MPEG stands for Motion Pictures Expert Group, and is a standard setting commitee.
The actual algorithms used in MPEG-1,2,4 etc are called block-based DCT, Arithmetic coding, MAX-quantization, VOP segmentation etc... Of these, I think at least block-based DCT is fit to be on the list, mainly beacuse its widespread useage (JPEG, MPEG-1, MJPEG, MPEG-2, H.263 etc.
Personally I don't see why they even bother to teach BS. Insertion sort, and selection sort are just as easy to comprehend, and much faster. Lots of people in my class had actually reinvented these before they were taught algorithms in class...
That said, I was also shown a video of sorting algorithms in CS class. The only thing that made the video worth watching was that the algorithms made all sorts of weird noises. We actually borrowed the tape from our teacher some years later, and had a good laugh watching it B) IIRC the soundtrack was made by Wendy Carlos.
Some special variant of wavelets might use FFT, but in general your statement is not correct. FFT (or actually DFT) is a special case of wavelets, and thus it is not covered. Most wavelet bases aim to be compact both in spatio-temporal domains, and in the frequency domain, and can thus be said to land somewhere in between.
The second reason that occurs to me is that having the source available would make reverse engineering an application relatively trivial.
Actually, if you have the source, it's not reverse engineering anymore. Reverse engineering means finding out how something works, ie. going from implementation to specs (or in this case, code), as opposed to engineering which is designing specs, and implementing them.
Is the code-morphing software open sourced? Unlikely, since they have one of the most closed systems on the planet.
Is the code-morphing patented? If it is I see only advantages for them if they make it open source. After all, they sell the entire kit (CPU+code-morpher) don't they? If they make the code-morpher open source they would still sell as many CPUs.
That the main advantage with open source is that you get lots of people to track bugs, and customize the software. Many O-S projects (XMMS, GIMP etc.) have evolved as closed source initially. The source is only released once the product is working and stable. So if Transmeta actually wanted to open-source the code-morpher, it might just be natural to wait until there are machines with Crusoe chips on the market.
By releasing the source code, what's to stop a competitor of from using it in their closed source-project? It will be rather impossible to prove it.
If you know which compiler was used, you could always compare the binaries. Something similar actually appeared in court a few years ago, and it worked. IIRC is was stolen closed-source code that time, but in principle its the same.
So, goodbye Marvin Minsky! So long, later John McCarthy! We'll see you in the Open Source AI!
This is a very good idea that should be adopted more widely I think. (Things are moving in this direction already though. RoboCup has helped a lot to get people started.) In today's AI, there are so many people working on inventing new programming structures or ways of arranging "knowledge", without ever bothering to think about how they should obtain this knowledge. If they had the programming structure given, they could concentrate on trying things out on real problems. The AI community suffers from way too much theory way too little practice IMHO.
One major difference between me as the subject in an IQ test and a robot is that the robot has a clear record of its algorithms in memory and I do not...
And this is supposed to be an advantage to the robot? I'd say the contrary, you should be glad you have the precise details of how to move your hands, how you sense smells etc hidden from consciousness.
Consider this analogy: Knowing the rules of chess does not make you a good chess player. Or this one: Birds can fly, but they cannot tell you how they do it:-)
According to Dr Poley, you should just 'ask it'... In fact, the AI 'being' may have a better understanding of what makes itself tick than you do.
You can always "ask" yourself how you perform things as well, but as most aspects of our intelligence are non-linguistic, you cannot expect to get an answer in linguistic terms, you will have to settle for a non-linguistic answer. That is, a set of examples.
It sounds as if Dr Poley, just as so many others is a victim of the consciousness fallacy. There is more to our intelligence than consciousness, but as we in general don't need to know about these parts, we tend to forget that they exist at all.
Hmm if I compiled a C program for Linux and mailed it to a Linux user, couldn't the same thing happen if he/she ran it? Where's the patch??????
You can't just double click the attachment to run it, or can you? At least I can't do that. If you have to save it, then run it you will have some more time to think, and you will definitely know that you are executing a program by then. As someone pointed out here earlier, the executable extension (.exe) is hidden by default in Wxx, so you might not even know that you are executing a binary if the file is called niftypicture.gif.exe or similar.
