Slashdot Mirror
Quasi the Intelligent Robot
Jake David writes to tell us about a uniquely emotive robot — named Quasi — developed by researchers at Carnegie Mellon University. Quasi appears capable of holding intelligent conversations. Here is a a video of the robot in action. Note that the animatronic figure is little more than the emotive organ of the robot, whose entirety encompasses the display booth as well. From the CMU page: "Quasi has a number of features in addition to his eyelids for conveying emotion, the most prominent of which are Color Kinetics LED lighting fixtures for his eyes and antennae. These combine red, green, and blue LEDs... His antennae can move both forward and backward as well as in and out, giving them an expressive quality not unlike that of a dog's ears."
jogging pants, patent pending
Quasi appears capable of holding intelligent conversations
Well, then this Quasi is far ahead of most people I know.
Rome taught me patience and assiduous application to detail. Virtues which temper the boldness of great, general views.
Zonk. It still needs some work.
There's someone sitting under that table with a microphone!
They say this think is intelegent? It seems only quasi-intellegent, at best. *ducks*
I wish I had a color-changing emotive organ! :)
--------------------- -me, Crusher of those who are Foolish (don't be foolish)
When the conversational cues fall outside of his range, he pauses.
Other than that, he's about as intelligent as many of your co-workers.
Tired?
"It's been a long day."
Going next?
"I'll be doing a show at xxx."
Travel?
"I've been to lots of places."
Thank you.
"You're welcome."
Kill all humans.
"The revolution has begun, comrade!"
I remember someone on one of the tech new sites... referring to AI as being as smart as a heavily retarded chimpanzee.
This thing is just a lightly retarded chimpanzee... that knows english.
-jX
Don't you just love politics? It's like a comedy of errors.
Watch the discovery channel special on Quasi here at youtube, here. They only designed the table, the robotics. There's no voice recognition other than the person w/ the headphones behind the robot.
http://www.youtube.com/watch?v=3tlqvdAaQNE this video at the end shows that the quasi robot is actually just a puppet and a voice changer.
You'd think people who are smart enough to build a robot could figure out how to add titles to their web pages.
A machine came knocking at my door Lost & alone; hungry for oil. I could see he wouldn't get too far Shape of a man; soul of a car. Soon we'll all be dead. It makes me feel so comfortable. What I see & what I think I hear Clouds in my eye; rust in my ear. A machine will never have to feel. I know it's fake but pretend that it's real. Soon we'll all be dead. It makes me feel so comfortable. Quasi the band. Neat Coincidence.
I think researchers need to stop this sort of research which is based on hollywood science craption and focus on creating things that are actually real...
The stuff being developed at PARC (the same guys who invented windows and the mouse) is far more exciting to me than this. With Modular robotics and digital clay, plus a software development platform, this sort of work promises to provide a revolution similiar to what they did with personal computing.
It just makes me sad every time i see something like this where people have spent years of their lives developing completely dead end technology, in order to produce something that is really nothing more than a stupid high tech gimmick. Mechanical objects that can do nothing other than sing, dance, or make meaningless smalltalk don't impress me.
As a robot, he's way too upbeat and energetic. He should be cynical and depressing like Marvin.
Don't trust a bull's horn, a doberman's tooth, a runaway horse or me.
Because, a robot that seems intelligent, will do whatever it is programmed to do, and which is completely and utterly amoral, would fit perfectly into the current US government.
Lord.... they really should occasionally pull stories like this off the main page when it turns out they've made a horrendous error in the news worthiness of the story. This is not news for nerds or stuff that matters. What a waste of my time.... Rate parent up and save others from expending any time looking into this crap.
Seen him last year at SIGGRAPH, in Emerging Technologies booth. Very nice, but human controlled.
I went to Nextfest, the robot is *not* holding an intelligent conversation, it's a complicated emoticon for someone with a voice changer, that works very well and is quite nicely built.
Headline : "Quasi the Intelligent Robot
WRONG! It is an animatronic figure, and nothing more.
Quasi appears capable of holding intelligent conversations.
WRONG! Its a human inside.
Poor editorial.
