Actually the difference is very clear. Imperative languages modify state while they are running. Functional languages create values. Just because some functional feature have been added to imperative languages, and vica versa does not weaken this distinction. This is why imperative monads are referred to as "pollution" in functional programs and it is common practice (enforced by the type system) to restrict their use to the "outside" parts of a program, ensuring that the core is purely functional.
Logic programming languages have no functions, or even a sense of time or order.
This is not true in practice. Logical relations are a superset of functions, and so most real logic code is written in predicates that are somewhat similar in both form and function to functional languages. The most common logic language Prolog actually has operational semantics similar to a procedural language. It is very easy to write purely procedural code within Prolog, or to use the extra features that a logic language provides.
Prolog does have a well-defined evaluation ordering for reasons of efficiency, and also because it is quite hard to express the semantics of a language which weakens this ordering. Mercury is one language that moves in that direction, also replacing the builtin arithmetic with a CLP library allows more flexibility in ordering of evaluations.
The meta-programming that you describe is not a specific feature of functional languages. Reflection and Reiterfication can be added to any type of language. Some functional languages (such as lisp) have it as a built-in feature, while some (Haskell if I remember correctly) does not. Amongst the logic languages Prolog has very powerful meta-programming features (there is no separation between code and data and both are manipulated in the same way), and amongst the procedural languages I think that Smalltalk has strong meta-programming facilities althought I haven't used it.
Err... have you got some weird exotic hardware in mind?
In x86 protected mode memory is a single contiguous address range. Sure it is split into different pages but that is not something that you have to deal with as a programmer. Hardware does maintain a stack for you on most architectures. On x86 it's done with a dedicated register and standard addressing modes to build the frame.
It's been a while since I've programmed in Z80, or in 68k, but most other CPU architectures also have contiguous addressing and hardware stacks. The only chips that I've used that don't are PIC microcontrollers.
No this is completely wrong. Non-proliferation is completely unilateral and is aimed at preventing all non-nuclear states from developing the bomb. Have a read if you are unsure of the terms - but don't make up half-baked analogies to support your incorrect assumptions.
So in your terms, the signatories to the NPT who possess nukes are saying "Nyeh, we want to be the only ones with the bomb". Which is why the poster that I replied to was making such a contorted point, why the US is hypocritical in its policy, and why you are flat out wrong.
That's a contorted argument designed to win a cheap point in an argument. You know exactly what the GP meant: The point of proliferation is that it leads to possession. A country possessing nukes cannot argue against proliferation without being a hypocrite - it is specifically arguing that other countries should not be able to do what it has.
Whoops my bad. It's not in the summary at all, it's in one of the articles. No one could be expected to read that far. Carry on, nothing to see here...
Or maybe use the hop count so that you're measuring network distance rather than geographical distance. You know, the way they describe IN THE SUMMARY.
Erm, it's not actually a mistake. It depends on whether you parse "increase" as being additive, or multiplicative. In everyday english it can be either. So what you're describing is ambiguity rather than error.
Yeah I know. I submitted a followup but I don't think it's going to get voted up enough within firehose. Shame as I think this is a really important topic - how the first major transition of a broadcast network to internet programing gets lumbered wit DRM.
I was quite suprised as I read the story about your father in law because I'd implicitly assumed that you were American, from your libertarian sounding views. Then I read as far as you being Canadian and oddly that story wasn't a suprise. Your country has possibly one of the best international reputations. I think it's really good for them to do that for him, very generous and considerate.
It's an interesting question when you compare no-national-army against no army at all, rather than private armies. At one point I would have thought you naive and idealist for suggesting it, but after the past five years of watching Bush & Blair's wars it doesn't seem like such a bad idea.
Why has the AC been modded down? He explained (with direct quotes) what the Loebner Prize that the researchers are going for is.
Reading the description it sounds like it could be done with current methods and brute-force. In particular restricting the scope of the questions to avoid tripping up the AI removes a large chunk of the problem. It is only a small first step towards the real Turing Test.
