The interesting thing is that in our last federal election, they counted all the paper and pencil ballots the same night, and then recounted several jurisdictions that same night.
I truly cannot figure out why the recount of a few jurisdictions in Florida required so much ridiculous work. A recount here, of paper ballots, can happen in a few hours. A recount in Florida takes weeks?! Either it was intentionally made impossible, or the system is just *#@!ed.
>People that buy Macs for general use are in general people who don't care about cost, or cost/benefit ratios. >If they did, they wouldn't buy a Mac in the first place
or maybe they just have a different idea of what is a 'benefit' than you do. when I bought my first mac, i counted the fact that it didn't crash at least once every day as a benefit worth quite a bit of money. perhaps you don't. (this was before whatever windows version is now available that allegedly reduces crashes considerably. thank god i don't have to find out whether that's true.)
As I took the proposal, the idea was that users become *members* of the corporation, not employees. This would require some sort of formal shareholding arrangement.
Unfortunately, this difference doesn't make the proposal any more legal. By this argument, all the shareholders of a company which had purchased a copy of Microsoft Office could use that copy on their home computers *simultaneously*.
Even if it were restricted to one user at a time (which significantly reduces the utility of the proposal), this would not fly. While the license restrictions on software are more stringent than on music, when you purchase a CD you are buying a license to play the content contained on it---you are not actually buying the content. More specifically, you are buying a *single user* license.
So, two problems (there may be others) with "The Idea":
1. The fact that a corporation is a single legal person doesn't obliterate the individual personhood of all its members. If the members listen to the music individually, it is not the legal equivalent of only the corporation listening to the music (whatever that might mean).
2. The corporation does not own the content. It owns a license to play (and possibly store) the content. The corporation cannot serve the content to many persons simultaneously, even if they are members, employees, or well, anybody.
Um, people... this is a joke. This article is *satire*. (and even if it's not, it's pretty dang funny.) However, a not to slashdot editors would be appropriate: this belongs in humour (the disembodied foot thing) not in science (the disembodied head thing)
Perhaps you mean _Consciousness Explained_. Anyway, I don't think that believing in the possibility of AI indicates a lack of understanding of the current debate. Perhaps what was wanted was a more substantive picture of HOW AI might be possible. I haven't read the book, and i'm just going from the content of the comments in this thread, but personally, if I read an 800 page book whose conclusion was AI is possible, I too would be disappointed. If Hofstader did more than that, good for him. If he didn't, then I would suspect the book is indeed 'masturbatory'.
And I think it would be worthwhile to avoid the Dennett in any case. He's a skilfull metaphor crafter, but not philosophically very deep.
this is an interesting question, which actually ties into a lot of what elman has to say about non-representational innateness... there -are- ways to get interesting constraints on development which don't involve prespecification of synaptic connections (what elman calls architectural and chronotopic constraints, as opposed to the representational constraints which pinker and others seem to want). In this way, representations can be acquired through minimal genetic representations which -interact- with information present in the environment, or with physical laws, etc. to produce effects which contain much more information than the genome by itself. But the environmental interaction necessary here is just the sort of thing that people like pinker want to deny; what's innate here is the method of development, not what's developed. Now, the question i take it you're asking is whether there isn't possibly some extremely compressed form in which to store the information in the genes--eg. the fractal equation. I'm not sure exactly how plausible this is; one question is whether the synaptic connection specification that actually gets you the results you want can be specified with that type of thing. This is an empirical question (and i'm inclined to be sceptical until there's some kind of demonstration; though i certainly wouldn't rule it out), but i think one of the strength's of elman's account is that we no longer -need- this amount of information. the architectural and chronotopic constraints do all the work; we don't need to specify the representations. So, if elman is right, that type of information compression would be superfluous. Though that isn't to say wrong; all this is just speculation on my part.
