cool as this may be, it's definitely not the first singing synthesis software. CNMAT at berkeley had a neural-net based additive synthesis engine in the mid 90's that did a pretty good human voice (it could even reproduce the voices of specific individuals, as I recall), and did other instruments as well (a mean viola).
I can't find a link to an actual demo of it simulating a human voice, but here's a page that documents its use to reproduce the sound of a suling (javanese wooden flute). Does a good job too. I've heard it demo'd with a human voice, and it was pretty good (though the neural net needed additional input - the syllable being sung - obviously).
i'm sure that many of the other academic computer music labs around the world had similar software long before yamaha introduced this package. still cool, though.
1. open com.apple.mail.plist and backup 2. expand "root" 3. expand "DeliveryAccounts" 4. expand "0", "1", etc 5. find the account you want to delete 6. select the corresponding array number 7. press the delete button 8. save as com.apple.mail.plist
just did it on my machine (running Jaguar), works a charm.
Here's the thing though - the difference in quality between live performance and *any recorded medium* far exceeds the difference in quality between CD and mp3 (assuming reasonable bitrate) or whatever.
I have no difficulty picking out the difference between mp3 and CD, between CD and 192kHz/48bit, and between 192kHz/48bit and live music (by spatial imaging, not quality, obviously). I suspect that this is true of most people who work with music on a day to day basis, if not the broader population.
My point is more that most people, myself included, just don't care. I can tell that 192k/sec mp3 isn't CD quality, but I don't mind - if I like the music, I like the music. I'd like it if it was played through a tin can. I admit that if I listen to 128k/sec for a *long* time (like, maybe 2 hours), at loud volume levels, I get fatigued. So I use 192k/sec, and it doesn't bother me in the slightest. I *know* that the quality isn't perfect, but that doesn't concern me; I have better things to focus my listening on.
Yeah, but some of us can hear sound above 22kHz. And even those who can't can still perceive fast transients at higher frequencies.
The human ear doesn't just do fourier analysis. Something similar, but it's a lot more complicated. Thus the "audible band" is more nuanced then a mere numerical frequency range.
Are CD's lossless, even within the bounds of human perception? no. But are they "good enough" for 99.999% of the population? yes. For that matter, so is mp3/aac/ogg/whatever is in style this month.
Re:There is one important limitation.
on
Digital 35mm SLRs?
·
· Score: 1
It's mostly silly to worry about this sort of manipulation of an audio signal -- almost nobody wants audio "enlargements" (what would that even be? playing it really slowly at high volume?) or "zoomed crops", fractally zooming in on just a sliver of a snippet of sound.
Except that we commonly do these things with audio; tempo shifting, compression, mixing, intonation-matching, equalizing - all of these things correspond to applying a scaling function to one or more dimensions of the input sound signal.
That's why people record at 192kHz - the audio equivalent of large format - to be able to apply more of these transformations before you get substantial signal degradation (and for other reasons, sure, but this is a biggie).
of course you can tell the difference between titanium, steel, carbon frames. (at least I've always found it pretty clear, without looking at the stickers). hell, there's a noticable difference between different aluminum frames. ride a specialized frame, then a cannondale, and tell me you don't notice a difference.
me, i wouldn't ride either one. i've got a 10 year old carbon frame that's buttery-smooth on the beaten up roads where i live:?)
now, as for audiofiles, full of shit, every last one. but i think you might find more than one golden ear among those 1000 professional engineers.
yeah, but scientific computation *does* consist entirely of multiply-adds (ish). How do you computationally solve a problem? You find approximations in some convenient linear space - you reduce the problem to a whole lot of linear algebra problems. And linear algebra problems are solved by a whole lot multiply-adds.
i'm being incredibly simplistic here. really mindbogglingly simplistic. and anyway, why bother to speculate when there will be real numbers in a month?
"I can't believe that people still view politics on the one dimensional axis of economically conservative-economically liberal."
