The big difference here is that a backprop model needs to essentially be told what the correct phonemic boundaries are, etc., either in the way it represents the phonemes (possibly as a localist representation) or as target outputs. The model discussed in PNAS is simply given raw speech sounds and learns the categories itself (and how many there are) with no given targets. It's the difference between an unsupervised and supervised model that makes the difference here. In the earliest stages of language learning, I don't think parents correct their children's babbling!
What I find so strange about the whole fuss over Google Book Search is that they're not doiing anything different than they do with their web search. As you should know, if you know anything about copyright, just about anything you create is automatically copyrighted (of course there are some things that are excluded). Certainly, webpages fall under this category. Thus essentially every webpage on the internet is automatically copyrighted. When Google crawls the web, they add updated pages to their cache. In doing so they are creating a copy of the page itself with which to analyze links, etc, as well as offering it up to the world under the "Cached" link. Site owners have the choice of opting out of Google's system, though few choose to do so. This is exactly what they are doing with the book search. The only difference is that website owners don't haved an industry front group to fund lawsuits.
Now, one could argue that the opt out system is not in keeping with the Copyright Act, whereby creators should choose to allow their use of their works, not have to make a point of disallowing it. If that argument holds in the Book Search case, however, then it certainly should hold for Google's web search.
I do think that many (though probably not all) of those textbook writers would agree that the brain itself does not operate in a discrete manner on a structural level. It's just that they don't care. Their argument rests on the hypothesis that the relevant mental and behavioral states are best described symbolically. Now, much of the evidence I've seen recently seems to refute this belief (of course, I hardly have an unbiased background based on who teaches the classes i've taken and those with whom I've done research). But it's just important for those reading this article to understand that the symbolic diehards are making a distinction between the anatomical brain structure and the functions they believe that structure instantiates. The connectionists and to a greater extent dynamicists (to break everyone up into schools) see structure and function as more tightly linked, and in general just see more similar processes at work in different areas of cognition.
I also think it's short-sighted for Birge to argue that there's no point whatsoever to studying biological systems from a higher-level perspective. His statement that we need to fully understand a worm brain before tackling any apparent aspects of a human one seems akin to arguing that we need to have a complete understanding of particle physics in order to build a basic machine. Biology itself serves as an even more useful example. The general scientific study of living things certainly preceded molecular biology and produced wortwhile findings, and similarly Darwin far preceded the discovery of DNA, etc. Reductionism is an extremely useful methodology, but often it needs to be informed by a higher-level perspective. That's why the field of psychology even exists. If Birge believes that we should ignore all high-level concepts and start our understanding completely from the lowest level neuronal networks, then he's espousing eliminative materialism, and the problem with that philosophical outlook is that it's extremely hard to explain anything when you've thrown out all the concepts that are purported to need explaining.
I'm working on the steering control program for our DARPA Grand Challenge vehicle, and it does that, too. Doesn't mean it's not "digital".
Clearly in the case of the DARPA challenge, the underlying physical machine is a digital processor. But I'm wondering, what type of algorithms does your control program use? The reason I ask is that the question Spivey's considering here isn't whether the physical processing mechanism resembles a digital computer, as clearly the anatomical brain structure itself does not, but instead whether the functional mechanisms (which in your example are directly defined as the algorithms instantiated by your code) are compatible with the serial and symbolic models posited in the classic era of AI as well as cognitive psychology.
In other words, that a computer program could model these functions is for the most part logically trivial and in fact Spivey employs a computational model in the article itself in order to test whether his theory accounts for the data. Similarly, many modern machine learning algorithms are qualitatively "un-digital" even though they are being modeled on a digital computer (hence why they require so much processing power). Do you see the distinction?