For those who want to judge progress themselves, a much better link than the article is the researcher's page. It has videos of their system in action (or it did when I typed this, I suspect it may die shortly...)
What they've done is no more impressive than many current chatbots. The two main differences (that created this new story) are that they are doing it with avators (yay! newmedia-whores2.0), and that they're going to try it out on much bigger hardware than has been tried previously. But even the false beliefs demo is not that impressive if you strip off the Second Life frontend. It is only as complex as a typical demo in blocks world.
Yes, that was my point. Given that there doesn't seem to be prior art they will have to argue obviousness. But showing that something was obvious in hindsight is much harder than showing prior art.
It will be hard to find. The patent was filed on Oct 5th 1999. I can't find exactly when Napster started, but it was first sued in Dec 1999 - so shortly before that. The problem is that Napster is not prior art for this patent. The service described is basically iTunes. It uses a central server with a library of music, and client devices use licenses to download the tracks that they can prove they are entitled to.
I suspect that this patent will be very hard to fight in the US courts because the defense relies on how obvious the invention is. There may be prior art for music distribution, and for downloading licensed content from a central server, but the issue is whether or not there is prior art for the combination. I can't think of any iTunes-like servers back before Napster. Although there were plenty of illegal ftp/web sites they didn't use licenses for obvious reasons.
Looks like these guys saw Napster, guessed which way the wind was going to blow and rushed to file a patent before the music industry organised themselves. I bet that they never implemented / released a product before iTunes. So Apple may have some scope in that it is an obvious combination of well-known pieces, and that ZapMedia didn't even have to invent the product (at least build a prototype) to file the patent.
I'd agree with what you say, but I would go further. They are short-sighted idiots. Who do they think differential bandwidth pricing is aimed at? Seeing them weigh in on the side of the ISPs is like watching someone hand a knife to a mugger in a dark alley. It's almost as if they haven't realised that in a few years they're going to want to provide high-bandwidth services to customers over those same connections...
Why would it be overselling when they can't meet your arbitrary contention ratio of 10%? Why not 1%? Why not 0.1%? If you're going to stick an arbitrary constant in an equation then you can't use it to back up what our saying without some sort of justification for that constant.
What I would like to see is the removal of unlimited* offers and FUPs and explicit contention ratios written into service contracts. How can a free market function when the ISPs are refusing to explain what they are offering to customers? This contention ratio would define the transfer per month that the ISP can hold you to, and the level of service that you can expect from them.
You don't seem to have considered the effects of changing bandwidth in the channel. Speech is a type of brain to brain communication, but because we need to convert the symbols into something that can be transmitted and recognised through sound the bandwidth is quite low. What if you had another type of brain to brain communication that wasn't restricted by the abilities of the vocal cords and the ear? If the bandwidth were higher then we could transmit a more complex set of symbols that allowed richer communication.
One staple of telepathy in sci-fi is transmitting mental images or memories. This is something that could be achieved with your wireless commms, although the question of how to plug the transmitter and receiver into the brain is quite complex. No doubt the neurobiologists will have fun decoding that schematic..
Thank you, I do try my best although Zonk has ruined it in a couple of places.
The basis of the BBC's argument has been (as you've stated) that platforms without DRM cannot be trusted. But the reason that this argument is false, and why it is now weakened in the light of their iPhone hole is that you're not worried about redistribution within the trusted enironment. The point is that once something has been stripped out of its DRM covering it can be freely passed around. The idea that Linux is less capable of supporting DRM than windows is a Red Herring. They are both equally capable when it comes to running snakeoil.
The "strong" DRM that the BBC relies on is in fact security through obscurity. The annoying features (such as downloads timing out and self-destructing) can be worked around easily. The "protection" that the BBC has for content distributed through the windows platform is illusionary. Plenty of people had already extracted the FLAC containers from the streaming service before, it was just more of a pain in he arse to do so.