If only that were true... problem is, people now (by and large) view politics on a spectrum of socially conservative - socially liberal. Thus "conservatives" are willing to vote for people like the damn idiot in the white house, ignoring the fact that he's the most fiscally-irresponsible big-government wanker of them all, merely because of his conservative social positions, which really, he has little to no power with regards to. (uh, check that: is *supposed* to have little to no power over... now if congress would only stop bending over to give the fuck whatever he wants.)
different from anything you've every used: not if you get out much:?)
once i figured out the basic idea (about 2 minutes), i found the interface makes perfect sense, and is quite easy to use. could be better, of course, but perfectly adequate, and better than most in some ways.
I'm curious as to your standard for "full featured" - aside from a couple of input/playback tricks, I've never seen finale or sibelius do that Lime can't. On the other hand, I know of many things that are trivial in Lime but take an absurd amount of work to recreate in finale. (I'm less familiar with Sibelius).
For just plain-old putting notes and text on a page, I'll use Lime over anything else. In fact, for almost anything, I'll use Lime over anything else. And for someone who doesn't have a copy of Finale or Sibelius lying around, you can't argue with the unlimited-trial-period shareware nature of Lime.
Lime is (approximately speaking) god's gift to the composer. It lacks some of the nifty input features of Finale, and it has less sophisticated playback. But, when it comes to laying out scores, and even composing, there's nothing that I'd rather use. It's wonderfully customizable (many 20th century notation conventions are much easier than in finale, you can use finale's alternate fonts, you can make it do damn near anything you want), and *it's free*.
Let me repeat that: ***it's free***
Finale sells for what, over $200, maybe $300 these days? Short of nifty input and playback tricks that most people never use, I'm yet to find anything that Finale or Sibelius can do that Lime can't. And it's OS X native. And stable.
Sorry if this is a bit over the top, but I really really love this program, and have wasted (in retrospect) many hours on finale. *maybe* if I was writing very large scale works I'd prefer finale, but I've written an entire musical in Lime without any problems.
did I *ever* say that I though LZW deserved patent protection? no. LZW, "1 click ______", etc, are exactly the sort of software that shouldn't be patentable. Just as some equally trivial physical invention isn't patentable. I think I made it perfectly clear that this was my opinion in my earlier posts.
your out-of-hand dismissal of the very idea of patents, just because of some horrendous examples, however, deeply troubles me. What the situation calls for is better training for patent examiners, coupled with a higher bar for what constitues a patentable invention. Not elimination of the patent system altogether. As I explained in my original post, elimination of the patent system would deal a major blow to research, and slow down (not speed up) the rate at which algorithms become available to the public.
The patent system is certainly broken. This is the fault of lawers who will seek a patent on anything and a lack of funding to the patent office, etc. But the situation can be remedied without throwing out the baby with the bathwater.
I assume that you mean to say that Free Software *couldn't* use my algorithm, if I made it free for non-commercial use only. You seem to have missed the fact that the GPL is far from the only licensing scheme available.
Please, don't confuse the Free Software Foundation with the open source community as a whole. There's a tremendous bulk of open source software released under exactly the licensing model I described, especially from academic researchers, who would be in exactly the position that I described.
Don't get me wrong. I'm a big fan of the FSF, and I whole-heartedly support it's goals. But the GPL is pretty clearly intended to cover exactly the sort of code that I assigned to the "isn't / shouldn't be patentable" category. Algorithms that truly represent legitamite scientific advances should be patentable, and clearly if patented, they wouldn't be released under the GPL (in all likelihood). However, they could very easily be released under the licensing scheme I describe, unassociated with the FSF.
Software patents might be incompatible with Free Software, but they are *not* inherently incompatible with open source.
Of course algorithms should be patentable - suppose I come up with a great new video codec tomorrow, much better than anything else, but I can't patent it, I can only copyright the source code.
Now, I'm faced with a choice. I could develop closed-source software implementing the codec, and refrain from publishing my new algorithm, thus protecting my innovation with the copyright I'm allowed - if I do this, I'll probably make myself a tidy profit (assuming I know someone who knows the least bit about marketing), and be quite happy with the situation.
Or, I could publish the details of my algorithm, ensuring that no one will be willing to pay for my implementation (well, almost no one) - I can't get paid for my innovation, and the companies that already dominate the market will get my R&D efforts for free. They'll put their own implementation into their products, and pocket the proceeds.
