What's the point of robot pets?
This page created August 2002. Last modified 03-Dec-2011.
Artificial pets. What a brilliant idea!
OK, I admit that the artificial animals we've managed to make so far aren't quite up to J.F. Sebastian standards. They're clumsy battery-powered sci-fi-robot-looking things, and the best one commercially available is the Sony Aibo. The second generation Aibo has been out for a while now, but it's still no more nimble than a puppy in a suit of lead armour.
There are other "synthetic pets". The Omron Tama robot cat, for instance, which is apparently meant for pet therapy purposes, not as a mass market item. And there's NEC's not-for-sale R100 and PaPeRo, which keep most of their brain in a PC and connect to it via a wireless network interface.
There are lots of cheaper toy-shop versions, too. Tiger Electronics, makers of the Furby, now also make the questionably named but cheap ($US30) Poo-Chi dog. There are many other low cost, less capable Aibo clones.
For the tinkerers, there are various little wheeled sensor platforms that look nothing like any living thing. Like this one, for instance. And Lego's doing very nicely with their startlingly capable Mindstorms system.
None of these, even Aibo, have anything remotely approaching the complexity of the synthetic kitten Robokoneko. But that's an artificial intelligence experiment, not a product, and it exists only as a computer simulation, so far.
Not many people have played with robo-pets yet, because so far they're all either cheap and disappointing, or very expensive. The current model Aibo's down to a mere $US1300 now.
Which'll buy you a puppy from the RSPCA, all medical care up to and including desexing, and well over a tonne of tinned food.
But the Mark 2 Aibo's a bargain. The original Aibo was $US2500!
Because not many people have played with them, most people's opinion about robot pets is along the lines of "oh, what a sad thing, imagine trying to play with a robot dog instead of a real one". The robot dogs and cats and... other things... are a very long way indeed from being self-aware. Really, they're not even on the road to consciousness, any more than aeroplanes are on the way to being birds.
And current robo-pets indeed do a very poor job of pretending to be a real dog. If you want something that actually does the sorts of things that a dog will do, an Aibo is an appallingly bad substitute.
But robot pets, so far, are not meant to be substitutes for real pets in the general sense. Nobody, including the manufacturers, is pretending that they are. What they are, are interesting toys in the shape of an animal. Which is what leads to all of the emotional confusion.
The abovementioned NEC robots don't have this problem. They're not pretending to be anything alive - they're just cute little perambulating alien-things. Your own personal R2-D2, except about a millionth as cool.
Some R100 descendant or other will make it to market soon enough, and will probably be received with the same initial enthusiasm and then boredom as all "personal robots" so far, back to the old Heathkit Hero, about which you can find more information than you could probably ever desire here.
The very best of the personal robots you can get these days are still only slightly useful, really, for normally-abled owners. Spy-Cye can act as a roving Net-connected security camera, trundle around carrying letters or drinks, and drag a vacuum cleaner. But for $US995 you'd want it to at least be able to mix a martini, as well as slowly deliver it to somewhere.
Friendly Robotics make a $US525 robot lawn mower that uses a buried wire around your lawn to stop it from wandering off down the street. It can, genuinely, slowly but quite effectively mow most lawns. But it isn't supposed to be left unsupervised, and isn't rain-proof.
So, in summation, don't hold your breath for a Cylon butler. Independent robots that do something useful are a very difficult problem indeed.
All of the world's working robots operate in very tightly controlled environments, like automotive assembly lines, and have nothing that can fairly be called "brainpower" at all.
But robo-pets don't have to be useful. Sure, it'd be cool if they could fetch the newspaper from the verandah and scare off intruders. But most cats don't do anything useful, and they're popular enough.
Think about the problem of making a robot companion-thing, as opposed to a robot butler. Think about the practical pointlessness of the cat. And also think about people's ability to love all sorts of things, alive and dead.
Nobody's too likely to anthropomorphise their automatic mower, but look at the relationships people have with cars, computers and teddy bears. Human beings tend to perceive consciousness and personality in things which do not actually have them. If a gadget sets out to seem alive, lots of people will treat it that way.
Look at the low cost keyring "computer pet" Tamagotchi, for heaven's sake. As is cogently explained here, Tamagotchi have only four variables that determine their behaviour. The Tamagotchi game - and it is just an elementary hand-held game - is only barely more complex than tic-tac-toe.
And yet, when the Tamagotchi fad was in full swing, any number of people swore the things had distinct personalities.
Humans are good at finding complex patterns, even where none really exist.
As the owner of Thomas, a cat whose behaviour could be pretty easily flowcharted (with a large number of arrows leading to the boxes marked "keep armchair warm" and "inhale large bowl of fish"), I have more than once found myself wondering how much of the personality I see in his roughly spherical fuzzy form is of my own invention.
Brain size maps poorly to intelligence, and intelligence itself is a massively slippery concept, but my beloved feline's imposing almost-ten-kilogram body (a testament to the awesome power of Whiskas) is steered by a roughly 30 gram brain, which, to be frank, doesn't seem to be in any danger of failure due to over-use. Tom practically never does anything I don't expect him to do.
As I type this, a flowcharted version of Tom would be using its randomiser, weighted by its hunger index, to determine whether it would stay curled up on the couch or come and whinge for food. The chance of him doing anything else at all for the next six hours approaches zero.
I don't know. Maybe the many tiny nuances of Tom's behaviour could not be economically simulated, and would by their absence irreparably damage the illusion of life. Maybe deleting responsibility from the owner-pet relationship would make it hugely less meaningful. And, of course, making a robo-cat that could actually eat and excrete would be gigantically challenging.
