Your comparison seems to suggest that the computer is much worse than the brain because to store enough info to recreate the brain would be prohibitive. However, if you look at it the other way around, and asked what how big a disk the brain could remember the contents of, you'd get a much less flattering answer. For example, a computer could easily remember the names and telephone numbers of 100 million people, while a human brain would probably have a lot of trouble with this, since you have to read the data out USING the brain, not by dissecting it and looking at the synapses.
Actually, it depends on your requirements. For a lot of nontrivial applications (matrix inversion and 3D Fourier transforms, for example) the entire dataset needs to be transmitted at each iteration.
I know that for what I do (pseudopotential plane wave calculations) one 100bT switch would be too slow to connect an 8 node cluster. As in, you'd be better off with all the memory in one computer, and forget parallelization.
Also, I think that typically you don't want more than two hops between nodes. Of course, it all depends what you're doing. If you're doing monte carlo stuff, you could probably get by with 9600 baud modems if it were cheaper than ethernet.
Actually, quite a few cluster environments run SMP for one very good reason: communication overhead. Communication is much faster in a shared memory environment.
NERSC, for example, has recently purchased an IBM SP system which has two processors (or was is 4?) per node, with plans to upgrade to 16 processors per node.
The problem with SMP and clusters is that the message passing software has to be smart emnough to take advantage of the shared memory situation, and needs to and this can also complicate things when you try to optimize your code.
No, the big cray machines are not, in fact, clusters. I have a supeconmputer expert friend who laments that there are no new supercomputers since Cray is dead. All the new "supercomputers" are just big clusters. The Cray T3E (my favorite supercomputer) runs a single operating system which controls all the nodes.
A cluster runs a separate operating system on each node. This generally (again, this is hearsay) makes it much harder to maintain a cluster than a supercomputer (meaing one with one operating system). We purchased a small 8 node IBM sp2 computer six months ago, and still haven't figured out how to make it act like a single computer.:( Oh well.
Provided you aren't dealing with a crashed hard drive or some other issue that can't be solved remotely.
Of course, on this machine the compute nodes aren't equipped with hard drives...:) Seriously, in a huge cluster like this, if a node fails, they will take it out, and may not even bother trying to fix it, I imagine.
I have found it very usable over the last year or so for day to day work
I'm just wondering, what is it that ALSA are very usable for? Just playing sounds, or for recording as well? And if for recording as well, for full duplex recording?
I'm asking because one of the two applications that keeps me attatched to the MacOS is ProTools, a very nice HDD recorder. I would defninitely like to be able to switch over to Linux entirely at some point. But it's really hard to record duets by oneself if you don't have full duplex recording!:)
I assume the other recorder they are referring to is Multitrack. (I don't have a link for it handy.) However, as far as I can tell, Multitrack is distributed binary only, and only for x86.
You are right, there is a center of the universe, or at least one could define one. For example, the center of mass of the universe. However, according to relativity there is no need for us to know what the center of the universe is or in which direction it is moving relative to us. We can just as easily treat ourselves as the center of the universe...
Don't even get me started on the cool-ass things that your barin does with the STEREO signal from your two ears. Bin-aural hearing is totally bad-ass.
Alas, my brain never receives a stereo signal.:( But it can do some pretty incredible things with the power spectrum of my mono signal, and I can often tell (albeit pretty crudely) where a sound is coming from. Higher frequencies have more trouble refracting around my head to reach my ear. Of course, this requires a sound with a broad and known (by me) power spectrum. But I do wish I could hear stereo.
There are in fact differences, and until people make and investigate biological neural nets, we won't know all the differences. One caveat is that the last time I really looked into neural nets was 4-5 years ago, so there may have been some improvements since then.
Computational neural nets have a floating point activation of each neuron, which is independent of time. In reality, the activation of a neuron is a very time dependent phenomenon. Traditionally the frequency of activation was considered the important thing, and this would correspond to the "activation" in the neural net models. Relatively recent research in the optical nerve pathways indicates that the timing of the individual eactivation potentials is also important. Specifically, simultaneous action potentials of several neurons carried a different meaning than the same frequency of uncorrelated action potentials.
I don't know what impact this would have on a neural net, but that is precisely the point. Biologists do not fully understand neural networks, and so there is no reason to believe that computational ones will learn as well as the real thing.