Actually, in the PNAS journal article that the Cornell News story refers to, Professor Spivey compares the data he obtained in the mouse-movement experiment to a computer simulation that uses both the TRACE model of spoken-word recognition (an old-school interactive activation model that still seems to be pumping out good data, especially for a localist model) and his own normalized recurrence attractor network (another localist statistical model). Both of these models, as well as the combined one that he creates for the simulations, are by their very nature dynamical systems models, and definitely not Turing machines.
Not totally coincidentally, Spivey is also my undergrad psych major advisor.
I'd have said "RTFA", but since slashdot doesn't directly link to the journal article PNAS, I guess you're kind of off the hook.
Re:Similarities between democrat party, communists
on
Joe Trippi Interviewed
·
· Score: 1
Let me guess; you're a college professor.
What's with the anti-academic attitude so prevalant with conservatives? Didn't you guys used to love Milton Friedman?
The road to being a professor is hardly one paved with gold. Grad school takes at least 5 years, during which you are expected to TA, work with professors on their research and find time to do your own research toward your dissertation, while being given about $15000 to live on per year. After that you have to spend another few years at least as a post-doc, being paid perhaps $35000/year. By the time you're even actually a faculty member (easily not until you're in your early to mid 30s), you'll make between $40,000-$65000, and that's only for those who get full-time faculty positions at major research universities (certainly the minority among PhDs).
Professors aren't the fat-cats. I'm sure you can point out some very high-paid professors, because they certainly exist, but they're definitely not in the majority.
I'll have to check it out further later, but at the moment I can't handle all that functionalism. I swear they're breeding me to be a total connectionist around here.
Linking music and language is hardly a new idea. The language analogy has been the focus of much of the research on music cognition in cognitive science in general. For one of the most direct examples of such a paradigm, look at Lerdahl & Jackendoff's generative grammar for Western tonal music, which sought to apply Chomskyan linguistic theory (hotly debated itself) to musical structures. While the syntax can be dealt with rather easily, pinning down a concept of "musical semantics" has been rather elusive. And to think Chomsky said, "If you take care of the syntax, the semantics takes care of itself." Not only is this not the case in music, a lot of people don't think it's the case in language anymore either. Ok this is the best I can think of for 3 AM. Maybe I'll try again later.
That sounds pretty misinformed. Have you ever filled out a financial aid application? The most important part of it is your family income, not your savings. Though if you have a great deal of assets saved for college they will expect you to use them, the actual EFC (expected family contribution) is calculated mostly from your adjusted gross income on your tax return. So what that has to do with a consumer culture, I have no idea.
Essentially anthing I write, say, and or do in class, for class, or remotely connected to class in anyway they can take for their own property. That's very strange. I go to Cornell and I know for a fact that students essentially retain all copyright rights to their academic work. Look at the section entitled "Student" in the university's copyright policy.
I think the reason "internet law" wasn't necessary in the early days is that the internet wasn't actually used for all that much, and those who used it tended to be scientists and/or people very technically-inclined. You can't bemoan the fact that the internet is now used by the general public, that's just how it's happened. As a result, innovation on the internet boomed in the late '90s, and created a lot of issues which are now being considered. Along with this, general interest in the Internet of course created major commercial interests.
So the important thing to do now is to get educated individuals with a "modicum of internet savvy," as you put it, to represent the public interest to our officials (yes I know that's their job, but if they aren't doing it correctly, we need to help instruct them) in order to counter the often amoral and anti-innovative opinions of the invested commercial interests.
Here's a good example of a cd that came with a nice bonus. The latest Queens of the Stone Age album, Songs For The Deaf, was released with a bonus DVD featuring recorded live performances and clips from the recording studio. The DVD was attached to the first 100,000 copies. To top that off, it was being sold for ~$10 at most stores. So for $10 you got their CD plus a DVD. That's a good package, and perhaps as a result, their album went gold. Their previous album had not even made it onto the Billboard 200.
I guess this doesn't really count, but I know that here at Cornell they're doing a Christmas sale on iPods for students. They're charging $239/$329/$429. The rest of the year you get $30 off the retail price. Just one of the perks of college, of course they make up for it by charging insanely high tuition and attaching fees to everything.
