Posted by
Hemos
on from the fun-with-machines dept.
Cy Guy writes "ZDNET had an article on CYE, a personal robot that will vacuum, collect dirty dishes and serve cocktails. " This is exactly what Rob needs-and if it can serve Bushmills or Jamison up, I'll be happy as a clam.
No Blue Screen of Death?
by
daviddennis
·
· Score: 2
Humph. Windows?
Is there a Linux version?
I'll bet a Beowulf cluster of those would be... well, pretty dangerous, actually.
D
----
$800 is almost cheap enough
by
SEWilco
·
· Score: 2
I'm sure there will be a number of/. readers who will use this device as a platform. Assorted other sensors and programming will sprout from the original device.
Of course, if they publish the programming and communication specs then there will be Perl controllers in no time. And they'll sell more the cheaper it is.
Hmm. The web pages mention openness. The communication protocol is supposedly available, and the interface jack on the robot also carries the RF serial data so that other devices can use the RF link.
The web site also mentions a price of $700, not 800. I'm sure there will be a bunch of hardware hackers bolting various things to the basic unit.
AI = Heuristics today
by
squarooticus
·
· Score: 2
In academia, AI is just a synonym for heuristic algorithm design; there is nothing "intelligent" about it. I can see two primary reasons for AI not progressing in the past twenty years:
All the major AI researcher jumped ship into industry in the early 80's. This left the brainless, the unmotivated, and the senile to carry on AI research in academia.
It isn't clear to anyone (except the all mighty Marvin Minsky with LOTS of sarcasm) that it is even possible to model human intelligence with a digital computer. It's a nice idea...for a philosophy book. Additionally, the human brain is so massively parallel that even if we had a "human intelligence algorithm," there might not exist enough digital computers today to run it.
So, AI became a study in heuristic algorithms and machines that learn in a very isolated environment, and strong AI kind of fell by the wayside. Unlike foundational 70's research in cryptography, algorithms, systems, and even graphics, I'm sure we'll be laughing at 70's AI research a hundred years from now.
AI went through a spell, sometimes called the "AI winter", when it was hard to get support to do R&D (unless you hid the fact that you considered it to be AI -- lots of products, from games to elevator controllers, use techniques derived from AI to solve fairly simple problems). That's since eased somewhat, though as far as I know nobody's aiming at the Turing test right now. Nevermind having the intelligence of a dull chimp -- we're at the cockroach level, and accurately imitating mouse's brain would be a breakthrough
Symbolic (traditional) AI got into trouble for two substantial reasons. First, the real world is a noisy place and you can never see all of it. So there is always uncertainty and ambiguity in the input, and strictly symbolic AI doesn't deal with uncertainty very well. Fuzzy logic tries to patch this, but it doesn't seem to scale very well. What's needed is to mix in some of statistics, which exists to cope with ambiguous data. One of the newer AI inference technologies does this quite nicely; it's called Bayesian Inference. (My work involves Bayesian inference, and I don't claim to be unbiased. It's works, and it feels right.)
Second, AI needs some knowledge about how the world works -- a knowledge base -- in order to reason about its input, and building a knowledge base is usually a major project. Writing a knowledge base is essentially programming in a wierd special-purpose language, and the rest of the project may be as simple as writing I/O wrappers for a generic inference engine. (If you want to try it, get clips, an expert-system shell, built around a rule-based inference engine.) Building knowledge bases is gradually getting easier, just as programming got easier after people had been doing it and building better tools for a couple decades. And machine learning techniques can do some amazing things, but with some caveats. How well it learns, and what it learns, depends on how you set up the learning problem and how you package the input. And I don't know how well the techniques scale.
Cye is cute, but it has nothing to do with state-of-the-art AI. I'd call it a pretty impressive example of minimalist engineering. The gear-shaped wheels both give traction and minimize navigation uncertainty, and they use the motor fan blades as shaft encoders. IIRC, it has one board and two moving parts. I don't see why it costs more that $75.
Isn't this the Asimov paradox? I always thought he was the first to say (or at least be credited with), "Time travel can't exist, because what happens if I go back and time and kill my own grandfather? Therefore I couldn't have been born, therefore I could not have gone back in time..."
I always liked Heinlein's approach: there are no paradoxes. Because these things were invented means, by their very existence, that you didn't kill the inventor.
I always like to wonder about the flip side, which is "What will happen if time travel ever is invented? How come no one has shown up claiming to be from the future?"
Can it do anything really useful?
by
Knight
·
· Score: 2
It seems that thanks to a particularly prosperous period in our history, we have moved away from creating useful things, to creating things solely for our entertainment. While that is not all bad, would it hurt to take this technology and turn it into something truly useful, like a robotic help for individuals who are bed-ridden, or other types of helpful uses? I'd love to see that. ------------------------------------------- ------------ If you need to point-and-click to administer a machine,
The article mentions that Cye whistles and chirps like R2-D2. Clearly, the next step is your own personal C3PO -- for when you need a twittering yet comedic companion with a British accent.:)
When it can put the dirty dishes in the dishwasher, get a beer out of the fridge and bring it to me, ald hitch the vacume cleaner to itself, it'll be ready for primetime.