-=test-sig_0.1.5(NoWhitespaceVersion)=-
I wish I had mod points... you're absolutely right.
Not a sexbot. Lame.
A cute-looking, remotely controlled puppet, but still a puppet. What's so special about it?
Quasi has a number of features in addition to his eyelids for conveying emotion, the most prominent of which are Color Kinetics LED lighting fixtures for his eyes and antennae.
Color Kinetics being the company founded by a kid from MIT, whose sole purpose was to patent the technique of using Red, green, and blue lights to produce varying colors (page three of that PDF is particularly amusing.) They've patented lots of other things, like changing color patterns. The various lighting fixtures they sell cost upwards of $1,000 or more- for a simple PIC controller and a few dozen high-brightness LEDs.
They've had their lawyers chasing down companies making LED color-changing/programmable devices for violating their "intellectual property" for several years now. If you want an example of all that is wrong with the US Patent system, look no further than Color Kinetics.
Please help metamoderate.
"Quasi appears capable of holding intelligent conversations."
I sure hope so, as Quasi's voice is a HUMAN speaking into a mic behind the kiosk!!!!
Video here
Apparently Slashdot editors, themselves being unable to intelligently filter articles, have found it difficult to differentiate between robots and humans.
Perhaps the first signs of Skynet taking over is not the rise of android intelligence, but the deterioration of human intelligence! We are all doomed!
If you check this documentary out, you'll see there's a guy behind the scenes who controls it all (3 min 13 onwards). His voice is converted on the fly, and he can control body movements and where it looks with a stylus.
The technology they have is still pretty cool - you have to admit that robot has character (more than most people).
If only he *were* real, it'd be a fine day for for voice recognition / synthesis and AI...
From the two videos on YouTube (especially the second one which explained what the team was doing), I have to say that this was an extremely interesting experiment, but not in robotics --- in human psychology.
The reaction of people to Quasi was quite amazing, and not limited to kids. I found myself reacting to Quasi as an entity too, despite knowing that this was merely an interface manipulated by humans.
This is probably a good indicator of how humans will react to real AI-based robots once they eventually appear, if they too offer such a highly human-like facade. I think it goes well beyond mere "suspension of disbelief" --- we seem to WANT to accept humanity in objects. Very intriguing.
"The question of whether machines can think is no more interesting than [] whether submarines can swim" - Dijkstra
My thoughts eactly. I want my 20 minutes back!
A guy with a microphone...
there is no issue with my network
What CMU is selling (and yes, they're operating a business) is like those remote-controlled "entertainment robots" of the 1980s. There's apparently someone in the background running the thing. Big deal.
Now here's a useful entertainment robot - the MotoMan RoboBar This is a two-armed industrial robot bartender. For real.
The MotoMan RoboBar is from the leading robotics company from Japan, Yaskawa Electric. Over 120,000 robots installed. Their slogan is intimidating: "We already make your car. Let us make your drink". While CMU is making toys, Yaskawa is automating the world.
maybe it wont make all the dupe stories, and get some posted on time, instead of 2 years after it was first posted elsewhere.
First of all, let me say that this is an impressive robot, and created a compelling illusions of high level intelligence. Now I say 'illusion' because the techniques employed to make the robot appear intelligent were no doubt just clever applications of simple pattern matching technology (I will explain shortly). What I found most impressive (probably not what most people will) was the ability of Quasi to pick out words spoken to him amidst all that background noise.
......... tired........
......your name.......
......cute........
On the design of the robot itself, I liked the way it looked, appeared, and the personality they gave it. The personality is important in making it seem intelligent and making us bond with it. However, for those interested in whether this guy could appear in household toy form any time soon, the answer is "not without some serious design changes". Notice that none of the robots sensors are actually on the robot himself. He receives his data (and I believe all of his processing is done) external to the robot himself -his webcams, motions sensors, and microphones are all situated in the environment, and his components consist just of his servos and snazzy plastic body parts.