Umm, I think you're mistaken about one thing. The right to create derivative works is a clearly outlined exclusive right given to copyright holders. If you want to make a movie of someone's book, you have to get permission. Under the GPL, the copyright holder gives up the exclusive right to create derivative works, and anyone is allowed to do so, as long as any derivative works are also licensed under the GPL (I'm pretty sure you knew that already). But if you wrote a story that was obviously about LOTR, even if you didn't use the names, they would most likely have grounds to sue for infringement.
Umm, I think if you live in violent neighborhoods, you tend to be relatively poor, and in poor communities there tends to be statistically more violence due to the fact that it goes along with the underground economy. Now, if we're talking about suburban white kids shooting their schools up, that's one thing, but they don't fit #3 anyway. But in general, I do think it is probably the biggest factor in the avg. level of violence in a given community, even moreso than the parents.
See I always thought this was about legal recognition, tax benefits, and making employer's health insurance plans cover the "spouse". I don't think the gay marriage issue is about keeping relationships from falling apart.
I remembger both of those last two games! I mean, everyone I know remembers Oregon Trail and I remember playing that game where you'd identify gems a million times in my 5th grade science class.
And you live on campus? So technically using an instant messenging program is against your school's policy? I've never heard of anything like that. Where do you go to school? Of course, my on-campus network connection here costs me $44 a month (they're finally bundling it with housing next year).
Are you so into the open source community that you think it's overrun the world and there's no proprietary software left? There's still plenty of warez out there.
But isn't an airplane to some degree a Faraday cage, seeing as it's completely encased in conductive metal? Wouldn't it therefore be hard to get good cellphone reception?
Re:OT: from the people that want "under God" remov
on
Can You Hear Me Now?
·
· Score: 1
Actually, yes....how many hikers have died stranded in snowstorms? This was bound to happen once.
Slashdot Mirror
The big difference here is that a backprop model needs to essentially be told what the correct phonemic boundaries are, etc., either in the way it represents the phonemes (possibly as a localist representation) or as target outputs. The model discussed in PNAS is simply given raw speech sounds and learns the categories itself (and how many there are) with no given targets. It's the difference between an unsupervised and supervised model that makes the difference here. In the earliest stages of language learning, I don't think parents correct their children's babbling!
What I find so strange about the whole fuss over Google Book Search is that they're not doiing anything different than they do with their web search. As you should know, if you know anything about copyright, just about anything you create is automatically copyrighted (of course there are some things that are excluded). Certainly, webpages fall under this category. Thus essentially every webpage on the internet is automatically copyrighted. When Google crawls the web, they add updated pages to their cache. In doing so they are creating a copy of the page itself with which to analyze links, etc, as well as offering it up to the world under the "Cached" link. Site owners have the choice of opting out of Google's system, though few choose to do so. This is exactly what they are doing with the book search. The only difference is that website owners don't haved an industry front group to fund lawsuits.
Now, one could argue that the opt out system is not in keeping with the Copyright Act, whereby creators should choose to allow their use of their works, not have to make a point of disallowing it. If that argument holds in the Book Search case, however, then it certainly should hold for Google's web search.
I do think that many (though probably not all) of those textbook writers would agree that the brain itself does not operate in a discrete manner on a structural level. It's just that they don't care. Their argument rests on the hypothesis that the relevant mental and behavioral states are best described symbolically. Now, much of the evidence I've seen recently seems to refute this belief (of course, I hardly have an unbiased background based on who teaches the classes i've taken and those with whom I've done research). But it's just important for those reading this article to understand that the symbolic diehards are making a distinction between the anatomical brain structure and the functions they believe that structure instantiates. The connectionists and to a greater extent dynamicists (to break everyone up into schools) see structure and function as more tightly linked, and in general just see more similar processes at work in different areas of cognition.