Until then, it's like Homer Simpson with the telepods, dragging one upstairs into the bathroom so he can run back, sit in the chair, and then effortlessly teleport to the bathroom.
I saw this robot last year... and wasn't impressed
by
SawdustBuffalo
·
· Score: 3
I saw this robot demonstrated about a year ago at a "Robotics Expo" in Boston. (and was asked to fill out a little card indicating my income and at what price I would consider buying it. I said $50. I think they said they planned to sell it for a few hundred dollars, but I could be remembering wrong on that one)
The design may have changed since then, so this may no longer be correct, but the version I saw used only dead-reckoning (counting the revolutions of each wheel) to determine its position (and to some extent, whether it is stuck against an object), and had no other sensors (no bump sensor, no compass to sense direction, nothing to tell if it had tipped over or if the vacuum had become caught on the edge of a rug or a small pet, etc...)
I asked the guy who was demonstrating it what happened if there was a bump in the floor, it ran over something (would change "apparent distance" as well as possibly direction), your pet/kid/you bumped it and moved it, etc.
He sort of stammered and avoided the question.
Never mind the fact that you have to hook the vacuum up to it (and remove it, unless you want your vacuum sitting out in the corner of your living room), pre-load its tray with drinks, and recalibrate its destination points every time you move a piece of furniture.
Sigh.
p.s. again, these are only comments on a demo version I saw last year
Actually the robot lawnmower has been around for about 20 years. The electronics have gotten better and cheaper, of course. This robot butler has spiked wheels for running on the carpet. I wonder how well it will work on grass.
Weren't they doing this back in the 80's?
by
konstant
·
· Score: 4
I could be wrong since I was only a littleun back then, but I strongly recall all sorts of hexagonal men and robodogs that supposedly would serve your drinks (how extremely...odd) and pick up your underwear. What ever happened to those?
It's a sad commentary on the state of AI that we haven't much progressed in 10 years. All we've succeeded in doing is rendering the creatures a little more anthropomorphic in appearance (think Furby), but their base intellect is still that of a dull chimp. The mean processing power has skyrocketed and there are more skilled people in tech than ever before, yet Teddy Ruxpin is more or less still the undisputed champion of artificial brains.
Is AI a dying field? I'm honestly curious. When I took my lone AI course in college, I was dismayed by the dronelike applications of DFS and BFS when I was expecting something a little more exotic. It seemed as though the professors lacked all spark of imagination - incredible when you consider the flare AI has made in the popular mind. And don't get me started on the affection these dodderers had for the sadly inadequate LISP family of languages.
Even this little robot cylinder thing doesn't do more than navigate a path through a virtual field. Nothing most slashdot readers couldn't code in under an hour. What ever happened to breathroughs in AI? Why are the serious researchers leaving it for other fields? Should we forget the dream?
-konstant
-- -konstant Yes! We are all individuals! I'm not!
I actually know a guy who got a wife from the Phillipines. He flew over there, spent 1 week scoring with underage girls, then went into some warehouse and picked out his "wife" out of 200 girls. Went to her house and met her family ("Yeah, I just bought your daughter"), then waited for 6 months for her to arrive. He was really excited and she was ok when she first got here. I haven't seen him lately, but I am really curious. It was his third wife and he looked like a 70's disco throwback. It was most humorous.
Time Travel Has Been Invented
by
SEWilco
·
· Score: 2
Obviously, time travel has been invented and a time machine constructed.
But every time someone defines a time travel theory which can be developed into a time machine, eventually someone working with time machines gets angry about their work and kills the inventor's grandfater.
That is why we don't see any workable time travel theories. When they are created they are eventually destroyed.
Slashdot Mirror
Humph. Windows?
... well, pretty dangerous, actually.
Is there a Linux version?
I'll bet a Beowulf cluster of those would be
D
----
Of course, if they publish the programming and communication specs then there will be Perl controllers in no time. And they'll sell more the cheaper it is.
The web site also mentions a price of $700, not 800. I'm sure there will be a bunch of hardware hackers bolting various things to the basic unit.
So, AI became a study in heuristic algorithms and machines that learn in a very isolated environment, and strong AI kind of fell by the wayside. Unlike foundational 70's research in cryptography, algorithms, systems, and even graphics, I'm sure we'll be laughing at 70's AI research a hundred years from now.
--
Kyle R. Rose, MIT LCS
[ home ]
Symbolic (traditional) AI got into trouble for two substantial reasons. First, the real world is a noisy place and you can never see all of it. So there is always uncertainty and ambiguity in the input, and strictly symbolic AI doesn't deal with uncertainty very well. Fuzzy logic tries to patch this, but it doesn't seem to scale very well. What's needed is to mix in some of statistics, which exists to cope with ambiguous data. One of the newer AI inference technologies does this quite nicely; it's called Bayesian Inference. (My work involves Bayesian inference, and I don't claim to be unbiased. It's works, and it feels right.)