But making one that just recharged itself by curling up in a special cat basket that came with the robot wouldn't be nearly as difficult, and I can't help but think that it really wouldn't be hard, if you had the technology to make a plausible artificial cat-body and connect it to sensors, to give it a surprisingly simple and thoroughly non-self-aware stimulus-response programmed Aibo-style control system that'd let it behave like a fat, lazy cat.
An active, outdoorsy sim-cat that could interact well with real animals would be far more problematic. But most of the persona of a peaceful lap-warming homebody like Tom might well be simulated adequately by an unrevolutionary expert system. The marketing people would say it had emotions, but they're saying that already about Aibo and company, and that lot make flatworms look like Einstein.
The robo-pets, that is. Not the marketing people. I'm sure every one of the marketroids has a towering intellect.
Ask a computer game programmer about computer-opponent algorithms that produce the appearance of intelligent opposition, while not using much CPU time. Provided the environment the "artificial intelligence" has to operate in is simple enough, it can give you a very realistic run for your money. Behind the scenes it may be cheating like mad, but it is the programmer's art to hide that.
An indoor robo-fuzzy-thing programmed for not much more than cheerful affection is not a monstrous coding task. It's the body that's the problem, not the mind. Making a legged robot that can jump up on a couch, and that doesn't break down all the time, and that doesn't cost as much as a fighter jet, is tricky.
Which is not to say that the body problem is insuperable.
With advances in conventional low-power processors - never mind fantasy ultra-complex brain-analogue neural network magic-chips - and in sensor and fast servo technology, a reasonably coordinated and agile robo-pet with far more possible behaviour patterns than an Aibo, yet a lower price tag, isn't a ridiculous thing to expect to see on the shelves in several years' time.
Advances in materials science permitting stronger and more elegant designs - with things like nitinol wire for muscles, instead of whining servo-motors - and existing technologies like solid state piezoelectric "gyroscopes", make the notion of a robo-pet that can judge the height of a table and the properties of the floor, then neatly leap from the latter to the former, less ridiculous than it may seem to someone observing Aibo's clumsy bumblings.
Mind you, a sim-creature that can just scramble, rather than leap, up onto a couch would do for most purposes.
Limitations like not being able to deal with the outdoors would actually be an advantage for a hypothetical future robo-cat. The whole idea of a true artificial pet (as opposed to a mere animal-shaped gadget), after all, is that it's a pet for people who can't have, or don't want, a real one. And one big reason why people don't have real pets is that they live somewhere, like a little home unit, in which mammalian pets are forbidden. Such places often have no outdoors that a real cat could safely access.
Plus, robo-pets can do computery things that real animals can't. Recording and playing back sound and video, playing games ("Rook to Q4, checkmate. Woof!"), counting how many times you open the fridge and telling you when your snack-free diet seems to have failed.
Now, I freely admit that I'd still be sad for someone who was stuck with a robo-thing that sat on their lap and purred realistically, but which only did it because it was programmed to, and was in reality no more self-aware than the remote personal computer which, most likely, was running its wireless-controlled body. A big part of the pleasure of patting a cat or dog is that you're genuinely making something else happy.
But if someone owns a robo-pet, and loves it, and gets joy from patting it, then I doubt that person's likely to care whether I'm sad for them or not. If you think your robo-pet's alive, it doesn't matter to you whether it really is or not.
Many people already get very attached to all sorts of creatures which are, arguably, only very dimly self-aware, or not self-aware at all. Your average insect pet, with a few hundreds of thousands of neurons at most in its whole body, for instance. Popular in Japan, but not exactly a roiling stew of complex emotions.
Remember, people fall in love with cars and teddy bears and houseplants. And most of those people aren't hard-core animists who believe there's a living spirit in everything, even their oil-bleeding English sports car. OK, Jaguar owners, I know that spirit possession is actually a pretty parsimonious explanation for the behaviour of your chosen mode of transport. But you know what I mean.
The if-you-love-a-machine-then-you're-not-truly-human crowd have a serious hole in their argument. Lots of people love machines already, and at least half-believe that the machines have personalities. These people are not, on the whole, soulless dead-eyed monsters. Most of them, of course, don't love machines that look like animals. Yet.
Importantly, robo-pet buyers might be the sort of people that, otherwise, might let their cat out to get squished on a busy road, or to eat the native wildlife. Or might simply abandon it when they got bored with it. Then I'd be very glad indeed that they had, instead of a real animal, something that can't suffer, any more than it can really be happy.
OK, this person could have donated the money to the local animal shelter and done more good, but I doubt that a robo-pet Aibo buyer would, if a robo-pet wasn't available, be very likely to be charitable instead.
And if later model robo-pets turn out to appeal to the kinds of people who'd otherwise be irresponsible real pet owners, with un-desexed animals that spit out a new generation of unwanted babies every ten minutes, then I think the net effect of even a high-priced robo-pet to stop them from polluting the world with yet more poor surplus animals might well be very positive indeed.
Perhaps, soon, having animals for companionship will be like using animals for transport. A bit peculiar. Inefficient, on the whole. More trouble than it's worth, you know. OK, I doubt the domestic dog and cat are at all likely to become endangered species. But perhaps artificial pets will be able to deliver a startlingly realistic dose of love to people who otherwise would have had to make do with a goldfish, at best, and will also give less responsible folk a pet that doesn't mind at all if you just stuff it in a cupboard when it bores you.
If people can derive joy from a worm-brained fuzzball that beetles around on four plush-furred balloon tyres and rubs up against warm things, then good luck to them, I say.
Heck, I'd buy one.
Tom needs the exercise.