Let's say I'm a noble minded researcher; I don't really care whether or not I get rich off my invention - on the other hand, I'm not stupid, I don't want anyone else getting rich off of it if I'm not getting some of the pie. Clearly, I'm going to keep my source closed, I'm not going to publish. This prevents my algorithm from being used in open-source products, blocks other researchers from extending and improving my results and generally holds back progress.
On the other hand, suppose that I can patent my great new codec. Now I have a third option. I can patent it, and set up a reasonable licensing scheme: you're free to implement my for private non-commercial use, research, etc. If you're getting paid from your implementation, then I want a cut of it, too. Now I can publish, its possible for the open-source and academic communities to use my great new invention, it's available for more research work, and at the same time, I can prevent other people from getting rich off of my work without also compensating me. This is *exactly* why patents exist: to allow people to profit from their work without impeding the flow of "progress".
Now, am I a loony git who thinks that *any* algorithm should be patentable? of course not. There's a reasonable standard, and it's illustrated perfectly by my previous example. An algorithm should be patentable only if the difficulty/effort in creating the algorithm sufficiently exceeds the difficulty/effort in implementing the algorithm.
Why this standard? If the "implementation cost" far exceeds the "invention cost", then no one's going to want to use their own implementation; they'll happily pay for mine, and a mere copyright on the source will suffice to protect me. Furthermore, the fact that the implementation cost far exceeds the invention cost is a strong indication that the algorithm in question fails to qualify as something that most people in the field wouldn't have thought up in the same situation (this should be a necessary standard for ALL patents - they should be INNOVATIVE).
On the other hand, if the "invention cost" exceeds the "implementation cost", then everyone else will develop their own implementations rather than use mine if the algorithm itself isn't protected; Conversely, since the invention cost/difficulty/effort/etc was so substantial, my invention is exactly the sort of thing that should be protected as innovative - something that the average person in the field wouldn't have thought of.
It should be noted that this requirement that the cost of invention far exceed the cost of implementation would actually eliminate the vast, vast bulk of software patents - which is a good thing. Patents such as "one-click ___", where you see it in operation without any knowledge of its guts and immediately know how it works, or the marching cubes algorithm (patented by HP, I believe), which is just what any sensible person with some background in computational geometry or algorithms would do, without much thought, should be gotten rid of. In all likelihood, this standard should be applied across the board, not just in software/algorithms. But should patents be done away with entirely, even in a restricted field? Of course not, and I think my example establishes that pretty clearly.
actually, you can get your license suspended - it's at the officer's discretion; if he thinks you were being truely reckless, he'll probably suspend your license. If, on the other hand, he thinks that you were merely stupid to have been caught, he'll just write you a ticket.
This is in CA - I know that there's a similar law in maine (20mph over = suspension), but the cops there are a lot happier to just take away your damn license, especially if you're from out of state (happened to a friend of mine).
Did you notice down at the bottom of the page where they compared the *actual* realized bit rate of each algorithm? 64kbit ogg turns out to be about 67-68kbit, once you factor in the header data, etc.
So, yes, the test took file size into consideration, and no, this doesn't make ogg look any better.
I'm a graduate student in math; while my preference in for pencil & paper, I *do* have my powerbook & wacom tablet set up for taking notes, for when I don't have pencil & paper, or I need to make an electronic copy anyway.
1/2 of the screen is an text window (handwriting recognition), the other half a window to draw the equations in. A scripted button to save the equations as pictures and drop them into the text file. Later on, I just go through and type the equations in by hand. Saves a whole lot of time over typing the whole thing from scratch.
Of course, it helps that I have very precise handwriting to start with, so I get near perfect performance from the handwriting recognition software. When my girlfriend tries to use the system, it doesn't go so well.
