What is Digg?Sponsored by Travelzoo
Take Advantage of Ridiculously Low Holiday Airfares view!
travelzoo.com - Flights $52 and up for Thanksgiving, Christmas & New Year. But move on it now.
36 Comments
- dsignr, on 10/12/2007, -0/+15"The same way Windows Speech Recognition does"
And we both know that it works flawlessly.
/sarc - Amablue, on 10/12/2007, -1/+15Does the wittle wobot want it bottle?
- scispaz, on 10/12/2007, -1/+12I thought my parents had too much tape of me growing up.
- PaddyTheChump, on 10/12/2007, -2/+12What the hell did I just read?
- Amablue, on 10/12/2007, -0/+5@dsignr
That's why they're doing this research. - jcaino, on 10/12/2007, -3/+7of course the robot wants a bottle.
they are fueled by alcohol, after all. - patricktu, on 10/12/2007, -1/+5"Moreover, how can a robot hear the word "ball" without confusing it with ambient noises and accent changes?"
The same way Windows Speech Recognition does. - squeevey, on 10/12/2007, -0/+3A child (read: the brain) also has to "learn" to see. I'd say it's much easier for a brain to see because it interprets everything as a direct electrical signal with minimal processing. Each rod and cone from my understanding goes into the brain on its on nerve pathway. The bundled group is called the optic nerve. So where a computer processes one at a time, the brain does concurrent operations. In short: the information is there at the same time and the repetition of something helps nerves grow and create more pathways. the more pathways a brain creates the more items a person has learned. The faster computers and more concurrent computer processors become the easier to "see".
http://en.wikipedia.org/wiki/Photoreceptor_cell - dsignr, on 10/12/2007, -3/+5Teaching a robot how to talk is extremely difficult, regardless of the amount of data it has. A number of factors have to be considered--referential language, the ability to correlate words or phrases to abstract objects (like love or time), and even syntax.
An infant can see a ball, and correlate that to his parent's annunciation of the word "ball". A robot will see a bunch of pixels and cannot distinguish the ball from its background. Moreover, how can a robot hear the word "ball" without confusing it with ambient noises and accent changes?
It seems a bit farfetching to me. - DiggingDeep, on 10/12/2007, -0/+2a little creepy imo
- dsignr, on 10/12/2007, -1/+3The research's ultimate goal is to teach a robot how to speak.
That's different (but builds on) the research of speech recognition. It SHOULD go 1) R&D Speech Recognition. 2) Teach a robot how to talk.
It's like teaching a toddler how to fly an aircraft when he hasn't grasped the knowledge of walking. - dancpsu, on 10/12/2007, -0/+2Yeah, I think they flipped direct objects.
But don't worry, I'm sure Toco's short-term memory is longer than yours. - Malcx, on 10/12/2007, -1/+3I had no opinion on Vista until i saw that overlay ad.
Now I hate it. - omnidugg, on 10/12/2007, -0/+2got a screenshot? I'm not seeing that add. (they're probably doing geo-ip lookup... i'm in egypt, and i bet they think (correctly) that no-one pays for software over here
- themarkshow, on 10/12/2007, -0/+1That video of the robot on the left of the page was the best!
- dancpsu, on 10/12/2007, -0/+1You can still mimic concurrency in computer vision. What is difficult is either separating out an object from the background (you can cheat by chroma filtering, but the brain does something more complex), or go all out and have the computer vision system generate a model from the images given. The brain does this in an instant with very sparse data, but a computer sometimes loses simple edges.
It does help to have the dynamic range of human vision though. - doctechnical, on 10/12/2007, -0/+1Darn it, I was hoping this was about a Korenthal Associate's Babble program.
- tech42er, on 10/12/2007, -0/+1@jc, dugg for the Futurama reference.
- msgyrd, on 10/12/2007, -0/+1Give the robot two cameras and it can determine depth of objects, where a ball would stand out as a closer object than the backgrounds. Add in chroma filtering, and you have the potential for a machine to recognize simple shapes and objects.