Second, AI needs some knowledge about how the world works -- a knowledge base -- in order to reason about its input, and building a knowledge base is usually a major project. Writing a knowledge base is essentially programming in a wierd special-purpose language, and the rest of the project may be as simple as writing I/O wrappers for a generic inference engine. (If you want to try it, get clips, an expert-system shell, built around a rule-based inference engine.) Building knowledge bases is gradually getting easier, just as programming got easier after people had been doing it and building better tools for a couple decades. And machine learning techniques can do some amazing things, but with some caveats. How well it learns, and what it learns, depends on how you set up the learning problem and how you package the input. And I don't know how well the techniques scale.
Cye is cute, but it has nothing to do with state-of-the-art AI. I'd call it a pretty impressive example of minimalist engineering. The gear-shaped wheels both give traction and minimize navigation uncertainty, and they use the motor fan blades as shaft encoders. IIRC, it has one board and two moving parts. I don't see why it costs more that $75.
Are we guaranteed that they won't rebel and revolt against their Owners in some post-apocalyptic manner?
Kagenin
"All warfare is based on deception."
Sun Tzu, "The Art of War"
It seems that thanks to a particularly prosperous period in our history, we have moved away from creating useful things, to creating things solely for our entertainment. While that is not all bad, would it hurt to take this technology and turn it into something truly useful, like a robotic help for individuals who are bed-ridden, or other types of helpful uses? I'd love to see that.- ------------
------------------------------------------
If you need to point-and-click to administer a machine,
The article mentions that Cye whistles and chirps like R2-D2. Clearly, the next step is your own personal C3PO -- for when you need a twittering yet comedic companion with a British accent. :)
When it can put the dirty dishes in the dishwasher, get a beer out of the fridge and bring it to me, ald hitch the vacume cleaner to itself, it'll be ready for primetime.
Until then, it's like Homer Simpson with the telepods, dragging one upstairs into the bathroom so he can run back, sit in the chair, and then effortlessly teleport to the bathroom.
The design may have changed since then, so this may no longer be correct, but the version I saw used only dead-reckoning (counting the revolutions of each wheel) to determine its position (and to some extent, whether it is stuck against an object), and had no other sensors (no bump sensor, no compass to sense direction, nothing to tell if it had tipped over or if the vacuum had become caught on the edge of a rug or a small pet, etc...)
I asked the guy who was demonstrating it what happened if there was a bump in the floor, it ran over something (would change "apparent distance" as well as possibly direction), your pet/kid/you bumped it and moved it, etc.
He sort of stammered and avoided the question.
Never mind the fact that you have to hook the vacuum up to it (and remove it, unless you want your vacuum sitting out in the corner of your living room), pre-load its tray with drinks, and recalibrate its destination points every time you move a piece of furniture.
Sigh.
p.s. again, these are only comments on a demo version I saw last year
p.p.s also at the robotics expo were Lego Mindstorms, some small solar robots, and NewtonLabs and their vision system (used on their winning robots in the Internation Micro Robot World Cup a few years back.
Actually the robot lawnmower has been around for about 20 years. The electronics have gotten better and cheaper, of course. This robot butler has spiked wheels for running on the carpet. I wonder how well it will work on grass.
I could be wrong since I was only a littleun back then, but I strongly recall all sorts of hexagonal men and robodogs that supposedly would serve your drinks (how extremely...odd) and pick up your underwear. What ever happened to those?
It's a sad commentary on the state of AI that we haven't much progressed in 10 years. All we've succeeded in doing is rendering the creatures a little more anthropomorphic in appearance (think Furby), but their base intellect is still that of a dull chimp. The mean processing power has skyrocketed and there are more skilled people in tech than ever before, yet Teddy Ruxpin is more or less still the undisputed champion of artificial brains.
Is AI a dying field? I'm honestly curious. When I took my lone AI course in college, I was dismayed by the dronelike applications of DFS and BFS when I was expecting something a little more exotic. It seemed as though the professors lacked all spark of imagination - incredible when you consider the flare AI has made in the popular mind. And don't get me started on the affection these dodderers had for the sadly inadequate LISP family of languages.
Even this little robot cylinder thing doesn't do more than navigate a path through a virtual field. Nothing most slashdot readers couldn't code in under an hour. What ever happened to breathroughs in AI? Why are the serious researchers leaving it for other fields? Should we forget the dream?
-konstant
-konstant
Yes! We are all individuals! I'm not!
I actually know a guy who got a wife from the Phillipines. He flew over there, spent 1 week scoring with underage girls, then went into some warehouse and picked out his "wife" out of 200 girls. Went to her house and met her family ("Yeah, I just bought your daughter"), then waited for 6 months for her to arrive. He was really excited and she was ok when she first got here. I haven't seen him lately, but I am really curious. It was his third wife and he looked like a 70's disco throwback. It was most humorous.
But every time someone defines a time travel theory which can be developed into a time machine, eventually someone working with time machines gets angry about their work and kills the inventor's grandfater.
That is why we don't see any workable time travel theories. When they are created they are eventually destroyed.