But... required computers in 7th grade? huh? why? hell, if I had my way, there wouldn't be calculators in high school. Maybe a computer science class, but most highschool CS classes are crap, and would probably be improved by getting rid of the computers (i.e. then they might actually learn some CS instead of writing "hello world" in pascal, then playing counterstrike all day - not that that isn't a valuable skill, per se)
or, rather, all the 3D display technologies I've seen so far, is that there's no "hidden line removal", so to speak. Every technology i've seen is inherently transparent, and uses some means to generate perceived light sources within a volume.
Unfortunately, the human optical system isn't really built to deal with this on a regular basis; we expect *most* things to be at least somewhat opaque, and have a considerably easier time processing visual information that adheres to those expectations. So what's really needed is a way to not only change the color of a voxel, but also it's opacity; basically an "alpha channel". (You can't just do old school hidden line (surface) removal because you don't know where the observer is).
Clearly, this is impossible with any of the spinning disc/helicoid techniques; with some of the other ideas (like crystal activated by non-visible-wavelength, etc) it seems like it should be possible; use one wavelength to produce light, another to turn pixels opaque. Make the interior of an "object" opaque, illuminate the boundary, and you've got a display that's much easier for the human visual system to process.
Prediction: until this happens, no real 3-D displays except for highly specific industry applications.
Amen. I'm in applied mathematics, doing complicated numerical simulations, and I'm also a musician - I use my computer as a real-time effects processor. Both very computation-intensive tasks. Do I use the fastest CPU? heck no. I have a 3 year old G3 Powerbook, running at 400mhz.
Why? Because it's bombproof. Completely stable, entirely predictable, and has never given me the slightest problem. I hate to imitate the stupid commercials, but "it just works". It's fast enough for everything except *serious* numerical simulations; things "ordinary" users will never ever throw at their computer. And for those, well, no desktop system is fast enough, so it's a moot point to compare.
Are there people with nice stable Windows or Linux installations? of course. But a much bigger percentage of mac users I've talked to have a problem-free life than on any other platform. For me, the productivity and piece of mind gains from having a completely stable system more than offset the productivity losses of having a 3 year old 400 mhz processor (which, let's face it are minimal - the vast bulk of computer time is spent waiting for user response, not system response).
** of course, if you're someone who just needs to run the newest games at the highest framerate, you might be an exception. me, i'll stick to using my ps/2 or gamecube for gaming and be perfectly happy.
Slashdot Mirror
I can't find a link to an actual demo of it simulating a human voice, but here's a page that documents its use to reproduce the sound of a suling (javanese wooden flute). Does a good job too. I've heard it demo'd with a human voice, and it was pretty good (though the neural net needed additional input - the syllable being sung - obviously).
i'm sure that many of the other academic computer music labs around the world had similar software long before yamaha introduced this package. still cool, though.
... hypnosis?
unlikely. the ez-pass system was designed to be able to handle cars moving at over 100mph.
1. open com.apple.mail.plist and backup
2. expand "root"
3. expand "DeliveryAccounts"
4. expand "0", "1", etc
5. find the account you want to delete
6. select the corresponding array number
7. press the delete button
8. save as com.apple.mail.plist
just did it on my machine (running Jaguar), works a charm.
Oh, I don't deny that at all.
Here's the thing though - the difference in quality between live performance and *any recorded medium* far exceeds the difference in quality between CD and mp3 (assuming reasonable bitrate) or whatever.
I have no difficulty picking out the difference between mp3 and CD, between CD and 192kHz/48bit, and between 192kHz/48bit and live music (by spatial imaging, not quality, obviously). I suspect that this is true of most people who work with music on a day to day basis, if not the broader population.
My point is more that most people, myself included, just don't care. I can tell that 192k/sec mp3 isn't CD quality, but I don't mind - if I like the music, I like the music. I'd like it if it was played through a tin can. I admit that if I listen to 128k/sec for a *long* time (like, maybe 2 hours), at loud volume levels, I get fatigued. So I use 192k/sec, and it doesn't bother me in the slightest. I *know* that the quality isn't perfect, but that doesn't concern me; I have better things to focus my listening on.
Yeah, but some of us can hear sound above 22kHz. And even those who can't can still perceive fast transients at higher frequencies.
The human ear doesn't just do fourier analysis. Something similar, but it's a lot more complicated. Thus the "audible band" is more nuanced then a mere numerical frequency range.