The real point is that it can be done, there is no real protection (only obscurity) and publically stating that there are technical barriers to Mac and Linux support that would take 2 years is laughable. I don't know if you had a look at the three source pages (they're on the second page of the reg article), but the people that are doing this have an agenda. They pay a license fee and they want support. Being told that artificial barriers have been erected to separate them from what they've paid for will not go down well. And if the beeb wants to continue support for the iPhone then they'll need to keep punching holes in the DRM that will be found and exploited.
Lastly, I've already bitched a couple of times about Zonk's incompetence but seriously: The Apple Section? This was supposed to be under YRO as it is a story about DRM being circumvented.
Not for long. The beeb will certainly close this hole quickly - but it weakens their argument that they can't provide a decent service to linux/mac users. Of course once it is closed people will start looking for other holes.
What I don't understand is why Zonk moved the link from the text at the end of the summary about people downloading it for themselves, and placed it on the unrelated text about about the bbc opening up the streaming version? Is this the famous Zonk editing technique that I've heard so much about. If you're listening Zonk, for the love god Move It Back!
Maybe, but our tv isn't huge. It's only 28" so we got it for £500 a few years back ($1000?) It's comfortable for us to watch on the couch, but it's just as convenient for us to curl up in bed and watch movies on my macbook. The guy who replied to you has a point about tv sizes and groups of people, but for day to day stuff we don't need a monster 42" tv. For when we are watching movies with friends it would make sense to get a giant screen just for those occasions - projectors are getting much cheaper and the contrast is getting better.
Two replies, pretty unusual! Your sig is excellent. When you say paraphrased did Godel say something quite similar (which I couldn't find on Google), or do you literally mean that it paraphrases Godel's work. I'm just curious as it's a very cool quote and I was wondering whether I should attribute it to Godel, or to you.
You're not quite putting it in the right terms for the slashdot audience. How about:
When you download a 5Gb Blue-Ray Rip it will look much better than a 1Gb DVD rip if you play it on the right equipment. The right equipment being a display to do it justice, and mplayer to do the upscaling nicely:)
Seriously though, on reading your post I'm shocked by just how much hassle everything is using legal components. We got our TV cheaply as it wasn't "HD-Ready". Apart from the lack of sticker it does do 1280x1024 stretched onto 16:9, and mplayer can do bicubic interpolation in software, so for our purposes it was high-def. Plays 720p wonderfully.
The paper is quite strange, it reads like someone very skilled in physics and chemistry trying (and failing a little) to explain basic computational theory. Their results are really interesting, and this does look like a leap forward, but not the one being reported. The authors don't understand SIMD processing and keep claiming that their advance is the ability to control the 16 molecules in parallel. But the computational model applies the same state change to each of the 16 molecules (which are esssentially a 4-state FSM).
Obviously we can do SIMD manipulations in silicon so that claim is a little hollow. What they've built is strange - it's not quite a memory and not quite a computer. But it is obviously an advance in the state of the art. The STM is used for I/O but all of the processing is done within the constructed molecules. This is quite exciting and the authors are confident that they can expand it to a bigger system. So although the I/O is currently infeasible, the actual computing part does seem realistic.
If I've understood the paper properly then by firing 2-bit commands at the CCU (central molecule) they can cause 16 related commands to be sent to the connected molecules. So this isn't direct storage / retrieval into a memory. They've shown that they can walk through 2^32 distinct states in the array, but getting to a particular point would require knowledge of your starting position and how to walk the state-space. The controlled flipping of the external molecules does look like the basis of control logic for a processor (although it doesn't yet do "computation").
If anyone else has access to the paper then I would recommend reading it as the description of how chemical changes in the molecules are used to encode state changes is amazing. One of the most impressive things that I've read and it sounds like the basis for real computer engineering at the molecular level.
PS. I actually logged into work to read the paper and answer your question about temperature but I can't see it in the paper. On a quick second pass I do notice that the STM is only used to precisely apply a voltage difference over the CCU, which sounds like something that could be replaced to take it out of the lab - the temperature range and noise in the system is probably the killer
Slashdot Mirror
Actually the difference is very clear. Imperative languages modify state while they are running. Functional languages create values. Just because some functional feature have been added to imperative languages, and vica versa does not weaken this distinction. This is why imperative monads are referred to as "pollution" in functional programs and it is common practice (enforced by the type system) to restrict their use to the "outside" parts of a program, ensuring that the core is purely functional.