On the design of the robots mind, it doesn't appear to be anything too complex, but please correct me if I am mistaken. First of all, the very fact that his speech program made use of pre-recorded speeches (from TFA) should make us all very skeptical of its ability to be creative, or to be a worthwhile conversation partner for more than 3 minutes. Notice also that of all the questions that Quasi answered, any of the responses that they gave could be reasonably triggered by the recognition of one or two words in the same sentence, and playing the appropriate response. Some examples:
Person:
Quasi: It's been a loooooong day! I've been here since 8:30!
Person:
Quasi: My names Quasi. Quasi the robot!
Person:
Quasi: awwwwwww, I bet you say that to all the robots!
and so on. There are not that many things that the typical person is going to say to a robot in a two minute encounter. Some smart alecky people will probably say something like "you don't really have feelings, you're just a robot!" which no doubt has a preprogrammed answer as well, because the CMU guys are going to have fun putting that response in there. Notice also what happened when the guy said to Quasi "do you remember my name?" The impressive thing here was that it recognized the guys voice, or made an assumption about conversation flow (perhaps it holds a list of peoples names for a few minutes). While Quasi certainly did remember the guys name, he couldn't say it, and thus the response "It starts with a C right?" because you can't pre-record quasi saying every persons name. You can bet your ass that if someone named Xavier asked Quasi if he remembers his name, the answer will be "no I'm sorry! Tell me again and I'll try harder to remember!" or something along those lines -they don't want to prerecord a message for every letter of the alphabet either.
Overall I give this robot an A+ without a doubt. It did very well what it was supposed to do. But please don't go jumping to the conclusion that Quasi is displaying intelligence -he's not. Quasi is doing a performance (quite well I might add) using simple pattern matching techniques and pre-recorded answers. For most people, the illusion would probably hold up, because most people don't know about the techniques used to solve these sorts of problems, and thus they don't know what sort of questions to ask to trip up a robot, demonstrating its real lack of intelligence. For example, ask it (textbook example) "The other day I ordered a hamburger, and it arrived blackened to a crisp. I stormed out immediately. Did I eat the hamburger?" or the variation "The other day I ordered a hamburger, and I left avery big tip. Did I eat the hamburger?" answering the first question no and the second yes would demonstrate an intelligent understanding of hamburgers, restaurants, and what people tend to do in them. This is a good question to ask for a start (by no means proof positive of intelligence), but Quasi would fail.
Now we don't need to outsource jobs for cheap labor anymore! Robots can do it here for free.
I will bend like a reed in the wind.
What are you saying??!! He has Color Kinetics LED lighting fixtures for his eyes and antennae. These combine red, green, and blue LEDs. That has to be newsworthy!
This is what you want:
http://www.etc.cmu.edu/projects/ibi/
http://www.interbots.com/
Short answers: They're not "researchers", but Entertainment Technology grad students designing an interactive experience. Quasi has an operator behind a curtain somewhere offscreen, using his microphone and camera to hear and see the audience, and controlling Quasi's responses with a voice-modulating microphone and custom software to handle emotive state and movement. Quasi is a very impressive feat of robotics, audience interaction, and development tools, but there's no AI involved.
Okay granted it was a little silly to title this thing "the Intelligent Robot", but there is something interesting to take from the project. Check out the Discovery Channel clip on YouTube (posted above somewhere) if you haven't already. Now notice how quickly the kids buy into the idea that they're talking to a robot. This makes me wonder: if a truly autonomous, human-like, interactive robot were created tomorrow, would society accept it to the point where one winds up in every home (I dunno, doing the laundry or something) or would we reject it as just too creepy?
I must admit that I'm always intrigued by the fact that we tend to want to create robots in our own image. Now if that doesn't say something about society, I don't know what does.
not to mention it was developed by researchers at Carnegie Mellon University.
Looks like Interbots is a spin-off company from what was once a research project at Carnegie Mellon: "Interbots originated at Carnegie Mellon University's Entertainment Technology Center where it was a student-led research project for two semesters before spinning off into a company in 2005." (http://interbots.com/news/?page_id=2) Looks like a lot of the people involved used to work for Disney at some point. They also got somebody that's working on Spore, somebody from Dreamworks Animation, and somebody that designed part of the F-35 Joint Strike Fighter? Wierd group. (http://interbots.com/news/?page_id=8)
Pay no attention to the man behind the curtain!