I also think it's short-sighted for Birge to argue that there's no point whatsoever to studying biological systems from a higher-level perspective. His statement that we need to fully understand a worm brain before tackling any apparent aspects of a human one seems akin to arguing that we need to have a complete understanding of particle physics in order to build a basic machine. Biology itself serves as an even more useful example. The general scientific study of living things certainly preceded molecular biology and produced wortwhile findings, and similarly Darwin far preceded the discovery of DNA, etc. Reductionism is an extremely useful methodology, but often it needs to be informed by a higher-level perspective. That's why the field of psychology even exists. If Birge believes that we should ignore all high-level concepts and start our understanding completely from the lowest level neuronal networks, then he's espousing eliminative materialism, and the problem with that philosophical outlook is that it's extremely hard to explain anything when you've thrown out all the concepts that are purported to need explaining.
I'm working on the steering control program for our DARPA Grand Challenge vehicle, and it does that, too. Doesn't mean it's not "digital".
Clearly in the case of the DARPA challenge, the underlying physical machine is a digital processor. But I'm wondering, what type of algorithms does your control program use? The reason I ask is that the question Spivey's considering here isn't whether the physical processing mechanism resembles a digital computer, as clearly the anatomical brain structure itself does not, but instead whether the functional mechanisms (which in your example are directly defined as the algorithms instantiated by your code) are compatible with the serial and symbolic models posited in the classic era of AI as well as cognitive psychology.
In other words, that a computer program could model these functions is for the most part logically trivial and in fact Spivey employs a computational model in the article itself in order to test whether his theory accounts for the data. Similarly, many modern machine learning algorithms are qualitatively "un-digital" even though they are being modeled on a digital computer (hence why they require so much processing power). Do you see the distinction?
Actually, in the PNAS journal article that the Cornell News story refers to, Professor Spivey compares the data he obtained in the mouse-movement experiment to a computer simulation that uses both the TRACE model of spoken-word recognition (an old-school interactive activation model that still seems to be pumping out good data, especially for a localist model) and his own normalized recurrence attractor network (another localist statistical model). Both of these models, as well as the combined one that he creates for the simulations, are by their very nature dynamical systems models, and definitely not Turing machines.
Not totally coincidentally, Spivey is also my undergrad psych major advisor.
I'd have said "RTFA", but since slashdot doesn't directly link to the journal article PNAS, I guess you're kind of off the hook.
Let me guess; you're a college professor.
What's with the anti-academic attitude so prevalant with conservatives? Didn't you guys used to love Milton Friedman?
The road to being a professor is hardly one paved with gold. Grad school takes at least 5 years, during which you are expected to TA, work with professors on their research and find time to do your own research toward your dissertation, while being given about $15000 to live on per year. After that you have to spend another few years at least as a post-doc, being paid perhaps $35000/year. By the time you're even actually a faculty member (easily not until you're in your early to mid 30s), you'll make between $40,000-$65000, and that's only for those who get full-time faculty positions at major research universities (certainly the minority among PhDs).
Professors aren't the fat-cats. I'm sure you can point out some very high-paid professors, because they certainly exist, but they're definitely not in the majority.
I'll have to check it out further later, but at the moment I can't handle all that functionalism.
I swear they're breeding me to be a total connectionist around here.
Linking music and language is hardly a new idea. The language analogy has been the focus of much of the research on music cognition in cognitive science in general. For one of the most direct examples of such a paradigm, look at Lerdahl & Jackendoff's generative grammar for Western tonal music, which sought to apply Chomskyan linguistic theory (hotly debated itself) to musical structures. While the syntax can be dealt with rather easily, pinning down a concept of "musical semantics" has been rather elusive. And to think Chomsky said, "If you take care of the syntax, the semantics takes care of itself." Not only is this not the case in music, a lot of people don't think it's the case in language anymore either. Ok this is the best I can think of for 3 AM. Maybe I'll try again later.