Stephen Pinker is notorious in many circles for his sometimes wild speculation about innate knowledge, predispositions, reactions, etc. (cf. his recent New York Times article explaining exactly why, in evolutionary terms, women might discard newborn babies (as has happened in a few very publicised cases).) There are a few problems with the arguments used to support all of this innateness (or:nativism). Most importantly, it's generally based on questionable assumptions about the data (e.g. the language samples to which children are exposed) and the power of the learning device (the brain) which has to extract patterns from the data. Jeffrey Elman et. al. present a very interesting criticism of this common nativist argument in "Rethinking Innateness: A Connectionist Perspective on Development". Basically, they show that many of the functions that are supposedly impossible, and many of the strange patterns in learning which supposedly can only be accounted for by innate predispositions and knowledge can actually be simulated very well in neural networks which don't build in -any- knowledge of the subject area (e.g., they aren't born knowing Universal Grammar, yet Elman provides an example of a network which has managed to sort linguistic data according to grammatical category. This ability was not programmed in--it was independently learned by the network). Of course, all this is preliminary--we certainly don't have neural nets that can simulate human behaviour in total, or even close. But the assumption that we -need- innate knowledge to account for this type of stuff just doesn't fly.
An interesting point that Elman et. al. mention is that the genome just doesn't contain enough information to specify brain representations--it's not possible for this sort of knowledge to be encoded in the genes. 'Wiring of the microcircuitry'--i.e. specification of synaptic connections--'is essential if language, the quintessential higher cognitive process, is an instinct...' (Pinker, The Language Instinct, 93, 97). But, genetic prespecification of synaptic weights in the brain is just too much information for genes to carry--such specification could reasonably require on the order of 10 trillion base pairs of DNA just for our brain--and that's more than we have for our whole body.
Elman et.al. make another major contribution to the debate by describing an interactionist framework, where outcomes may be highly constrained and universal, but not themselves directly contained in the genes (or the brain) in any domain-specific way. (Notice that this is exactly the faulty inference of which Pinker and many other nativists are guilty.)
In sum, I think that Rethinking Innateness is an extremely valuable contribution to the old nativist/empiricist debate, most especially for their dissection of the concept 'innateness', which nativists use with abandon, but which really is ambiguous between a large number of possible interpretations.
Slashdot Mirror
Here in Canada it's paper and pencil, too.
The interesting thing is that in our last federal election, they counted all the paper and pencil ballots the same night, and then recounted several jurisdictions that same night.
I truly cannot figure out why the recount of a few jurisdictions in Florida required so much ridiculous work. A recount here, of paper ballots, can happen in a few hours. A recount in Florida takes weeks?! Either it was intentionally made impossible, or the system is just *#@!ed.
>People that buy Macs for general use are in general people who don't care about cost, or cost/benefit ratios.
>If they did, they wouldn't buy a Mac in the first place
or maybe they just have a different idea of what is a 'benefit' than you do. when I bought my first mac, i counted the fact that it didn't crash at least once every day as a benefit worth quite a bit of money. perhaps you don't. (this was before whatever windows version is now available that allegedly reduces crashes considerably. thank god i don't have to find out whether that's true.)
Unfortunately, this difference doesn't make the proposal any more legal. By this argument, all the shareholders of a company which had purchased a copy of Microsoft Office could use that copy on their home computers *simultaneously*.
Even if it were restricted to one user at a time (which significantly reduces the utility of the proposal), this would not fly. While the license restrictions on software are more stringent than on music, when you purchase a CD you are buying a license to play the content contained on it---you are not actually buying the content. More specifically, you are buying a *single user* license.
So, two problems (there may be others) with "The Idea":
1. The fact that a corporation is a single legal person doesn't obliterate the individual personhood of all its members. If the members listen to the music individually, it is not the legal equivalent of only the corporation listening to the music (whatever that might mean).
2. The corporation does not own the content. It owns a license to play (and possibly store) the content. The corporation cannot serve the content to many persons simultaneously, even if they are members, employees, or well, anybody.
Um, people... this is a joke. This article is *satire*. (and even if it's not, it's pretty dang funny.)