I'm not sure I can help it all make sense, but I may be able to clear up one small difficulty for you.
if both are accelerating at nearly equal speeds, would either one gain mass if there were no 'landmarks' to juge by?
You are referring to the increase in mass as a person approaches the speed of light. Actually, that is no longer the preferred way of thinking about it. This comes down to the question of definition of mass. Nowdays physicists define the mass to be the rest mass, and don't say that the mass changes. Unfortunately this has been a wee bit slow to catch on. The reason we define the mass this way is that the older definition was confusing because the mass was not a scalar (relativistically speaking). It also was not the same as the "Newtonian mass" as in the "m" in "F = ma", which makes it a rather deceptive quantity.
I guess these distinctions get a little confusing, but I guess my point would be that part of your problem might be that you have had relativity explained in a way that even Einstein agreed 50 years ago was not the best way.
I'm afraid you seem to have misunderstood the principle of relativity. There is no "REAL velocity". The whole point of relativity (at least, where it got its name from) was that it doesn't matter at all which observer you consider.
Of course, your speed will depend on the observer, just because it is defined relative to the observer. But the physics (i.e. what actually happens) is independent of the observer. In your example of a tree and an observer falling in the woods, the observer would certainly see that the tree is stationary, but the ground would seem to be travelling at a very disconcerting speed!
So this Belgian bloke has come up with a absolute wheeze: keep the surface area of the warp bubble you create really small but expand its volume to something you could reasonably fit a Volkswagen Polo into. Try this at home and you may run into the slight problem that volume usually increases to the 3/2 with area.
Actually, this really is not a problem, relative to the other stuff he's proposing. The only reason that volume typically increases with surface area is that we typically live in a (fairly) flat spacetime. The whole point of the warp drive is to theorize a nonflat spacetime. Actually creating such a spacetime is of course not so easy.:)
I'm afraid I don't know too much about general relativity, although I'm not too bad at special relativity, so I can hint at the answer to your question, but I can't really explain it.
Special relativity is valid only in a flat space-time. In practice, this means it is valid whenever gravity doesn't play a significant role. General relativity is necesary when space-time gets warped, as it does near massive objects. The Alcubierre Warp Drive works by having an incredible density of negative mass which warps spacetime in front and behind you in just the right manner to let you travel faster than the speed of light.
The Alcubierre Warp Drive creates a relativistically correct (my term) space time, which means you can go faster than the speed of light without violating any causality questions. So by construction (but using very hairy math) there is no causality paradox. However, it also warps space time all around the traveller, which could be a bit disruptive.
Also, there is no theory of how one could turn the warp drive on or off. It seems quite possible that one would age significantly either when turning on or off the warp drive. Or maybe just annihilate the populations of any nearby planets.:)
If current flows in the direction of the thumb of the right hand, the magnetic field lines wrap around the wire in the direction of the fingers.
Correct.
So given two wires there is an attractive force between them if the currents are co-directional, and a repulsive one if the currents are opposite.
Correct.
It's been a long time, so I've forgotten what goes on with regular AC, but as I remember - this principle, with extremely high currents, extremely close proximity of wires, and alternation in currents at a quarter-phase, would generate unidirectional forces == motion of the whole assembly.
I'm afraid this wouldn't work, since momentum is conserved under electromagnetism. Of course, if your apparatus were to radiate in one direction, it would get a bit of a push in the other, but that is just because the photons would be carrying away momentum.
One last thing... Most of the benchmarks I've seen have compared different clock speed chips on the same graph! What's the deal with this? I'd a least prefer it if the benchmarkers divided the score by the clock speed, for instance, to give a more meaningful measure of the efficiency of the chip itself!
Or perhaps they should instead divide it by the price of the processor?:) But more seriously, while it might be convenient to be able to think K7 as being equal to a fixed fraction times a PII of the same clock speed, there are so many other effects coming into play in the benchmarks that I'm not sure this would really be useful.
Nope, I don't think this will happen, and if it does, I am quite sure there will be protests when people discover that you do not only see weapons of different kinds but also the shape of people's bodies. Just look at the picture up to the right and say that doesn't look like a women...