The transmeta cpu actually executes x86 instructions...
Not at all. The Transmeta CPU:s execute their own set of instructions. The x86 code is interpreted by a small piece of software, that can be replaced. Thus no more CPU bugs, like in Pentium and K6. Simply upload a new code parser.
Although another instruction set could be developed, this would provide little improvement over emulation of x86. (at least according to initial tests as reported by Linus.)
This can't be the Crusoe chips you are talking about! Transmeta change the instruction set between each model of the Crusoe chips, the original posting was actually more correct than your "correction".
I think so too. My guess is that the Hermann's grid effect is enhanced by the presence of edges surrounding the spots that appear darker. It is well known that edges affect how we perceive shades of grey, see for instance this one for an example of what can happen if the edges aren't there.
I have to check out that article in Vision Research. Anyway, thanks for the links.
"...the devices are modelled exactly on the way the human eye works, based on the work of thousands of previous researchers into how all human eyes function. 'It was remarkably easy to implement,' say the inventors, 'once others laid the groundwork.'
It's always a good idea to stop reading when you see things that are not there. This it what the article really said:
...models the human perceptual process at the hardware level by mimicking the separate temporal and spatial functions of the eye-to-brain system.
AFAIK models by mimicking and models exactly is not the same. To write your interpretation as a citation is really bad manners BTW.
Okay, so what they're claiming is that their brand-new, $6 ($50 in total) device can do things which long-standing scientific projects costing $10,000 cannot? Am I the only one who thinks that this sounds somewhat fishy?
Apparently they have concentrated on a subset of vision problems. The device detects moving objects during varying lightning conditions, which is quite a feat. But, there is more to a generic vision system than that. For instance, you often want to know what kind of object you are tracking. Is it a ball or a bird?
The problem is, it doesn't look like there's that much room left in either the panel or the device on his head.
His implant was bulky because it was old-fashioned. If they were to do the implant now it would be much smaller. The reason the implant was old-fashioned is that they have been testing it for a long time (something like ten years).
Seems the price has gone down "a bit" since '97 though:
The modules measure 40 mm2, have 100 pins, and can handle 20-MHz video signals. The chip is priced at $960. On a card with a socketed GVPP and 64 kilobytes of Flash RAM, the price comes to $1,500.
I doubt that. A lot of things happen in the brain during sleep. Sure, the activity is slightly lower, but if the sole purpose of sleeping was to recharge, then why does not all activity stop?
If the brain does use up more energy than it receives during wake hours, I find it more likely that this is just the brain utilizing the wake hours more since it "knows" (in an evolutionary sense of course) that it can regain the lost energy during sleep.
IMHO, a more realistic theory about sleep is the one that states that sleep, especcially dream periods are used to sort the experiences of the say and store them. The weirdness of dreams would be caused by events being replayed faster than real-time by the subconscious, at a speed that consciousness cannot keep up with.
And about cells dying and others taking place - that's also not true. Neurons have several axons forming synapses, and if one dies, the axons go with it - if one were to take it's place, it would have lost all the connections of the previous one. And since learning, memory, our personality, etc, is all due to the way the various axons are arranged, after some time we'd have these parameters altered: different personality, loss of memory, etc, and that does NOT happen.
Actually brain cells do regenerate. There was a lot of talk about this in journals this autumn. Do a search on neurogenesis to find out more. Brain cells have been found to regenerate in the hippocampus (an area associated with some aspects of memory) in human brains. Previously neurogenesis has also been found to occur each season in some song birds, once the need to learn a new song arises.
Slashdot Mirror
Sorting Out Sorting
Yeah, that's the one!
MPEG stands for Motion Pictures Expert Group, and is a standard setting commitee.
The actual algorithms used in MPEG-1,2,4 etc are called block-based DCT, Arithmetic coding, MAX-quantization, VOP segmentation etc... Of these, I think at least block-based DCT is fit to be on the list, mainly beacuse its widespread useage (JPEG, MPEG-1, MJPEG, MPEG-2, H.263 etc.
Those sound more like tools than algorithms to me, but I admit that it's difficult to decide where to draw the line.
But in the same category as those you listed, I'd also like to put B-splines and least squares...