- thesixthdesign, on 10/12/2007, -0/+1Shouldn't he figure out a way for robots to think first before he starts trying to get them to talk?
- Darkhacker, on 10/12/2007, -0/+1"Thinking" is very open ended. What is thinking? Is performing mathematical calculations not thinking? Or were you referring to having robots act independent and make their own decisions? Acting independent is thousands of times more difficult to teach a robot than is teaching it to talk. Acting independent could mean almost anything but a language has a solid set of rules (like math) that a computer can much more easily interpret.
- dancpsu, on 10/12/2007, -1/+2At least the Bible quote could have been the part about the Tower of Babel, to be more on topic.
- redlemon, on 10/12/2007, -0/+1ahead it's time... but a lot of research is. imagine how many cores the human brain has! and i bet every one uses hyper hyper hyper threading. it's only a matter of time before we are able to mimic parts of the human brain using electronics (maybe even organics), but it's a matter of a very long time. regardless of your belief in a higher power, humans took a lot longer than 10,000 years to debug.
that's my thinking anyway. - conna, on 10/12/2007, -0/+1This is a DARPA funded project.
http://www.darpa.mil/ipto/Programs/gale/index.htm - dancpsu, on 10/12/2007, -0/+1Based on linguistic research dating all the way back to the 1950s, human brains have linguistic knowledge built-in. AI researchers balk at this because it makes the human brain effectively a black box that cannot be simulated by a computer.
I would imagine that after gathering all this data, that Dwayne's ability to learn language will still be mysterious, and Roy's robots will not be able to match Dwayne at age 2, even with arguably identical stimulation. A proper linguistic model may be built and put in place to fill the gap, but who is to know if it is identical or as powerful as Dwayne's? - redlemon, on 10/12/2007, -1/+1get of my tubes pissant!
- jcaino, on 10/12/2007, -1/+1i imagine this will be incorporated into image recognition technology.
sure, its not going to be able to get up in front of a room and deliver a dissertation, but baby steps man, baby steps. - dgblackout, on 10/12/2007, -1/+1@ keeblamos : welcome to my ever-growing blocked list.
- omnidugg, on 10/12/2007, -5/+5@keeblamos
Yeah. God would love you spamming this site. Nice work - lolo2007, on 10/10/2007, -0/+0ahead it's time... but a lot of research is. imagine how many cores the human brain has! and i bet every one uses hyper hyper hyper threading. it's only a matter of time before we are able to mimic parts of the human brain using electronics (maybe even organics), but it's a matter of a very long time. regardless of your belief in a higher power, humans took a lot longer than 10,000 years to debug. http://www.gwafi.com/news.html http://www.gwafi.com/story.html
- Sethwm2, on 10/12/2007, -1/+1Yeah patricktu it works great. http://youtube.com/watch?v=fV1kqthZf2g I mean.. DAMN I want my speech recognition to work this good on my Mac.. Oh wait.. It works like 100000 times better
- omnidugg, on 10/12/2007, -1/+0from the third page:
The idea was to supplement his robot's long-term memory with short-term memory. Both would be engaged in pattern recognition, searching speech input for recurring phonemes, but the short-term memory would focus on the recent past . By giving Toco a mild case of ADD, Roy made his son more like the robot he was trying to emulate.
ur... that's gotta be illegal? - LeNoir, on 10/12/2007, -1/+0...
- dsignr, on 10/12/2007, -3/+1@omnidugg:
i can't believe you didn't use "i'm in butt-f*ck egypt" when you had the chance. - JamesWilson, on 10/12/2007, -4/+1Text to speech has been around for years. ***** n00bs.
/joke - omnidugg, on 10/12/2007, -3/+0fta:
They can also switch off the cameras while Patel is breast-feeding or hit the "oops" button when something too personal gets recorded. In fact, a glowing, wall-mounted "oops" button can be found in every room, allowing them to make Total Recall's archive something less than total. Roy pressed it one day after emerging naked from the shower when the cameras were running.
© Digg Inc. 2009
— Content created and posted by Digg users is