Good to know that about measels. I may have been confusing german measels (which we did have, along with chicken pox) with real measels. Obviously we didn't have MMR back then, but maybe we were vaccinated against M&M.
Interesting also to know that about DT(a)P. I don't know if my childhood doctor was an idiot, but at least he made house calls!;-) As it happens I was later vaccinated against tetanus and diptheria as part of a school trip to the middle east.
Actually, given that one of the main symptons of autism is a lack of social and empathetic skills, I wouldn't totally discount that kids becoming less connected and spending so much time watching TV could conceivably have something to do with it. I would certainly anyway intend to limit the amount of TV my child watches, as most parents do.
Jenny McCarthy may not know what causes autism, but let's not forget that neither do doctors.
Tangentially related there was a documentary on TV last night about scientists discovering the mechanism of Gecko's adhesion to vertical surfaces (molecular Van de Waals forces). What was interesting - and hardly unique to this experiment - was that their initial experimental setup was unable to reproduce the adhesive effect leaving them scratching their heads, and it was only due to persistence and being sure that it was the experimental setup not the hypothesis that was wrong that caused them to persevere and eventually discover that you need to "back-up" the microscopic hair against a surface to get the "adhesive" effect.
The relevance? Maybe the experiments that have disproved any link between autism and vaccines are wrong, but scientists have abandoned them because they are so sure the connection is NOT there, rather than keeping at trying to prove it because of being sure that it IS the cause. One thing's for sure - I bet when the cuase is eventually discovered, there'll be some scientists somewhere having to admit they were wrong or at least surprised. In the meantime, don't trust a scientist who tells you that anything is impossible, because science can't prove impossibilities - it can only predict outcomes.
Since the cause of autism is unknown, but the incidence of it is up the last few decades, it seems quite likely that at least one of the assertions regarding what supposedly doesn't cause it may be wrong. Maybe the experiments that have "proved" MMR to be safe didn't reproduce the right conditions or test against the (unidentified) group most at risk of side effects... Maybe it's DTaP that's really the cause, maybe it's the cumulative effect of so many vaccines that does it (or maybe it is indeed unrelated to vaccines, and our source of ignorance lies elesewhere)..
In the meantime, until science has identified what is causing autism, it'd be better to have a bit more humility over what we don't know rather than asserting that for sure MMR/etc are safe. It may save you from a lawsuit to later be able to tell the parent of an autistic child ("but all the tests at the time said the vaccine was safe"), but that doesn't help the parent or the child at all. I'd say err on the side of caution and weight the risks. Don't just mindlessly vaccinate against everything possible, and use the most convenient multi-disease shots (MMR, DTaP).
I was born in the UK in 1961, and so grew up in the era where we weren't vaccinated against things like measels and chicken pox, and so of course we caught them, and we were fine. There may be rare side effects of these diseases, but the coincident rise in autism coupled with the rise in vaccination at least doesn't indicate autism as one of the side effects. As it happened I also almost died as an infant as a result of the DTP vaccine, and consequently wasn't given the 2nd shot of the series. I did subsequently catch whooping cough, and although it was unpleasant, it's sure better than being dead.
Since my wife is due with our first child in January, the topic of vaccinations has much personal relevance, and while I certainly intend to vaccinate our child in general, I will be looking for a pediatrician who isn't too blase about the potential risks given our lack of knowledge about what does autism.
First, imperative programming specifies the order of evaluation, whilst functional programming does not.
Yes, but there's nothing to stop you from building parallel processing libaries such as pthreads, OpenMP, CUDA or MapReduce using imperative languages - the semantics of your code are up to you. This seems to be a much more useful and practical way to take advantage of parallel hardware - keep the familiarity of the languages you know and then pick from a smorgasborg of parallel execution libraries as fits your needs. Relegating parallelism to libraries also has the advantage of freeing you from the limitations of the compiler... say that your Haskell compiler is able to generate "multi-threaded" code for multi-core processors, that's very nice, but what if you really want to spread your computation over multiple nodes in a cluster, or you want to use the power of your GPU-CPU video card to accelerate your application? Presumably you could write Haskell libraries to do this, but how is that much different from doing the same in C/C++ instead?
