be tagged. “Daddy, where is my blue doll?" “Don't know honey, ask Robbie."
“From the one who has been entrusted with much, much more will be asked."
―Luke 12:48 NIV
aka “ With great power comes great responsibility."
This might be a good time to bring up the issue of security and privacy. Imagine a legion of robots that collectively know everything about your home. They are mobile, interact with your family members, mix your drinks, have many eyes, ears, sensors, actuators, the “magic wand" … They are "plugged in" to the very heart of your personal life. Now picture those robots compromised and remotely controlled by a hijacker.
The devastating effects this might have far supersede the worst that can happen if a virus turned your PC into a zombie. We may not know much about how domestic robots of the future will be designed, but one thing is certain: security, safety, and privacy will not be thrown in as afterthoughts.
“I never predict anything, and I never will."
When will house robots start playing a significant role in our daily lives?
It's a tricky question. First, technology needs to be released. Then, if it survives present day, we adapt to use it, and it adapts back to our new usage patterns, which changes how we interact with it further, and so forth. In the midst of this infinitely recursive stochastic process, our views on what technology "should" look like is constantly redefined and quickly taken for granted in ways that neither consumers, nor technologists or sci-fi writers, can predict.
Sorry to disappoint, the household robot invasion is not happening tomorrow. Not only because the technology is not ready, but because our mentality and our surroundings have not been modified yet to welcome robots en masse. If that sounds like a catch-22, it's because it is.
But do not despair, the initial wave of robots will be designed (or is already designed) to survive the "stone age" environment of the modern man, and that's where it starts. Remember, the first successful electrical appliances [image, right] were powered through a light bulb screw adapter!
So, what is a household robot after all?
It's a class of machines without which we can't imagine our daily lives. The majority of them are common, cheap, and easily recognized; many are hidden, some are specialized, and others are cool and highly coveted. They handle dirty work, inspire our imagination, and do everything in between. Our homes are designed to accommodate them. Indeed, you can't tell where the home ends and robots begin. We depend on them, partner with them, and in return they empower us.
As of December 2011, they are just beginning to be imagined. Too bad I can't patent this description.
Images: Greg Shirakyan/Microsoft