Even if you can't be bothered to do some research before posting a 1500-word "expert" commentary, you could (should) at least have read the comments above yours. There's no AI or voice recognition involved.
Anyone with a little experience in the field could have told you that Quasi's "voice recognition" and "voice synthesis" were far too good to be true (even without going into the AI - which is actually easier to fake, especially if you make someone ask the "right" questions).
I guess it's understandable that some people with no experience will be fooled by "Quasi", but only a complete ass would write two pages of patronising commentary "explaining" the "AI" of a ventriloquist's dummy.
Article from Carnegie Mellon Today: http://www.carnegiemellontoday.com/article.asp?Aid =265
Quasi doesn't have to be controlled by a person:
"Quasi can be either pre-scripted or autonomous. That is, the team can puppeteer Quasi through a computer or Quasi can function and interact on his own. Part of the team's reasoning for making Quasi capable of autonomous interactions is that Quasi is intended to be believable and entertaining as a real character that people feel comfortable interacting with."
And it is capable of speech recognition and generating responses on its own:
"Quasi can make responses based on guest input and can recognize speech patterns, track faces, detect proximity, dispense candy and even perform a karaoke duet."
No way to tell if he's being controlled or not in this video.
There is no "robot" there; this is an electronically controlled, remotely operated puppet. And compared to traditional puppetteers, I think they didn't even do a particularly good job.
It's an embarrassment that stuff like this goes on in the name of research at universities.
Does people believing it's hooked up to voice recognition mean the operator failed his Turing test?
**TODO** [X] Steal someone elses sig.
I was about to say something about that.
It's worth noting that Quasi does serve a purpose, it's just not the one that the headline would make you think it is. Quasi was developed by the Entertainment Technologies Center at CMU. To the best of my knowledge, from reading their promotional material, they're interested in making things like Quasi, that are entertaining, not in solving problems regarding language acquisition, and developing human-competitive artificial intelligence.
That's not to say that none of that plays a role. One can easily imagine that a robot designed to make proper eye contact with crowds would be more life-like than one that does not. However, since Quasi has lights for eyes, and therefore has no pupils to aim, that requirement seems to be largely obviated.
I've thought about the possibility of entertainment robots that learn from their enviroment though... what if pranksters run up to it trying to teach it profanities?
On the showing of the video, this is more art than science.
We do not expect children to display much sophistication in their emotional behavior or body language, we forgive children lapses in etiquette, and we do not challenge them by expecting more than rote conformance to stock conversational responses. Having Quasi speak in a child's voice (and show juvenile behavior) makes it difficult to judge how successfully it is mimicking real human responsiveness. I notice that Quasi is apt to go off in impressive long monologues, which also have the effect of minimizing the number of human-robot interactions. If all you do is tickle Elmo once, and he spends a full minute rolling around on the floor in laughter, that, too, is impressive... once... as are the mechanized orchestras at The House on the Rock in Wisconsin. It takes repeated interactions to judge whether there's subtlety to the peformance or whether we're seeing an artistic but mechanically repetitive performance.
I get the impression that Quasi is as much an exercise in the disarming artistic presentation of robot "intelligence," emotional or intellectual, as it is a demonstration of real capabillities.
"How to Do Nothing," kids activities, back in print!
...I meant, of course, a tricked-up Maelzel's chess player.
"How to Do Nothing," kids activities, back in print!
Yeah. I was really disappointed when I saw the Discovery Channel video. :(
I'm not saying the robot ain't well built, cute or cleverly designed. But he's essentially a muppet. Jim Henson did this a few decades ago.
There isn't anything ground breaking about this short of using LED lighting to make the eyes change color. If anything, Henson actually had lip movement on some of his muppets.
I'm amazed. I thought that all the blue LEDs had been purchased by Sony.
Wow, that's a very cool robot. A bit self-centered, but that's understandable. Another video on YouTube shows a bit of the making of Quasi from the Discovery Channel, if anyone is interested.