Personally I don't see why they even bother to teach BS. Insertion sort, and selection sort are just as easy to comprehend, and much faster. Lots of people in my class had actually reinvented these before they were taught algorithms in class...
That said, I was also shown a video of sorting algorithms in CS class. The only thing that made the video worth watching was that the algorithms made all sorts of weird noises. We actually borrowed the tape from our teacher some years later, and had a good laugh watching it B) IIRC the soundtrack was made by Wendy Carlos.
wavelet uses FFT, so it's covered.
Some special variant of wavelets might use FFT, but in general your statement is not correct. FFT (or actually DFT) is a special case of wavelets, and thus it is not covered. Most wavelet bases aim to be compact both in spatio-temporal domains, and in the frequency domain, and can thus be said to land somewhere in between.
Want your own LISP machine? Code it up!
Yeah, or you could even make it parse Java bytecode. The code-morphing idea is pretty similar to JIT (Just in Time compilation) anyway.
But to be able to write your own code morpher you only need the specs for the Crusoe chip and the rest of the harware, not the code-morphing software.
The second reason that occurs to me is that having the source available would make reverse engineering an application relatively trivial.
Actually, if you have the source, it's not reverse engineering anymore. Reverse engineering means finding out how something works, ie. going from implementation to specs (or in this case, code), as opposed to engineering which is designing specs, and implementing them.
Is the code-morphing software open sourced? Unlikely, since they have one of the most closed systems on the planet.
Is the code-morphing patented?
If it is I see only advantages for them if they make it open source. After all, they sell the entire kit (CPU+code-morpher) don't they? If they make the code-morpher open source they would still sell as many CPUs.
That the main advantage with open source is that you get lots of people to track bugs, and customize the software. Many O-S projects (XMMS, GIMP etc.) have evolved as closed source initially. The source is only released once the product is working and stable. So if Transmeta actually wanted to open-source the code-morpher, it might just be natural to wait until there are machines with Crusoe chips on the market.
By releasing the source code, what's to stop a competitor of from using it in their closed source-project?
It will be rather impossible to prove it.
If you know which compiler was used, you could always compare the binaries. Something similar actually appeared in court a few years ago, and it worked. IIRC is was stolen closed-source code that time, but in principle its the same.
You can't run from satellites...
He, he... Sounds as if you've been watching Enemy of the state :-)
So, goodbye Marvin Minsky! So long, later John McCarthy! We'll see you in the Open Source AI!
This is a very good idea that should be adopted more widely I think. (Things are moving in this direction already though. RoboCup has helped a lot to get people started.) In today's AI, there are so many people working on inventing new programming structures or ways of arranging "knowledge", without ever bothering to think about how they should obtain this knowledge. If they had the programming structure given, they could concentrate on trying things out on real problems. The AI community suffers from way too much theory way too little practice IMHO.
I don't think so either. Written 20 years ago, and alas still way ahead of him...
One major difference between me as the subject in an IQ test and a robot is that the robot has a clear record of its algorithms in memory and I do not...
And this is supposed to be an advantage to the robot? I'd say the contrary, you should be glad you have the precise details of how to move your hands, how you sense smells etc hidden from consciousness.
Consider this analogy: Knowing the rules of chess does not make you a good chess player. Or this one: Birds can fly, but they cannot tell you how they do it :-)
According to Dr Poley, you should just 'ask it'... In fact, the AI 'being' may have a better understanding of what makes itself tick than you do.
You can always "ask" yourself how you perform things as well, but as most aspects of our intelligence are non-linguistic, you cannot expect to get an answer in linguistic terms, you will have to settle for a non-linguistic answer. That is, a set of examples.
It sounds as if Dr Poley, just as so many others is a victim of the consciousness fallacy. There is more to our intelligence than consciousness, but as we in general don't need to know about these parts, we tend to forget that they exist at all.
Hmm if I compiled a C program for Linux and mailed it to a Linux user, couldn't the same thing happen if he/she ran it? Where's the patch??????
You can't just double click the attachment to run it, or can you? At least I can't do that. If you have to save it, then run it you will have some more time to think, and you will definitely know that you are executing a program by then. As someone pointed out here earlier, the executable extension (.exe) is hidden by default in Wxx, so you might not even know that you are executing a binary if the file is called niftypicture.gif.exe or similar.
