You and I aren't going to see eye to eye on free will. I don't buy it and you do, for reasons that each of us considers convincing- but I have to wonder what your reasons are. I just don't see the evidence, and lets be honest here- proving randomness is a bitch. There are myriad possible explanations for our difficulty in modeling human behavior that do not depend on mystical intervention or unknown and unobservable phenomenon. Why do you believe something that flies in the face of all we know about the rest of the universe, when there is not a shred of evidence to support your claim? Is it because it is a comforting thought to believe that alone among all the collections of atoms in the Universe, we control our fate? Is it a justification of the notions of good and evil? Of God? I really don't understand, and would love to hear a cogent argument for its existence.
Part of the problem you've got in talking about AI is that you are talking about stuff you've seen in movies, not the actual state of AI. AI is creative now- look at the wifi antenna that it created, or the wind instrument playing robots that made headlines just a few weeks ago. Does it depend upon chance, upon permutation? Sure, just the same as a Monte Carlo simulation does, or a genetic algorithm, or a BPNN, but no more so, and you don't call any of those things "intelligent". In fact, I would venture so far as to say that you have a problem of definition on your hands- that you have decided that humanity alone holds the power of free will, of true intellect, and have derived all your subsequent reasoning from that point. That seems to me unworthy of somebody as obviously intelligent as you are, and I hope I'm wrong about it.
Your statement about latency coming from a random source and therefore being random is clearly false, as it is trivial to construct a distinguisher function which imposes arbitrary order onto a random input, as you yourself argue when discussing convergence.
I am familiar with neural networks, and am still not understanding how using a metric that depends upon the performance of your network to dictate the behavior of your network produces random behavior.
Your statement about modeling truly random behavior is, so far as I can tell, entirely in error. Perhaps you meant that it was possible to model systems whose state includes random information? It is most certainly not the case that you can model a truly random system with a non-negligible chance of success.
I'm glad you enjoy the mental exercise, I have too, but I'm still not getting your fundamental premise. You don't seem to be concerned in the slightest that there isn't any evidence for it, that it is impossible to build constructions based on, or that it is purely unfalsifiable. I find that strange, and look forward to your response.
Indeed, there are literally thousands of extant solutions to this problem- but since the asker is apparently too lazy to find any of them via Google, I figured I'd offer.
Honest to God, what you're talking about is a trivial task. Use ghostscript, or, if you don't have the time or interest, contact me with your requirements and I'll write it for you gratis, provided it remains F/OSS.
+ we have seen some instances where clause 3 was ignored by groups
and organisations Interesting take on it. Would that more folks saw common ground above license lawyering.
I'm afraid I can't follow you on your leap of faith, there. The difference between chaotic and random is BIG, a far more fundamental difference than you are making it out to be, and we just don't see anything that indicates any kind of "ghost in the machine".
As far as network latency goes, it is, once again, chaotic, not random. It has the property of being immensely difficult to predict with perfect accuracy, but, very importantly, has an excellent chance of being predicted given a rather limited margin of error and adjacent sample spaces. Your point may be that either form of indeterminacy is acceptable, but if that is the case why do you posit that "free will", if it exists, cannot emerge from deterministic systems?
Also, I'm not getting why you're using network latency to generate pseudorandomness for your MPNNs? Doesn't that cause bad loading behavior? And you're using 1 node per neuron?
I'm not sure why you say that modeling the brain exceeds the available power, since Blue Brain is already capable of modeling significant portions of a mammalian brain.
Honestly, and I hope you don't take this the wrong way, I hope DARPA doesn't fund your ideas. Maybe you're ahead of your time, but more likely I think you're headed down a path that is comfortably self-justifying but ultimately inaccurate. Will randomness be a part of any "solution" in the foreseeable future? Yes. Does that derive from any deep-seated meaning about the nature of man or the universe? That I find very dubious.
That is uncomfortably close to the conversation I had with them- except instead of pointing out the price, I pointed out the giant "GOING OUT OF BUSINESS SALE" sign outside their front window and declined.
I've actually liked the cheapo HP all-in-ones. I got one as part of a bundle when CompUSA went under, and since then have gotten two more (one for my grandparents, one for work use) and all of them work pretty well, especially for ~$40 each.
Last time I checked, all strong RNGs are based on either quantum or chaotic effects, with quantum effects being the mathematically stronger of the two.
Don't take this the wrong way, but I think you're drawing conclusions based on some serious misunderstandings, a large leap of faith, and an unfamiliarity with the fields in question.