Are CD's lossless, even within the bounds of human perception? no. But are they "good enough" for 99.999% of the population? yes. For that matter, so is mp3/aac/ogg/whatever is in style this month.
Except that we commonly do these things with audio; tempo shifting, compression, mixing, intonation-matching, equalizing - all of these things correspond to applying a scaling function to one or more dimensions of the input sound signal.
That's why people record at 192kHz - the audio equivalent of large format - to be able to apply more of these transformations before you get substantial signal degradation (and for other reasons, sure, but this is a biggie).
of course you can tell the difference between titanium, steel, carbon frames. (at least I've always found it pretty clear, without looking at the stickers). hell, there's a noticable difference between different aluminum frames. ride a specialized frame, then a cannondale, and tell me you don't notice a difference.
:?)
me, i wouldn't ride either one. i've got a 10 year old carbon frame that's buttery-smooth on the beaten up roads where i live
now, as for audiofiles, full of shit, every last one. but i think you might find more than one golden ear among those 1000 professional engineers.
yeah, but scientific computation *does* consist entirely of multiply-adds (ish). How do you computationally solve a problem? You find approximations in some convenient linear space - you reduce the problem to a whole lot of linear algebra problems. And linear algebra problems are solved by a whole lot multiply-adds. i'm being incredibly simplistic here. really mindbogglingly simplistic. and anyway, why bother to speculate when there will be real numbers in a month?
no, you're thinking of maine. the state of maine bought laptops for every 7th grade student, if i remember correctly.
"I can't believe that people still view politics on the one dimensional axis of economically conservative-economically liberal."
If only that were true... problem is, people now (by and large) view politics on a spectrum of socially conservative - socially liberal. Thus "conservatives" are willing to vote for people like the damn idiot in the white house, ignoring the fact that he's the most fiscally-irresponsible big-government wanker of them all, merely because of his conservative social positions, which really, he has little to no power with regards to. (uh, check that: is *supposed* to have little to no power over... now if congress would only stop bending over to give the fuck whatever he wants.)
uhhh... you're joking, right?
if you don't know what xfig is, you're much happier than you otherwise would be. jeez, i'd sooner just write postscript by hand.
my $.02: xfig is an incredible chore to use. (i really do just write postscript instead)
different from photoshop, yes.
:?)
different from anything you've every used: not if you get out much
once i figured out the basic idea (about 2 minutes), i found the interface makes perfect sense, and is quite easy to use. could be better, of course, but perfectly adequate, and better than most in some ways.
I'm curious as to your standard for "full featured" - aside from a couple of input/playback tricks, I've never seen finale or sibelius do that Lime can't. On the other hand, I know of many things that are trivial in Lime but take an absurd amount of work to recreate in finale. (I'm less familiar with Sibelius).
For just plain-old putting notes and text on a page, I'll use Lime over anything else. In fact, for almost anything, I'll use Lime over anything else. And for someone who doesn't have a copy of Finale or Sibelius lying around, you can't argue with the unlimited-trial-period shareware nature of Lime.
Lime is (approximately speaking) god's gift to the composer. It lacks some of the nifty input features of Finale, and it has less sophisticated playback. But, when it comes to laying out scores, and even composing, there's nothing that I'd rather use. It's wonderfully customizable (many 20th century notation conventions are much easier than in finale, you can use finale's alternate fonts, you can make it do damn near anything you want), and *it's free*.
Let me repeat that: ***it's free***
Finale sells for what, over $200, maybe $300 these days? Short of nifty input and playback tricks that most people never use, I'm yet to find anything that Finale or Sibelius can do that Lime can't. And it's OS X native. And stable.