This is not true in practice. Logical relations are a superset of functions, and so most real logic code is written in predicates that are somewhat similar in both form and function to functional languages. The most common logic language Prolog actually has operational semantics similar to a procedural language. It is very easy to write purely procedural code within Prolog, or to use the extra features that a logic language provides.
Prolog does have a well-defined evaluation ordering for reasons of efficiency, and also because it is quite hard to express the semantics of a language which weakens this ordering. Mercury is one language that moves in that direction, also replacing the builtin arithmetic with a CLP library allows more flexibility in ordering of evaluations.
The meta-programming that you describe is not a specific feature of functional languages. Reflection and Reiterfication can be added to any type of language. Some functional languages (such as lisp) have it as a built-in feature, while some (Haskell if I remember correctly) does not. Amongst the logic languages Prolog has very powerful meta-programming features (there is no separation between code and data and both are manipulated in the same way), and amongst the procedural languages I think that Smalltalk has strong meta-programming facilities althought I haven't used it.
Err... have you got some weird exotic hardware in mind?
In x86 protected mode memory is a single contiguous address range. Sure it is split into different pages but that is not something that you have to deal with as a programmer. Hardware does maintain a stack for you on most architectures. On x86 it's done with a dedicated register and standard addressing modes to build the frame.
It's been a while since I've programmed in Z80, or in 68k, but most other CPU architectures also have contiguous addressing and hardware stacks. The only chips that I've used that don't are PIC microcontrollers.
Why not some external module? That was the design that the Palladium group came up with to solve this very problem - whitelisting software.
No this is completely wrong. Non-proliferation is completely unilateral and is aimed at preventing all non-nuclear states from developing the bomb. Have a read if you are unsure of the terms - but don't make up half-baked analogies to support your incorrect assumptions.
So in your terms, the signatories to the NPT who possess nukes are saying "Nyeh, we want to be the only ones with the bomb". Which is why the poster that I replied to was making such a contorted point, why the US is hypocritical in its policy, and why you are flat out wrong.
That's a contorted argument designed to win a cheap point in an argument. You know exactly what the GP meant: The point of proliferation is that it leads to possession. A country possessing nukes cannot argue against proliferation without being a hypocrite - it is specifically arguing that other countries should not be able to do what it has.
Whoops my bad. It's not in the summary at all, it's in one of the articles. No one could be expected to read that far. Carry on, nothing to see here...
Or maybe use the hop count so that you're measuring network distance rather than geographical distance. You know, the way they describe IN THE SUMMARY.
Erm, it's not actually a mistake. It depends on whether you parse "increase" as being additive, or multiplicative. In everyday english it can be either. So what you're describing is ambiguity rather than error.
Yeah I know. I submitted a followup but I don't think it's going to get voted up enough within firehose. Shame as I think this is a really important topic - how the first major transition of a broadcast network to internet programing gets lumbered wit DRM.
I was quite suprised as I read the story about your father in law because I'd implicitly assumed that you were American, from your libertarian sounding views. Then I read as far as you being Canadian and oddly that story wasn't a suprise. Your country has possibly one of the best international reputations. I think it's really good for them to do that for him, very generous and considerate.
It's an interesting question when you compare no-national-army against no army at all, rather than private armies. At one point I would have thought you naive and idealist for suggesting it, but after the past five years of watching Bush & Blair's wars it doesn't seem like such a bad idea.
Why has the AC been modded down? He explained (with direct quotes) what the Loebner Prize that the researchers are going for is.
Reading the description it sounds like it could be done with current methods and brute-force. In particular restricting the scope of the questions to avoid tripping up the AI removes a large chunk of the problem. It is only a small first step towards the real Turing Test.
For those who want to judge progress themselves, a much better link than the article is the researcher's page. It has videos of their system in action (or it did when I typed this, I suspect it may die shortly...)