As far as the requirement for "free will" in computer systems, you've put the cart before the horse and assumed that free will must exist for a system to simulate the mind, without ever proving that the mind is anything other than a deterministic system of unbelievable complexity. To presume that it is nondeterministic because you cannot adequately predict its behavior is pretty obviously bad logic.
The human brain does not take advantage of any known large-scale quantum effects, and, so far as we know, does not exploit any of them to produce random behavior. Once again, the inability to demonstrate a pattern is not evidence that a pattern does not exist.
Asynchronous computing does not produce or take advantage of quantum uncertainty. The levels of quantum uncertainty involved are swallowed by the impact of the deterministic systems they are filtered through, and drowned out by the impact of chaotic but deterministic variations in process scheduling, resource locking, and timing conflicts. The same goes for parallel computing for the same reasons- network latency is a chaotic, not random, phenomenon.
In terms of the use of quantum uncertainty for intelligent systems, there is no doubt that quantum computing holds tremendous promise, but also that its applications are hugely misunderstood. It is not a cure-all for general computing problems, and it particularly does not solve the problem of being insufficiently able to describe the your problem.
Bottom line is that chaos != randomness, and unpredicted != unpredictable. What you've got is good philosophy, but does not accurately depict the state of AI or what we know about the systems you are describing.
A lot of the older AI research is pure theory, but in the last 20 years or so it has been driven by the realization that we don't really have the tools to meet some of the early expectations of the field. If you are interested in the theoretical foundations of AI, though, you might want to look into compression, data representation, and computability, as well as general information theory. Claude Shannon's work would be a good place to start, and is cited frequently enough to give you a guided tour through AI.
Re:Google Andriod is about to be hit by a steamrol
on
Android Phones Delayed
·
· Score: 1
I'm not ready to call Android the second coming, but if Google stays true to form, it is likely to be a very good product with an amazing marketing drive behind it. Nokia is really going to have to make something happen to make Qtopia more than just another cell phone platform, and so far it doesn't seem like a lot of people are convinced- especially the rival cell phone manufacturers who probably feel better about licensing software from Google than from a competitor. My guess is that Android and Qtopia will coexist for some time, with Qtopia eventually settling for a percentage of the low-end smartphone and high-end cellphone markets, and Android competing more with the iPhone and Windows Mobile devices.
Let me preface this by saying that I am not an expert in TPMs. I haven't looked into the technology in more than a year, and it may have changed significantly since then. You probably know more about them than I do. Having said that:
1) TPMs are far from universal, and have far from universal OS support.
2) The level of support needed to verify that there was not an X proxy in operation would be extremely burdensome. In addition to having to verify the existence and correctness of software- which TPMs were, last time I checked, pretty good at- it would have to verify the absence of software, a much more difficult task, and one historically prone to circumvention.
3) Last time I checked, no desktop-ready system capable of verifying all running software existed. That may no longer be the case- I'm not sure- but it seems unlikely to exist in an environment characterized by a large proportion of free/Free or open source software.
4) To extend the earlier argument, DRM utilizing TPMs in such a way in heterogeneous environments would have to be able to account for different versions of different software, fail open, or simply decline to support most environments. For some vendors that's acceptable, but the annoyance factor to end users would probably get out of hand pretty quickly. It seems more likely to me that most vendors utilizing DRM would verify only the components under their direct control- and that the X server probably isn't on that list.
5) TPMs, last time I looked, were vulnerable to some of the same data recovery methods that SSDs were. I don't remember where I saw it, or if today's would be vulnerable as well, but you probably know what I'm referring to.
6) In the worst case scenario- that the OS cooperates totally with DRM, that the vendor is capable of verifying all the hardware and software on your system against any possible change, that they are additionally able to verify the nonexistence of any software, that the TPM itself cannot be compromised or circumvented, that the remote attestation mechanism cannot be circumvented or forged, and that all data storage is unbreakably encrypted, the *worst case scenario* is still that you get the same as dead tree format. Combined with OCR, it is entirely possible that limited search functions could be obtained, making the performance of the worst case scenario system equivalent to or slightly better than the dead tree version.
All of this is predicated on the idea that there is no more specific attack on the software, TPM, etc. If history is any guide, there will be, even if (and this is a big if) this system actually makes it off the ground. On the other hand, I look forward to hearing how behind the times I am on TPMs.