The transmeta cpu actually executes x86 instructions...
Not at all. The Transmeta CPU:s execute their own set of instructions. The x86 code is interpreted by a small piece of software, that can be replaced. Thus no more CPU bugs, like in Pentium and K6. Simply upload a new code parser.
Although another instruction set could be developed, this would provide little improvement over emulation of x86. (at least according to initial tests as reported by Linus.)
This can't be the Crusoe chips you are talking about! Transmeta change the instruction set between each model of the Crusoe chips, the original posting was actually more correct than your "correction".
Now that's what I'd call a good flamebait ;)
Ought to be marked as funny imho. :)
It's similar to the Hermann's grid illusion...
I think so too. My guess is that the Hermann's grid effect is enhanced by the presence of edges surrounding the spots that appear darker. It is well known that edges affect how we perceive shades of grey, see for instance this one for an example of what can happen if the edges aren't there.
I have to check out that article in Vision Research. Anyway, thanks for the links.
Sigh. I stopped reading after I saw
"...the devices are modelled exactly on the way the human eye works, based on the work of thousands of previous researchers into how all human eyes function. 'It was remarkably easy to implement,' say the inventors, 'once others laid the groundwork.'
It's always a good idea to stop reading when you see things that are not there. This it what the article really said:
AFAIK models by mimicking and models exactly is not the same. To write your interpretation as a citation is really bad manners BTW.
Okay, so what they're claiming is that their brand-new, $6 ($50 in total) device can do things which long-standing scientific projects costing $10,000 cannot? Am I the only one who thinks that this sounds somewhat fishy?
Apparently they have concentrated on a subset of vision problems. The device detects moving objects during varying lightning conditions, which is quite a feat. But, there is more to a generic vision system than that. For instance, you often want to know what kind of object you are tracking. Is it a ball or a bird?
The problem is, it doesn't look like there's that much room left in either the panel or the device on his head.
His implant was bulky because it was old-fashioned. If they were to do the implant now it would be much smaller. The reason the implant was old-fashioned is that they have been testing it for a long time (something like ten years).
Just did a quick search, and found out that GVPP isn't exactly new (the article mentions that it was invented in 1992):
http://www.techweb.com/wire/news/1997/09/0913visio n.html
Seems the price has gone down "a bit" since '97 though:
The modules measure 40 mm2, have 100 pins, and can handle 20-MHz video signals. The chip is priced at $960. On a card with a socketed GVPP and 64 kilobytes of Flash RAM, the price comes to $1,500.
$6 sounds much better to me :)
It is. From the article:
Microphones embedded in the cat's head enable her to recognise her own name and react by turning her head and blinking coyly.
The traces on "modern" chips can't take the raditation. Errors all over the place...
Then, of course, they should be shielded. This is common practise in military electronics to withstand nuclear radiation...
Energy. That's why we sleep.
I doubt that. A lot of things happen in the brain during sleep. Sure, the activity is slightly lower, but if the sole purpose of sleeping was to recharge, then why does not all activity stop?
If the brain does use up more energy than it receives during wake hours, I find it more likely that this is just the brain utilizing the wake hours more since it "knows" (in an evolutionary sense of course) that it can regain the lost energy during sleep.
IMHO, a more realistic theory about sleep is the one that states that sleep, especcially dream periods are used to sort the experiences of the say and store them. The weirdness of dreams would be caused by events being replayed faster than real-time by the subconscious, at a speed that consciousness cannot keep up with.
And about cells dying and others taking place - that's also not true. Neurons have several axons forming synapses, and if one dies, the axons go with it - if one were to take it's place, it would have lost all the connections of the previous one. And since learning, memory, our personality, etc, is all due to the way the various axons are arranged, after some time we'd have these parameters altered: different personality, loss of memory, etc, and that does NOT happen.
Actually brain cells do regenerate. There was a lot of talk about this in journals this autumn. Do a search on neurogenesis to find out more. Brain cells have been found to regenerate in the hippocampus (an area associated with some aspects of memory) in human brains. Previously neurogenesis has also been found to occur each season in some song birds, once the need to learn a new song arises.