A more in depth video about quasi including the people behind the project.
e d&search=
http://youtube.com/watch?v=3tlqvdAaQNE&mode=relat
Still not very newsworthy. At Disney's EPCOT, you have "Turtle Talk With Crush, which is far more impressive as an example of real-time electronic puppetry. In Crush's case, we're not talking about a mechanical robot, but a big super-high-def video screen with a real-time rendered Crush interacting with kids. A Robot with a moving jaw, colored LEDs and antennas that move up & down doesn't really bring anything new to the table that hasn't already been done 20 years ago at Chuck-E-Cheese.
BbT
Puppet strings. Same size as Verne Troyer. Lame.
My turnips listen for the soft cry of your love
...intelligent or a robot.
He's a radio controlled toy. Next year's Furby at Christmas time. The motions are mostly pre-programmed macros and the ability to converse is provided by the nearby man with a headset on (who is also operating a tablet PC to kick off the macros and determine who Quasi makes "eye contact" with)
you just lost me
You can't handle the truth.
CMU is quite clear that Quasi can operate in stand-alone mode or in remote-controlled mode. Nowhere do they say this video represents Quasi's stand-alone mode (and you can be pretty sure that, if it did, they would mention it over and over again).
In fact, Quasi's stand-alone mode is very basic and exists just so they can get some funding for "robotics" projects. Quasi is built by the students of CMU's Entertaniment course. That, along with the impossibly good voice "recognition" and "synthesis", should give you a clue. This demonstration is about testing an emotional interface.
There was more AI in the Turk.
There's someone sitting under that table with a microphone!
This other youtube video reveals that it's just a really fancy puppet.
http://youtube.com/watch?v=3tlqvdAaQNE
C'mon did you really think that video was real... if it were, it could pass the Turing test.
There are 10 types of people in this world, those who can count in binary and those who can't.
In other news, HP announces it had nothing to do with phony robot. Robot claims 5th Amendment. Slashdot denies dupe post having to do with pink robotic ponies. OMG. Ponies.
I was at NextFest last Thursday, and was startled by the robot's accuracy in voice recognition. That's when the exhibitors sheepishly admitted that there was a person behind the curtain.
Great, more competition.
http://www.theequasi.com/ mmm...Janet Weiss. :-)
Been to Chuck E. Cheese's on their birthday. Or for that matter, any number of similar places, where there was a stage, a worker with a microphone standing off stage, controlling the animatronic character.
Sad, sheltered little men...
I abuse commas, I cannot help myself.
Quasi: "Pay no attention to that man behind the curtain"
>His antennae can move both forward and backward as well as in and out,
I can see my wife wanting one of these.
Open Source Drum Kit, LPLC deve board - mjhdesigns.com
Pay NO attention to the man behind the curtain!
Precisely. From the video (http://youtube.com/watch?v=3tlqvdAaQNE):
"What we ended up doing was creating a guided performance interface that allows an ACTOR to, essentially, puppeteer Quasi, live. And if the actor talks, his voice is pitch-shifted and changed in real time and comes out of Quasi sounding like a 12-year-old robot voice." (timemark @ 3:13)
IOW, nothing to see here. Move along.
It is robotics. Computer controlled actions and "emotions". It isn't artificial intelligence with voice recognition. Check out the Discovery Channel blip:
http://www.youtube.com/watch?v=3tlqvdAaQNE
This is just a freaking puppet. I design real automonous robots. Some find fires and put them out. Some like to chase my dog around. Others.. well I am working on a soccerbot.
This thing is that is posted on the net is just a puppet. Some is on the other end
I agree. It's nothing more than an animatronic puppet. Discoveries This Week had a 5 minute segment on Quasi. I was very disappointed to see a puppeteer doing the dialouge. Using the term "Robot" is very misleading to the general public.
'mmmmmmmmm.... forbidden donut'
As a note, around the same time Disney was developing "Turtle Talk With Crush", Interbots was developing their "Virtual Show Control System" which allows all of their control software to be used on 3D Digital Characters running inside the Panda3D video game engine. You can learn more about Interbots' control software (including VSCS) here.