Believe it or not, properly engineered 802.11a has the same or better range, throughput, and scalability. That was the original post, to which I replied. I understand that any radio technology can, with unlimited gain and/or power, be delivered at any range. I still haven't seen anything indicating that 802.11a can match the range of WiMAX in their common application- point to multipoint- without breaking FCC regulations. Throughput is undisputedly WiFi's, but scalability goes to WiMAX without question. If that's what we're arguing, let's argue it out, if not, don't be a jackass.
Like I say earlier in this thread, more complex solutions yield correspondingly more powerful results. A screenshot can only display information, but a custom X proxy or modified graphics toolkit would be able to harvest text and formatting information.
Very few formats are designed for cryptographic strength. They do provide small amounts of diffusion and confusion, burying the signal in noise and raising the difficulty in obtaining the information- but not providing much protection against a determined and skilled attacker. In the best case, it winds up moving the point of attack from the transmission to the points at which it is stored, interpreted, or presented. An example would be the idea of packing your data and a custom reader application into a signed virtual machine- on its face, it sounds good, and it certainly makes the existing attacks more difficult to pull off, but a simple screenshot application obviates the entire system. Variations on this theme have plagued DRM systems in the past- recording from the line out, copying videos from a browser cache, etc etc etc, and unless the fundamental information theoretic problems with DRM are solved, will continue to pose problems.
I hear you- but you still have to contend with being in unlicensed spectrum, which limits your peak gain. Those fancy shootout rigs get dispensations or sublease spectrum, as do the satellite rigs, and all are point to point. WiMAX is point to multipoint, and operates more efficiently inside of its spectrum allocation, thus allowing more total throughput and more total users. The range on WiMAX scales better than does 802.11a, requiring less power and lower gain for the same throughput, and in the real world, working inside of the legal and logistical hurdles is much more important than the theoretical capabilities of the radio technology. It is also worth pointing out that none of the wifi shootout records are set with A, preferring B and G for the lower frequency, which behaves better in NLOS scenarios. Again, please demonstrate your point.
While I admit that you raise the difficulty of breaking such a scheme considerably by doing all of that, it is far from "unhackable". My first thought would be to run the entire VM under an X proxy, which would permit me to capture image files of each individual page. Another possibility would be to take snapshots of the VM in operation and correlate the opening of the file to memory allocation. Either way, you're fighting against Knuth's third law, and you just aren't going to win that one.
Slashdot Mirror
You and I aren't going to see eye to eye on free will. I don't buy it and you do, for reasons that each of us considers convincing- but I have to wonder what your reasons are. I just don't see the evidence, and lets be honest here- proving randomness is a bitch. There are myriad possible explanations for our difficulty in modeling human behavior that do not depend on mystical intervention or unknown and unobservable phenomenon. Why do you believe something that flies in the face of all we know about the rest of the universe, when there is not a shred of evidence to support your claim? Is it because it is a comforting thought to believe that alone among all the collections of atoms in the Universe, we control our fate? Is it a justification of the notions of good and evil? Of God? I really don't understand, and would love to hear a cogent argument for its existence.
Part of the problem you've got in talking about AI is that you are talking about stuff you've seen in movies, not the actual state of AI. AI is creative now- look at the wifi antenna that it created, or the wind instrument playing robots that made headlines just a few weeks ago. Does it depend upon chance, upon permutation? Sure, just the same as a Monte Carlo simulation does, or a genetic algorithm, or a BPNN, but no more so, and you don't call any of those things "intelligent". In fact, I would venture so far as to say that you have a problem of definition on your hands- that you have decided that humanity alone holds the power of free will, of true intellect, and have derived all your subsequent reasoning from that point. That seems to me unworthy of somebody as obviously intelligent as you are, and I hope I'm wrong about it.
Your statement about latency coming from a random source and therefore being random is clearly false, as it is trivial to construct a distinguisher function which imposes arbitrary order onto a random input, as you yourself argue when discussing convergence.
I am familiar with neural networks, and am still not understanding how using a metric that depends upon the performance of your network to dictate the behavior of your network produces random behavior.
Your statement about modeling truly random behavior is, so far as I can tell, entirely in error. Perhaps you meant that it was possible to model systems whose state includes random information? It is most certainly not the case that you can model a truly random system with a non-negligible chance of success.
I'm glad you enjoy the mental exercise, I have too, but I'm still not getting your fundamental premise. You don't seem to be concerned in the slightest that there isn't any evidence for it, that it is impossible to build constructions based on, or that it is purely unfalsifiable. I find that strange, and look forward to your response.
Indeed, there are literally thousands of extant solutions to this problem- but since the asker is apparently too lazy to find any of them via Google, I figured I'd offer.