What they've done is no more impressive than many current chatbots. The two main differences (that created this new story) are that they are doing it with avators (yay! newmedia-whores2.0), and that they're going to try it out on much bigger hardware than has been tried previously. But even the false beliefs demo is not that impressive if you strip off the Second Life frontend. It is only as complex as a typical demo in blocks world.
Yes, that was my point. Given that there doesn't seem to be prior art they will have to argue obviousness. But showing that something was obvious in hindsight is much harder than showing prior art.
It will be hard to find. The patent was filed on Oct 5th 1999. I can't find exactly when Napster started, but it was first sued in Dec 1999 - so shortly before that. The problem is that Napster is not prior art for this patent. The service described is basically iTunes. It uses a central server with a library of music, and client devices use licenses to download the tracks that they can prove they are entitled to.
I suspect that this patent will be very hard to fight in the US courts because the defense relies on how obvious the invention is. There may be prior art for music distribution, and for downloading licensed content from a central server, but the issue is whether or not there is prior art for the combination. I can't think of any iTunes-like servers back before Napster. Although there were plenty of illegal ftp/web sites they didn't use licenses for obvious reasons.
Looks like these guys saw Napster, guessed which way the wind was going to blow and rushed to file a patent before the music industry organised themselves. I bet that they never implemented / released a product before iTunes. So Apple may have some scope in that it is an obvious combination of well-known pieces, and that ZapMedia didn't even have to invent the product (at least build a prototype) to file the patent.
Wow. You support a private fire service, and private (corporate) armies. That's some mighty fine and mellow crack you're smoking..
I'd agree with what you say, but I would go further. They are short-sighted idiots. Who do they think differential bandwidth pricing is aimed at? Seeing them weigh in on the side of the ISPs is like watching someone hand a knife to a mugger in a dark alley. It's almost as if they haven't realised that in a few years they're going to want to provide high-bandwidth services to customers over those same connections...
Why would it be overselling when they can't meet your arbitrary contention ratio of 10%? Why not 1%? Why not 0.1%? If you're going to stick an arbitrary constant in an equation then you can't use it to back up what our saying without some sort of justification for that constant.
What I would like to see is the removal of unlimited* offers and FUPs and explicit contention ratios written into service contracts. How can a free market function when the ISPs are refusing to explain what they are offering to customers? This contention ratio would define the transfer per month that the ISP can hold you to, and the level of service that you can expect from them.
It's unsurprising but still an abuse of their position. I've already complained to them about the way that article is written.
The interesting question is does it still work on a real iPhone? If it does the hole will be reopening again soon...
You don't seem to have considered the effects of changing bandwidth in the channel. Speech is a type of brain to brain communication, but because we need to convert the symbols into something that can be transmitted and recognised through sound the bandwidth is quite low. What if you had another type of brain to brain communication that wasn't restricted by the abilities of the vocal cords and the ear? If the bandwidth were higher then we could transmit a more complex set of symbols that allowed richer communication.
One staple of telepathy in sci-fi is transmitting mental images or memories. This is something that could be achieved with your wireless commms, although the question of how to plug the transmitter and receiver into the brain is quite complex. No doubt the neurobiologists will have fun decoding that schematic..
Thank you, I do try my best although Zonk has ruined it in a couple of places.
The basis of the BBC's argument has been (as you've stated) that platforms without DRM cannot be trusted. But the reason that this argument is false, and why it is now weakened in the light of their iPhone hole is that you're not worried about redistribution within the trusted enironment. The point is that once something has been stripped out of its DRM covering it can be freely passed around. The idea that Linux is less capable of supporting DRM than windows is a Red Herring. They are both equally capable when it comes to running snakeoil.
The "strong" DRM that the BBC relies on is in fact security through obscurity. The annoying features (such as downloads timing out and self-destructing) can be worked around easily. The "protection" that the BBC has for content distributed through the windows platform is illusionary. Plenty of people had already extracted the FLAC containers from the streaming service before, it was just more of a pain in he arse to do so.