I think your getting just a tad paranoid here. It doesn't take an x-ray machine to record the shapes of people's bodies! There are security cameras all over the place that do that just fine. If you don't want anyone to know what you look like, you shouldn't leave your home.:)
As far as people with assorted metal in their bodies, they already set off metal detectors and are inconvenienced. Possibly with the new scanner they wouldn't have to stand aside and be searched each time they go to the airport, which would be much nicer, provided the x-ray dosage is low enough.
To use this as something approaching a supercomputer you have to custom write your applications using a message passing library such as MPICH.
This is sort of a moot point. To use just about any modern supercomputer you need to write your code using mpi, so this isn't really a downside to the Beowulf clusters. Trying to get around using mpi would be foolish, as it would result in nonportable code. The cool thing about Beowulf clusters (as I see it) is that you can pretty much take any mpi code you have and just recompile.
Of course, you would also need compilers and the relevant libraries, which might be a problem. The code I use (a plane wave pseudopotential DFT code, if that means anything to you) requires certain libraries (plus a f90 compiler) which I don't think are available for intel.:( But then again, it requires a lot of internode communication, so it would be really slow on a Beowulf cluster anyways.
Right now, linux is only really useful to hackers/programmers/geeks.
There are basically three kinds of users: novices, competent users and geeks. Here I'm just going to use the word 'geek' as a shorthand for 'hacker/programmer/geek/sysadmin.' It's shorter.
I fall under the category 'competent.' I use UNIX at work, write scripts to do mundane chores, program from time to time, and run jobs on supercomputers. I have not, however, ever sysadmined, and I hope to avoid ever doing so. I would like to be to easily install and configure Linux.
My mom falls under the category 'novice' as far as UNIX systems go. Not that she is new to computers, she has been using them as long as I have. But she doesn't go out and learn how to do anything unless she has a need to do it. She most certainly is not stupid. She has a PhD in Physics. She should be able to gain a lot from using Linux.
She hasn't upgraded her computer (or software, mostly) for eight or nine years, because she wants to keep it stable. She sees that my dad (who upgrades impulsively) has a computer that crashes all the time, while hers (since she just stuck with a version that happened to be stable) still works. I'm sure she would love to use a stable modern powerful operating system, if she could easily figure out how to do what she needs to do, which is just involves a spreadsheet, a word processor and a draw program. And they would have to be easy enough for her to figure out how to use them without reading the manual (since I've never seen her read a manual).
I think that it is possible to make software easy enough for us to use without dumbing it down. Of course, in my case I know I could learn to sysadmin if I wanted. I just would rather not. And I am sure possible to create a version of the OS that is easy to configure to run as a personal computer.
I kind of like my current weight. If I'm going to visit another planet, I'd rather not have to change my weight by more than a factor of two or three!:)
I'm glad they decided to include a good FPU. The 604 had a nice one, but when the "optimized for the MacOS" the G3 they decided that they would be better off without a fast FPU. Very sad, for a physicist like me. But at last they seem to have turned around and made a chip that is more like the old 604's.
Slashdot Mirror
Your comparison seems to suggest that the computer is much worse than the brain because to store enough info to recreate the brain would be prohibitive. However, if you look at it the other way around, and asked what how big a disk the brain could remember the contents of, you'd get a much less flattering answer. For example, a computer could easily remember the names and telephone numbers of 100 million people, while a human brain would probably have a lot of trouble with this, since you have to read the data out USING the brain, not by dissecting it and looking at the synapses.
I know that for what I do (pseudopotential plane wave calculations) one 100bT switch would be too slow to connect an 8 node cluster. As in, you'd be better off with all the memory in one computer, and forget parallelization.
Also, I think that typically you don't want more than two hops between nodes. Of course, it all depends what you're doing. If you're doing monte carlo stuff, you could probably get by with 9600 baud modems if it were cheaper than ethernet.
NERSC, for example, has recently purchased an IBM SP system which has two processors (or was is 4?) per node, with plans to upgrade to 16 processors per node.
The problem with SMP and clusters is that the message passing software has to be smart emnough to take advantage of the shared memory situation, and needs to and this can also complicate things when you try to optimize your code.