That's a mighty "if" you've got there, friend...
Honest to God, what you're talking about is a trivial task. Use ghostscript, or, if you don't have the time or interest, contact me with your requirements and I'll write it for you gratis, provided it remains F/OSS.
I'm afraid I can't follow you on your leap of faith, there. The difference between chaotic and random is BIG, a far more fundamental difference than you are making it out to be, and we just don't see anything that indicates any kind of "ghost in the machine".
As far as network latency goes, it is, once again, chaotic, not random. It has the property of being immensely difficult to predict with perfect accuracy, but, very importantly, has an excellent chance of being predicted given a rather limited margin of error and adjacent sample spaces. Your point may be that either form of indeterminacy is acceptable, but if that is the case why do you posit that "free will", if it exists, cannot emerge from deterministic systems?
Also, I'm not getting why you're using network latency to generate pseudorandomness for your MPNNs? Doesn't that cause bad loading behavior? And you're using 1 node per neuron?
I'm not sure why you say that modeling the brain exceeds the available power, since Blue Brain is already capable of modeling significant portions of a mammalian brain.
Honestly, and I hope you don't take this the wrong way, I hope DARPA doesn't fund your ideas. Maybe you're ahead of your time, but more likely I think you're headed down a path that is comfortably self-justifying but ultimately inaccurate. Will randomness be a part of any "solution" in the foreseeable future? Yes. Does that derive from any deep-seated meaning about the nature of man or the universe? That I find very dubious.
That is uncomfortably close to the conversation I had with them- except instead of pointing out the price, I pointed out the giant "GOING OUT OF BUSINESS SALE" sign outside their front window and declined.
Absolutely. I have a feeling that we're all going to feel very stupid about who we put up on pedestals when all is said and done.
Not sure I'm understanding what you mean by that. I will admit that the situation with GPL'd drivers and kernel modules is insane, though.
Trolling or serious? I can't even tell anymore. And isn't most artistic license code dual licensed under GPL?
So, I guess you'll be writing DirectX for Linux, then? And yes, I see your point. I just wish I didn't have to. Gamers make my teeth ache.
I've actually liked the cheapo HP all-in-ones. I got one as part of a bundle when CompUSA went under, and since then have gotten two more (one for my grandparents, one for work use) and all of them work pretty well, especially for ~$40 each.
Last time I checked, all strong RNGs are based on either quantum or chaotic effects, with quantum effects being the mathematically stronger of the two.
Don't take this the wrong way, but I think you're drawing conclusions based on some serious misunderstandings, a large leap of faith, and an unfamiliarity with the fields in question.
As far as the requirement for "free will" in computer systems, you've put the cart before the horse and assumed that free will must exist for a system to simulate the mind, without ever proving that the mind is anything other than a deterministic system of unbelievable complexity. To presume that it is nondeterministic because you cannot adequately predict its behavior is pretty obviously bad logic.
The human brain does not take advantage of any known large-scale quantum effects, and, so far as we know, does not exploit any of them to produce random behavior. Once again, the inability to demonstrate a pattern is not evidence that a pattern does not exist.
Asynchronous computing does not produce or take advantage of quantum uncertainty. The levels of quantum uncertainty involved are swallowed by the impact of the deterministic systems they are filtered through, and drowned out by the impact of chaotic but deterministic variations in process scheduling, resource locking, and timing conflicts. The same goes for parallel computing for the same reasons- network latency is a chaotic, not random, phenomenon.
In terms of the use of quantum uncertainty for intelligent systems, there is no doubt that quantum computing holds tremendous promise, but also that its applications are hugely misunderstood. It is not a cure-all for general computing problems, and it particularly does not solve the problem of being insufficiently able to describe the your problem.
Bottom line is that chaos != randomness, and unpredicted != unpredictable. What you've got is good philosophy, but does not accurately depict the state of AI or what we know about the systems you are describing.
We don't need a successor, we can't get the one we've got working.
Good work- you blew a fuse on my sarcasmomiter.
A lot of the older AI research is pure theory, but in the last 20 years or so it has been driven by the realization that we don't really have the tools to meet some of the early expectations of the field. If you are interested in the theoretical foundations of AI, though, you might want to look into compression, data representation, and computability, as well as general information theory. Claude Shannon's work would be a good place to start, and is cited frequently enough to give you a guided tour through AI.