Sorry if this is a bit over the top, but I really really love this program, and have wasted (in retrospect) many hours on finale. *maybe* if I was writing very large scale works I'd prefer finale, but I've written an entire musical in Lime without any problems.
did I *ever* say that I though LZW deserved patent protection? no. LZW, "1 click ______", etc, are exactly the sort of software that shouldn't be patentable. Just as some equally trivial physical invention isn't patentable. I think I made it perfectly clear that this was my opinion in my earlier posts.
your out-of-hand dismissal of the very idea of patents, just because of some horrendous examples, however, deeply troubles me. What the situation calls for is better training for patent examiners, coupled with a higher bar for what constitues a patentable invention. Not elimination of the patent system altogether. As I explained in my original post, elimination of the patent system would deal a major blow to research, and slow down (not speed up) the rate at which algorithms become available to the public.
The patent system is certainly broken. This is the fault of lawers who will seek a patent on anything and a lack of funding to the patent office, etc. But the situation can be remedied without throwing out the baby with the bathwater.
I assume that you mean to say that Free Software *couldn't* use my algorithm, if I made it free for non-commercial use only. You seem to have missed the fact that the GPL is far from the only licensing scheme available.
Please, don't confuse the Free Software Foundation with the open source community as a whole. There's a tremendous bulk of open source software released under exactly the licensing model I described, especially from academic researchers, who would be in exactly the position that I described.
Don't get me wrong. I'm a big fan of the FSF, and I whole-heartedly support it's goals. But the GPL is pretty clearly intended to cover exactly the sort of code that I assigned to the "isn't / shouldn't be patentable" category. Algorithms that truly represent legitamite scientific advances should be patentable, and clearly if patented, they wouldn't be released under the GPL (in all likelihood). However, they could very easily be released under the licensing scheme I describe, unassociated with the FSF.
Software patents might be incompatible with Free Software, but they are *not* inherently incompatible with open source.
if you really fully understand the implications of your suggestion, well, you have an interesting view of the way our economic system should work.
Of course algorithms should be patentable - suppose I come up with a great new video codec tomorrow, much better than anything else, but I can't patent it, I can only copyright the source code.
Now, I'm faced with a choice. I could develop closed-source software implementing the codec, and refrain from publishing my new algorithm, thus protecting my innovation with the copyright I'm allowed - if I do this, I'll probably make myself a tidy profit (assuming I know someone who knows the least bit about marketing), and be quite happy with the situation.
Or, I could publish the details of my algorithm, ensuring that no one will be willing to pay for my implementation (well, almost no one) - I can't get paid for my innovation, and the companies that already dominate the market will get my R&D efforts for free. They'll put their own implementation into their products, and pocket the proceeds.
Let's say I'm a noble minded researcher; I don't really care whether or not I get rich off my invention - on the other hand, I'm not stupid, I don't want anyone else getting rich off of it if I'm not getting some of the pie. Clearly, I'm going to keep my source closed, I'm not going to publish. This prevents my algorithm from being used in open-source products, blocks other researchers from extending and improving my results and generally holds back progress.
On the other hand, suppose that I can patent my great new codec. Now I have a third option. I can patent it, and set up a reasonable licensing scheme: you're free to implement my for private non-commercial use, research, etc. If you're getting paid from your implementation, then I want a cut of it, too. Now I can publish, its possible for the open-source and academic communities to use my great new invention, it's available for more research work, and at the same time, I can prevent other people from getting rich off of my work without also compensating me. This is *exactly* why patents exist: to allow people to profit from their work without impeding the flow of "progress".
Now, am I a loony git who thinks that *any* algorithm should be patentable? of course not. There's a reasonable standard, and it's illustrated perfectly by my previous example. An algorithm should be patentable only if the difficulty/effort in creating the algorithm sufficiently exceeds the difficulty/effort in implementing the algorithm.
Why this standard? If the "implementation cost" far exceeds the "invention cost", then no one's going to want to use their own implementation; they'll happily pay for mine, and a mere copyright on the source will suffice to protect me. Furthermore, the fact that the implementation cost far exceeds the invention cost is a strong indication that the algorithm in question fails to qualify as something that most people in the field wouldn't have thought up in the same situation (this should be a necessary standard for ALL patents - they should be INNOVATIVE).
On the other hand, if the "invention cost" exceeds the "implementation cost", then everyone else will develop their own implementations rather than use mine if the algorithm itself isn't protected; Conversely, since the invention cost/difficulty/effort/etc was so substantial, my invention is exactly the sort of thing that should be protected as innovative - something that the average person in the field wouldn't have thought of.