However, a not to slashdot editors would be appropriate: this belongs in humour (the disembodied foot thing) not in science (the disembodied head thing)
Perhaps you mean _Consciousness Explained_. Anyway, I don't think that believing in the possibility of AI indicates a lack of understanding of the current debate. Perhaps what was wanted was a more substantive picture of HOW AI might be possible. I haven't read the book, and i'm just going from the content of the comments in this thread, but personally, if I read an 800 page book whose conclusion was AI is possible, I too would be disappointed.
If Hofstader did more than that, good for him. If he didn't, then I would suspect the book is indeed 'masturbatory'.
And I think it would be worthwhile to avoid the Dennett in any case. He's a skilfull metaphor crafter, but not philosophically very deep.
this is an interesting question, which actually ties into a lot of what elman has to say about non-representational innateness... there -are- ways to get interesting constraints on development which don't involve prespecification of synaptic connections (what elman calls architectural and chronotopic constraints, as opposed to the representational constraints which pinker and others seem to want). In this way, representations can be acquired through minimal genetic representations which -interact- with information present in the environment, or with physical laws, etc. to produce effects which contain much more information than the genome by itself. But the environmental interaction necessary here is just the sort of thing that people like pinker want to deny; what's innate here is the method of development, not what's developed.
Now, the question i take it you're asking is whether there isn't possibly some extremely compressed form in which to store the information in the genes--eg. the fractal equation. I'm not sure exactly how plausible this is; one question is whether the synaptic connection specification that actually gets you the results you want can be specified with that type of thing.
This is an empirical question (and i'm inclined to be sceptical until there's some kind of demonstration; though i certainly wouldn't rule it out), but i think one of the strength's of elman's account is that we no longer -need- this amount of information. the architectural and chronotopic constraints do all the work; we don't need to specify the representations. So, if elman is right, that type of information compression would be superfluous. Though that isn't to say wrong; all this is just speculation on my part.
Stephen Pinker is notorious in many circles for his sometimes wild speculation about innate knowledge, predispositions, reactions, etc. (cf. his recent New York Times article explaining exactly why, in evolutionary terms, women might discard newborn babies (as has happened in a few very publicised cases).)
There are a few problems with the arguments used to support all of this innateness (or:nativism). Most importantly, it's generally based on questionable assumptions about the data (e.g. the language samples to which children are exposed) and the power of the learning device (the brain) which has to extract patterns from the data. Jeffrey Elman et. al. present a very interesting criticism of this common nativist argument in "Rethinking Innateness: A Connectionist Perspective on Development". Basically, they show that many of the functions that are supposedly impossible, and many of the strange patterns in learning which supposedly can only be accounted for by innate predispositions and knowledge can actually be simulated very well in neural networks which don't build in -any- knowledge of the subject area (e.g., they aren't born knowing Universal Grammar, yet Elman provides an example of a network which has managed to sort linguistic data according to grammatical category. This ability was not programmed in--it was independently learned by the network). Of course, all this is preliminary--we certainly don't have neural nets that can simulate human behaviour in total, or even close. But the assumption that we -need- innate knowledge to account for this type of stuff just doesn't fly.
An interesting point that Elman et. al. mention is that the genome just doesn't contain enough information to specify brain representations--it's not possible for this sort of knowledge to be encoded in the genes. 'Wiring of the microcircuitry'--i.e. specification of synaptic connections--'is essential if language, the quintessential higher cognitive process, is an instinct...' (Pinker, The Language Instinct, 93, 97). But, genetic prespecification of synaptic weights in the brain is just too much information for genes to carry--such specification could reasonably require on the order of 10 trillion base pairs of DNA just for our brain--and that's more than we have for our whole body.
Elman et.al. make another major contribution to the debate by describing an interactionist framework, where outcomes may be highly constrained and universal, but not themselves directly contained in the genes (or the brain) in any domain-specific way. (Notice that this is exactly the faulty inference of which Pinker and many other nativists are guilty.)
In sum, I think that Rethinking Innateness is an extremely valuable contribution to the old nativist/empiricist debate, most especially for their dissection of the concept 'innateness', which nativists use with abandon, but which really is ambiguous between a large number of possible interpretations.