That sounds pretty misinformed. Have you ever filled out a financial aid application? The most important part of it is your family income, not your savings. Though if you have a great deal of assets saved for college they will expect you to use them, the actual EFC (expected family contribution) is calculated mostly from your adjusted gross income on your tax return. So what that has to do with a consumer culture, I have no idea.
Essentially anthing I write, say, and or do in class, for class, or remotely connected to class in anyway they can take for their own property.
That's very strange.
I go to Cornell and I know for a fact that students essentially retain all copyright rights to their academic work. Look at the section entitled "Student" in the university's copyright policy.
I think the reason "internet law" wasn't necessary in the early days is that the internet wasn't actually used for all that much, and those who used it tended to be scientists and/or people very technically-inclined. You can't bemoan the fact that the internet is now used by the general public, that's just how it's happened. As a result, innovation on the internet boomed in the late '90s, and created a lot of issues which are now being considered. Along with this, general interest in the Internet of course created major commercial interests.
So the important thing to do now is to get educated individuals with a "modicum of internet savvy," as you put it, to represent the public interest to our officials (yes I know that's their job, but if they aren't doing it correctly, we need to help instruct them) in order to counter the often amoral and anti-innovative opinions of the invested commercial interests.
Here's a good example of a cd that came with a nice bonus. The latest Queens of the Stone Age album, Songs For The Deaf, was released with a bonus DVD featuring recorded live performances and clips from the recording studio. The DVD was attached to the first 100,000 copies. To top that off, it was being sold for ~$10 at most stores.
So for $10 you got their CD plus a DVD. That's a good package, and perhaps as a result, their album went gold. Their previous album had not even made it onto the Billboard 200.
I guess this doesn't really count, but I know that here at Cornell they're doing a Christmas sale on iPods for students. They're charging $239/$329/$429. The rest of the year you get $30 off the retail price.
Just one of the perks of college, of course they make up for it by charging insanely high tuition and attaching fees to everything.
Umm, I think you're mistaken about one thing. The right to create derivative works is a clearly outlined exclusive right given to copyright holders. If you want to make a movie of someone's book, you have to get permission. Under the GPL, the copyright holder gives up the exclusive right to create derivative works, and anyone is allowed to do so, as long as any derivative works are also licensed under the GPL (I'm pretty sure you knew that already). But if you wrote a story that was obviously about LOTR, even if you didn't use the names, they would most likely have grounds to sue for infringement.
Umm, I think if you live in violent neighborhoods, you tend to be relatively poor, and in poor communities there tends to be statistically more violence due to the fact that it goes along with the underground economy. Now, if we're talking about suburban white kids shooting their schools up, that's one thing, but they don't fit #3 anyway. But in general, I do think it is probably the biggest factor in the avg. level of violence in a given community, even moreso than the parents.
Well I do know that you are offered a "verifiable paper reciept." :)
See I always thought this was about legal recognition, tax benefits, and making employer's health insurance plans cover the "spouse". I don't think the gay marriage issue is about keeping relationships from falling apart.
I remembger both of those last two games! I mean, everyone I know remembers Oregon Trail and I remember playing that game where you'd identify gems a million times in my 5th grade science class.
And you live on campus? So technically using an instant messenging program is against your school's policy? I've never heard of anything like that. Where do you go to school? Of course, my on-campus network connection here costs me $44 a month (they're finally bundling it with housing next year).
Don't tell me you're one of those idiots who thinks the moon landing was faked.
Are you so into the open source community that you think it's overrun the world and there's no proprietary software left? There's still plenty of warez out there.
But isn't an airplane to some degree a Faraday cage, seeing as it's completely encased in conductive metal? Wouldn't it therefore be hard to get good cellphone reception?
Actually, yes....how many hikers have died stranded in snowstorms? This was bound to happen once.
Yes i was hoping someone would say that, it's just more fun to be trollish.
Here's some news, everyone lip-syncs in music videos, jerkoff, that's just the way it works.