I'm not ready to call Android the second coming, but if Google stays true to form, it is likely to be a very good product with an amazing marketing drive behind it. Nokia is really going to have to make something happen to make Qtopia more than just another cell phone platform, and so far it doesn't seem like a lot of people are convinced- especially the rival cell phone manufacturers who probably feel better about licensing software from Google than from a competitor. My guess is that Android and Qtopia will coexist for some time, with Qtopia eventually settling for a percentage of the low-end smartphone and high-end cellphone markets, and Android competing more with the iPhone and Windows Mobile devices.
No idea; force of habit, perhaps? You pretty much jumped in on a point somebody else was making.
Let me preface this by saying that I am not an expert in TPMs. I haven't looked into the technology in more than a year, and it may have changed significantly since then. You probably know more about them than I do. Having said that:
1) TPMs are far from universal, and have far from universal OS support.
2) The level of support needed to verify that there was not an X proxy in operation would be extremely burdensome. In addition to having to verify the existence and correctness of software- which TPMs were, last time I checked, pretty good at- it would have to verify the absence of software, a much more difficult task, and one historically prone to circumvention.
3) Last time I checked, no desktop-ready system capable of verifying all running software existed. That may no longer be the case- I'm not sure- but it seems unlikely to exist in an environment characterized by a large proportion of free/Free or open source software.
4) To extend the earlier argument, DRM utilizing TPMs in such a way in heterogeneous environments would have to be able to account for different versions of different software, fail open, or simply decline to support most environments. For some vendors that's acceptable, but the annoyance factor to end users would probably get out of hand pretty quickly. It seems more likely to me that most vendors utilizing DRM would verify only the components under their direct control- and that the X server probably isn't on that list.
5) TPMs, last time I looked, were vulnerable to some of the same data recovery methods that SSDs were. I don't remember where I saw it, or if today's would be vulnerable as well, but you probably know what I'm referring to.
6) In the worst case scenario- that the OS cooperates totally with DRM, that the vendor is capable of verifying all the hardware and software on your system against any possible change, that they are additionally able to verify the nonexistence of any software, that the TPM itself cannot be compromised or circumvented, that the remote attestation mechanism cannot be circumvented or forged, and that all data storage is unbreakably encrypted, the *worst case scenario* is still that you get the same as dead tree format. Combined with OCR, it is entirely possible that limited search functions could be obtained, making the performance of the worst case scenario system equivalent to or slightly better than the dead tree version.
All of this is predicated on the idea that there is no more specific attack on the software, TPM, etc. If history is any guide, there will be, even if (and this is a big if) this system actually makes it off the ground. On the other hand, I look forward to hearing how behind the times I am on TPMs.
Like I say earlier in this thread, more complex solutions yield correspondingly more powerful results. A screenshot can only display information, but a custom X proxy or modified graphics toolkit would be able to harvest text and formatting information.
Very few formats are designed for cryptographic strength. They do provide small amounts of diffusion and confusion, burying the signal in noise and raising the difficulty in obtaining the information- but not providing much protection against a determined and skilled attacker. In the best case, it winds up moving the point of attack from the transmission to the points at which it is stored, interpreted, or presented. An example would be the idea of packing your data and a custom reader application into a signed virtual machine- on its face, it sounds good, and it certainly makes the existing attacks more difficult to pull off, but a simple screenshot application obviates the entire system. Variations on this theme have plagued DRM systems in the past- recording from the line out, copying videos from a browser cache, etc etc etc, and unless the fundamental information theoretic problems with DRM are solved, will continue to pose problems.
I hear you- but you still have to contend with being in unlicensed spectrum, which limits your peak gain. Those fancy shootout rigs get dispensations or sublease spectrum, as do the satellite rigs, and all are point to point. WiMAX is point to multipoint, and operates more efficiently inside of its spectrum allocation, thus allowing more total throughput and more total users. The range on WiMAX scales better than does 802.11a, requiring less power and lower gain for the same throughput, and in the real world, working inside of the legal and logistical hurdles is much more important than the theoretical capabilities of the radio technology. It is also worth pointing out that none of the wifi shootout records are set with A, preferring B and G for the lower frequency, which behaves better in NLOS scenarios. Again, please demonstrate your point.
While I admit that you raise the difficulty of breaking such a scheme considerably by doing all of that, it is far from "unhackable". My first thought would be to run the entire VM under an X proxy, which would permit me to capture image files of each individual page. Another possibility would be to take snapshots of the VM in operation and correlate the opening of the file to memory allocation. Either way, you're fighting against Knuth's third law, and you just aren't going to win that one.