A crossbar switch in of itself isn't the solution to increased memory access bandwidth. What you need is to increase parallelism of memory access per CPU which means increasing the number of independent memory banks per CPU - one bank per N cores. You also then need (for global memory) to provide a way for each core group to access non-local memory memory banks (a shared bus would be sufficient).
e.g. something like this:
Cores 1 -4 = Memory bank 1 =+
=================+ Cores 5 -8 = Memory bank 2 =+
=================+ Cores 9-12 = Memory bank 3 =+
=================+ Cores 13-16 = Memory bank 4 =+
=================+
This type of memory/core architecture could easily map onto todays software. For example, have each process assigned to a single core group, such that pthread_create() assigns threads to cores only in that core group, and malloc() allocates memory only from the corresponding local memory bank. You'd presumably map part of the virtual address space onto the local memory bank and part onto the global address space (accessed via the shared bus/switch). The inter-process shared memory APIs (mmap, etc) would allocate memory from whereever they saw fit and then map it into the virtual address space accordingly.
If you wanted to allow a single process to create threads across multiple core groups (maybe controlled via processor affinity), then you could introduce something like pthread_malloc() to allocate local per-thread memory, have malloc() allocate per-process memory from the shared access space. You'd probably only want to have this cross-core-group thread allocation happen when explicity requested via the thread library so that existing binaries and recompiled applications would automatically only use the more efficient local memory, and only heavily threaded apps written to also use pthread_malloc() might ask to be spread across multiple core groups.
The idea here would be that each thread would use local memory as much as possible, thereby achieving maximum parallelism of memory access. Global memory access would result in less parallelism since you'd usually be competing for memory access with the local core group, as well as for global memory bus access. You could choose to implement a crossbar switch in place of the global memory bus, but that would be optional.
As both educators and researchers, we are amazed at the hype that the MapReduce proponents have spread about how it represents a paradigm shift in the development of scalable, data-intensive applications.
But MapReduce isn't about data-intensive applications... It's about scalable/massive primarily COMPUTE-intensive applications (which may or may not also be data-intensive - e.g. to compute the Madlebrot set you don't need any data at all - just massive compute power).
The problem MapReduce is solving is "My PC isn't fast enough to run my aplication, and I can't afford a supercomputer, so how can I make use of the aggregate COMPUTE (& I/O) speed of lots of cheap PCs to run my application faster?"
Writing parallel applications is hard, so what MapReduce does is provide one - limited, but yet quite widely applicable - easy to use way to parallelize compute intensive applications so that at run-time you can give them as many machines as you have available to run on and they'll automatically spread themselves over them (while incidently also providing for fault-tolerance and error recovery).
It's hardly any secret that CPU speed, even for single core processors, has been running ahead of memory bandwidth gains for years - that's why we have cache, and ever increasing amounts of it. It's also hardly any relevation to realize that if you're sharing your memory bandwidth between multiple cores then the bandwidth available per core is less than if you weren't sharing. Obviously you need to keep the amount of cache per core and the number of cores per machine (or, more precisely, per unit of memory sybsystem bandwidth) within reasonable bounds to keep it usable for general purpose aplications, else you'll end up in GPU-CPU (e.g. CUDA) territory where you're totally memory constrained and applicability is much less universal.
For cluster-based ("supercomputer") applications, partitioning between nodes is always going to be an issue in optimizing performance for a given architecture, and available memory bandwidth per node and per core is obviously a part of that equation. Moreover, even if CPU designers do add more cores per processor than is useful for some applications, no-one is forcing you to use them. The cost per CPU is going to remain approximately fixed, so extra cores per CPU essentially come for free. A library like pthreads, and different implementations of it (coroutine vs LWP based), gives you the flexibility over the mapping of threads to cores, and your overall across-node application partitioning gives you control over how much memory bandwidth per node you need.
Rather than dismissing Hawkins' HTM as just an ANN approach and moving on, I'd take a higher level view... what he's done is look at the undeniably regular, and perhaps surprisingly simple, structure of the neocortex, and try to figure out what the architecture is in abstract terms and what it's accomplishing. Of course it's true that in doing this he's making many simplifying assumptions, such as focusing on the Hebian function of neurons, but I think it's hard to argue that this is a reasonable starting point when one's goal is to understand the gross function of the connectivity pattern, and his results anyway rather justify the approach...
Hawkins' analysis of the cortex is that it's implementing an HTM, and this HTM level is the focus of both his book and of the Numeta software. It's only an ANN to the extent that the architecture is inspired by an ANN-like understanding of the cortical architecture.
In analyzing the cortex as an HTM, Hawkins has undeniably provided answers to many questions about how the cortex achieves things such as invarient representation, and marrying top down stored experience with bottom up sensory input to achieve perception. The Numenta SDK includes working examples of simple image recognition and speaker recognition. Does this view of the cortex as an HTM capture all it's nuances? Probably not, but I'm willing to bet that this is it's primary functionality and that further analysis will just add refinement to this model.