The real point is that it can be done, there is no real protection (only obscurity) and publically stating that there are technical barriers to Mac and Linux support that would take 2 years is laughable. I don't know if you had a look at the three source pages (they're on the second page of the reg article), but the people that are doing this have an agenda. They pay a license fee and they want support. Being told that artificial barriers have been erected to separate them from what they've paid for will not go down well. And if the beeb wants to continue support for the iPhone then they'll need to keep punching holes in the DRM that will be found and exploited.
Lastly, I've already bitched a couple of times about Zonk's incompetence but seriously: The Apple Section? This was supposed to be under YRO as it is a story about DRM being circumvented.
Not for long. The beeb will certainly close this hole quickly - but it weakens their argument that they can't provide a decent service to linux/mac users. Of course once it is closed people will start looking for other holes.
What I don't understand is why Zonk moved the link from the text at the end of the summary about people downloading it for themselves, and placed it on the unrelated text about about the bbc opening up the streaming version? Is this the famous Zonk editing technique that I've heard so much about. If you're listening Zonk, for the love god Move It Back!
Maybe, but our tv isn't huge. It's only 28" so we got it for £500 a few years back ($1000?) It's comfortable for us to watch on the couch, but it's just as convenient for us to curl up in bed and watch movies on my macbook. The guy who replied to you has a point about tv sizes and groups of people, but for day to day stuff we don't need a monster 42" tv. For when we are watching movies with friends it would make sense to get a giant screen just for those occasions - projectors are getting much cheaper and the contrast is getting better.
Two replies, pretty unusual! Your sig is excellent. When you say paraphrased did Godel say something quite similar (which I couldn't find on Google), or do you literally mean that it paraphrases Godel's work. I'm just curious as it's a very cool quote and I was wondering whether I should attribute it to Godel, or to you.
You're not quite putting it in the right terms for the slashdot audience. How about:
:)
When you download a 5Gb Blue-Ray Rip it will look much better than a 1Gb DVD rip if you play it on the right equipment. The right equipment being a display to do it justice, and mplayer to do the upscaling nicely
Seriously though, on reading your post I'm shocked by just how much hassle everything is using legal components. We got our TV cheaply as it wasn't "HD-Ready". Apart from the lack of sticker it does do 1280x1024 stretched onto 16:9, and mplayer can do bicubic interpolation in software, so for our purposes it was high-def. Plays 720p wonderfully.
The paper is quite strange, it reads like someone very skilled in physics and chemistry trying (and failing a little) to explain basic computational theory. Their results are really interesting, and this does look like a leap forward, but not the one being reported. The authors don't understand SIMD processing and keep claiming that their advance is the ability to control the 16 molecules in parallel. But the computational model applies the same state change to each of the 16 molecules (which are esssentially a 4-state FSM).
Obviously we can do SIMD manipulations in silicon so that claim is a little hollow. What they've built is strange - it's not quite a memory and not quite a computer. But it is obviously an advance in the state of the art. The STM is used for I/O but all of the processing is done within the constructed molecules. This is quite exciting and the authors are confident that they can expand it to a bigger system. So although the I/O is currently infeasible, the actual computing part does seem realistic.
If I've understood the paper properly then by firing 2-bit commands at the CCU (central molecule) they can cause 16 related commands to be sent to the connected molecules. So this isn't direct storage / retrieval into a memory. They've shown that they can walk through 2^32 distinct states in the array, but getting to a particular point would require knowledge of your starting position and how to walk the state-space. The controlled flipping of the external molecules does look like the basis of control logic for a processor (although it doesn't yet do "computation").
If anyone else has access to the paper then I would recommend reading it as the description of how chemical changes in the molecules are used to encode state changes is amazing. One of the most impressive things that I've read and it sounds like the basis for real computer engineering at the molecular level.
PS. I actually logged into work to read the paper and answer your question about temperature but I can't see it in the paper. On a quick second pass I do notice that the STM is only used to precisely apply a voltage difference over the CCU, which sounds like something that could be replaced to take it out of the lab - the temperature range and noise in the system is probably the killer