A cluster runs a separate operating system on each node. This generally (again, this is hearsay) makes it much harder to maintain a cluster than a supercomputer (meaing one with one operating system). We purchased a small 8 node IBM sp2 computer six months ago, and still haven't figured out how to make it act like a single computer. :( Oh well.
Of course, on this machine the compute nodes aren't equipped with hard drives... :) Seriously, in a huge cluster like this, if a node fails, they will take it out, and may not even bother trying to fix it, I imagine.
I'm just wondering, what is it that ALSA are very usable for? Just playing sounds, or for recording as well? And if for recording as well, for full duplex recording?
I'm asking because one of the two applications that keeps me attatched to the MacOS is ProTools, a very nice HDD recorder. I would defninitely like to be able to switch over to Linux entirely at some point. But it's really hard to record duets by oneself if you don't have full duplex recording! :)
I assume the other recorder they are referring to is Multitrack. (I don't have a link for it handy.) However, as far as I can tell, Multitrack is distributed binary only, and only for x86.
You are right, there is a center of the universe, or at least one could define one. For example, the center of mass of the universe. However, according to relativity there is no need for us to know what the center of the universe is or in which direction it is moving relative to us. We can just as easily treat ourselves as the center of the universe...
Alas, my brain never receives a stereo signal. :( But it can do some pretty incredible things with the power spectrum of my mono signal, and I can often tell (albeit pretty crudely) where a sound is coming from. Higher frequencies have more trouble refracting around my head to reach my ear. Of course, this requires a sound with a broad and known (by me) power spectrum. But I do wish I could hear stereo.
Computational neural nets have a floating point activation of each neuron, which is independent of time. In reality, the activation of a neuron is a very time dependent phenomenon. Traditionally the frequency of activation was considered the important thing, and this would correspond to the "activation" in the neural net models. Relatively recent research in the optical nerve pathways indicates that the timing of the individual eactivation potentials is also important. Specifically, simultaneous action potentials of several neurons carried a different meaning than the same frequency of uncorrelated action potentials.
I don't know what impact this would have on a neural net, but that is precisely the point. Biologists do not fully understand neural networks, and so there is no reason to believe that computational ones will learn as well as the real thing.
if both are accelerating at nearly equal speeds, would either one gain mass if there were no 'landmarks' to juge by?
You are referring to the increase in mass as a person approaches the speed of light. Actually, that is no longer the preferred way of thinking about it. This comes down to the question of definition of mass. Nowdays physicists define the mass to be the rest mass, and don't say that the mass changes. Unfortunately this has been a wee bit slow to catch on. The reason we define the mass this way is that the older definition was confusing because the mass was not a scalar (relativistically speaking). It also was not the same as the "Newtonian mass" as in the "m" in "F = ma", which makes it a rather deceptive quantity.
I guess these distinctions get a little confusing, but I guess my point would be that part of your problem might be that you have had relativity explained in a way that even Einstein agreed 50 years ago was not the best way.
Of course, your speed will depend on the observer, just because it is defined relative to the observer. But the physics (i.e. what actually happens) is independent of the observer. In your example of a tree and an observer falling in the woods, the observer would certainly see that the tree is stationary, but the ground would seem to be travelling at a very disconcerting speed!
Actually, this really is not a problem, relative to the other stuff he's proposing. The only reason that volume typically increases with surface area is that we typically live in a (fairly) flat spacetime. The whole point of the warp drive is to theorize a nonflat spacetime. Actually creating such a spacetime is of course not so easy. :)
Special relativity is valid only in a flat space-time. In practice, this means it is valid whenever gravity doesn't play a significant role. General relativity is necesary when space-time gets warped, as it does near massive objects. The Alcubierre Warp Drive works by having an incredible density of negative mass which warps spacetime in front and behind you in just the right manner to let you travel faster than the speed of light.
The Alcubierre Warp Drive creates a relativistically correct (my term) space time, which means you can go faster than the speed of light without violating any causality questions. So by construction (but using very hairy math) there is no causality paradox. However, it also warps space time all around the traveller, which could be a bit disruptive.
Also, there is no theory of how one could turn the warp drive on or off. It seems quite possible that one would age significantly either when turning on or off the warp drive. Or maybe just annihilate the populations of any nearby planets. :)
If current flows in the direction of the thumb of the right hand, the magnetic field lines wrap around the wire in the direction of the fingers.