It should be noted that this requirement that the cost of invention far exceed the cost of implementation would actually eliminate the vast, vast bulk of software patents - which is a good thing. Patents such as "one-click ___", where you see it in operation without any knowledge of its guts and immediately know how it works, or the marching cubes algorithm (patented by HP, I believe), which is just what any sensible person with some background in computational geometry or algorithms would do, without much thought, should be gotten rid of. In all likelihood, this standard should be applied across the board, not just in software/algorithms. But should patents be done away with entirely, even in a restricted field? Of course not, and I think my example establishes that pretty clearly.
actually, you can get your license suspended - it's at the officer's discretion; if he thinks you were being truely reckless, he'll probably suspend your license. If, on the other hand, he thinks that you were merely stupid to have been caught, he'll just write you a ticket.
This is in CA - I know that there's a similar law in maine (20mph over = suspension), but the cops there are a lot happier to just take away your damn license, especially if you're from out of state (happened to a friend of mine).
Did you notice down at the bottom of the page where they compared the *actual* realized bit rate of each algorithm? 64kbit ogg turns out to be about 67-68kbit, once you factor in the header data, etc.
So, yes, the test took file size into consideration, and no, this doesn't make ogg look any better.
I'm a graduate student in math; while my preference in for pencil & paper, I *do* have my powerbook & wacom tablet set up for taking notes, for when I don't have pencil & paper, or I need to make an electronic copy anyway.
1/2 of the screen is an text window (handwriting recognition), the other half a window to draw the equations in. A scripted button to save the equations as pictures and drop them into the text file. Later on, I just go through and type the equations in by hand. Saves a whole lot of time over typing the whole thing from scratch.
Of course, it helps that I have very precise handwriting to start with, so I get near perfect performance from the handwriting recognition software. When my girlfriend tries to use the system, it doesn't go so well.
But... required computers in 7th grade? huh? why? hell, if I had my way, there wouldn't be calculators in high school. Maybe a computer science class, but most highschool CS classes are crap, and would probably be improved by getting rid of the computers (i.e. then they might actually learn some CS instead of writing "hello world" in pascal, then playing counterstrike all day - not that that isn't a valuable skill, per se)
or, rather, all the 3D display technologies I've seen so far, is that there's no "hidden line removal", so to speak. Every technology i've seen is inherently transparent, and uses some means to generate perceived light sources within a volume.
Unfortunately, the human optical system isn't really built to deal with this on a regular basis; we expect *most* things to be at least somewhat opaque, and have a considerably easier time processing visual information that adheres to those expectations. So what's really needed is a way to not only change the color of a voxel, but also it's opacity; basically an "alpha channel". (You can't just do old school hidden line (surface) removal because you don't know where the observer is).
Clearly, this is impossible with any of the spinning disc/helicoid techniques; with some of the other ideas (like crystal activated by non-visible-wavelength, etc) it seems like it should be possible; use one wavelength to produce light, another to turn pixels opaque. Make the interior of an "object" opaque, illuminate the boundary, and you've got a display that's much easier for the human visual system to process.
Prediction: until this happens, no real 3-D displays except for highly specific industry applications.
Why? Because it's bombproof. Completely stable, entirely predictable, and has never given me the slightest problem. I hate to imitate the stupid commercials, but "it just works". It's fast enough for everything except *serious* numerical simulations; things "ordinary" users will never ever throw at their computer. And for those, well, no desktop system is fast enough, so it's a moot point to compare.
Are there people with nice stable Windows or Linux installations? of course. But a much bigger percentage of mac users I've talked to have a problem-free life than on any other platform. For me, the productivity and piece of mind gains from having a completely stable system more than offset the productivity losses of having a 3 year old 400 mhz processor (which, let's face it are minimal - the vast bulk of computer time is spent waiting for user response, not system response).
** of course, if you're someone who just needs to run the newest games at the highest framerate, you might be an exception. me, i'll stick to using my ps/2 or gamecube for gaming and be perfectly happy.