Correct.
So given two wires there is an attractive force between them if the currents are co-directional, and a repulsive one if the currents are opposite.
Correct.
It's been a long time, so I've forgotten what goes on with regular AC, but as I remember - this principle, with extremely high currents, extremely close proximity of wires, and alternation in currents at a quarter-phase, would generate unidirectional forces == motion of the whole assembly.
I'm afraid this wouldn't work, since momentum is conserved under electromagnetism. Of course, if your apparatus were to radiate in one direction, it would get a bit of a push in the other, but that is just because the photons would be carrying away momentum.
One last thing... Most of the benchmarks I've seen have compared different clock speed chips on the same graph! What's the deal with this? I'd a least prefer it if the benchmarkers divided the score by the clock speed, for instance, to give a more meaningful measure of the efficiency of the chip itself!
Or perhaps they should instead divide it by the price of the processor? :) But more seriously, while it might be convenient to be able to think K7 as being equal to a fixed fraction times a PII of the same clock speed, there are so many other effects coming into play in the benchmarks that I'm not sure this would really be useful.
different kinds but also the shape of people's bodies. Just look at the picture up to the right and say that doesn't look like a women...
I think your getting just a tad paranoid here. It doesn't take an x-ray machine to record the shapes of people's bodies! There are security cameras all over the place that do that just fine. If you don't want anyone to know what you look like, you shouldn't leave your home. :)
As far as people with assorted metal in their bodies, they already set off metal detectors and are inconvenienced. Possibly with the new scanner they wouldn't have to stand aside and be searched each time they go to the airport, which would be much nicer, provided the x-ray dosage is low enough.
This is sort of a moot point. To use just about any modern supercomputer you need to write your code using mpi, so this isn't really a downside to the Beowulf clusters. Trying to get around using mpi would be foolish, as it would result in nonportable code. The cool thing about Beowulf clusters (as I see it) is that you can pretty much take any mpi code you have and just recompile.
Of course, you would also need compilers and the relevant libraries, which might be a problem. The code I use (a plane wave pseudopotential DFT code, if that means anything to you) requires certain libraries (plus a f90 compiler) which I don't think are available for intel. :( But then again, it requires a lot of internode communication, so it would be really slow on a Beowulf cluster anyways.
There are basically three kinds of users: novices, competent users and geeks. Here I'm just going to use the word 'geek' as a shorthand for 'hacker/programmer/geek/sysadmin.' It's shorter.
I fall under the category 'competent.' I use UNIX at work, write scripts to do mundane chores, program from time to time, and run jobs on supercomputers. I have not, however, ever sysadmined, and I hope to avoid ever doing so. I would like to be to easily install and configure Linux.
My mom falls under the category 'novice' as far as UNIX systems go. Not that she is new to computers, she has been using them as long as I have. But she doesn't go out and learn how to do anything unless she has a need to do it. She most certainly is not stupid. She has a PhD in Physics. She should be able to gain a lot from using Linux.
She hasn't upgraded her computer (or software, mostly) for eight or nine years, because she wants to keep it stable. She sees that my dad (who upgrades impulsively) has a computer that crashes all the time, while hers (since she just stuck with a version that happened to be stable) still works. I'm sure she would love to use a stable modern powerful operating system, if she could easily figure out how to do what she needs to do, which is just involves a spreadsheet, a word processor and a draw program. And they would have to be easy enough for her to figure out how to use them without reading the manual (since I've never seen her read a manual).
I think that it is possible to make software easy enough for us to use without dumbing it down. Of course, in my case I know I could learn to sysadmin if I wanted. I just would rather not. And I am sure possible to create a version of the OS that is easy to configure to run as a personal computer.
enough to have an atmosphere...
I think that all these planets are too close to the sun, so their moons wouldn't be able to retain an atmosphere if they are earthsized. :( Alas.
I kind of like my current weight. If I'm going to visit another planet, I'd rather not have to change my weight by more than a factor of two or three! :)
I'm glad they decided to include a good FPU. The 604 had a nice one, but when the "optimized for the MacOS" the G3 they decided that they would be better off without a fast FPU. Very sad, for a physicist like me. But at last they seem to have turned around and made a chip that is more like the old 604's.