Of course the understanding of the neo-cortex as an HTM would fall totally flat if this didn't also explain cognition and intelligence, but you can't judge that until you've distilled the gazillion phenomenal manifestations of intelligence down to a minimalist abstract definition of what it actually is and therefore are in a position to attempt to bridge whatever divide you've left between mechanism (HTM) and fucntionality (intelligence). I can't recall how sucinctly Hawkins himself defines intelligence, or how well he explains it, but I can say that having thought about this myself - seriously - for a couple of decades, I would myself define intelligence in a single short sentence, and do indeed believe that an HTM in essence is all you need to achieve it.
If Hawkins has failed to explain things well enough to convince many people, then I think the failure is not of his architecture (not unique, but a useful summing up of ideas many others have had) but rather a failure of the audience to realize what it is that needs explaining!;-)
Hawkin's isn't saying that complexity or computer power alone will get us to intelligence.
It's obvious that simulating a human brain, even at a high level of abstraction, will take a lot of compute power, memory, etc.
If you take, say, a neural net of a worm's brain and speed it up by a factor of a few million all you will have is a very fast worm.
However, if you take a architecture that is representative of our own brain (but say, with a cortex only 1/1,000,000 the size of ours, and only overall running at 1/1,000,000,000 of the aggregate speed), and you speed/scale THAT up accordingly, then you will in fact have something that has human like intelligence.
You are saying, and are claiming that Hawkin's is saying, that complexity/scaling won't lead to intelligence, but you are wrong. Scaling the RIGHT architecture will lead to it, scaling the wrong architecture will lead to super fast worms, etc.
It sounds as if you're describing a feed-forward network. Things get much more interesting once you bring feedback paths into the picture. Try googling "Adaptive Resonance Theory (ART)" for one particular architecture, or consider your own grey noodle as the ultimate proof of concept of the power of neural nets!
The term AI is so nebulous that it doesn't really mean much of anything. It's more of a functional goal (computer-based human-like ability) than anything more concrete, and as anything that may fall under that general umbrella does become better understood or more concrete, then it tends to be no longer regarded as part of AI (e.g. machine learning, expert systems, speech recognition).
It's also worth noting that natural intelligence is also a rather nebulous concept - you'll find many definitions offered (e.e. an ability to learn and generalize), but almost any definition is doing to be met by some people with "well, that's PART of it...".
The larger problem is that the goal of AI really isn't artificial intelligence(!) - it's an artificial brain/mind. The emphasis on intelligence/cognition is perhaps why so many attempts to achieve AI have failed - because there's no theory of mind or overall brain architecture behind them. Artificial cognition itself is anyways arguably an already solved problem via general symbolic problem solvers like Allen Newell's SOAR, but the ability to manipulate and further refine knowledge isn't of much use if you can't aquire it in the first place. Pairing a cognition capability with an artificially aquired body of real-world knowledge, as done by Cyc, is no better because it is brittle and inflexible - an artifical brain needs to be able to distill it's own ever-changing categories and knowledge out of raw analog sensory input, and this means that symbolic approaches (fuzzy or not) are at best only a small part of the solution (more likley no part at all).
What you really need to do before getting into any technologies that may help implement "AI" is to define what it is that you want to implement. If you want to create an artificial brain, then you will need to figure out what a brain is and how to decompose it into functional blocks - an architecture. Biology texts may be of more use than "AI" ones, as even AI luminaries such as Marvin Minsky have produced laughably simple "theories of the mind". Hawkin's "On Intelligence" is a better start than many, but he also is really focusing on the problem of intelligence, but at least addressing the issue of how to mate this to the perceptual system.
It's only once you know what you want to build (i.e. the hard part of figuring out how the brain works) that the issue of how to implement it comes up, and what existing technologies may or may not be of use (Neural nets, HTMs, SVMs, etc). If you start to implement before having a fairly complete theory of the brain/mind, then your results will only be as good as your goals.
Sure most people will run whatever OS their new computer comes with (and there's not usually a choice), and one side effect of that is that Vista must be reducing sales of new hardware. Case in point is myself... I was casually browsing for black friday laptop deals to replace my wife's aging Toshiba Satellite (400MHz Celeron - still just about tolerable for web browsing), but was put off by the fact that all the ones I say came with Vista (and no mention of XP downgrade option, so who knows if they have XP drivers available).
Sure I'll eventually have to replace that laptop, and probably run whatever it comes with, but for the time being the fact that the new ones come with Vista didn't make me become a Vista user - it meant that I chose not to purchase a new laptop!
Maybe with a tanking economy and slowed sales, the hardware manufacturers will put pressure on Microsoft to fully revive Windows XP - rather like Coke reviving "Coke Classic" after the "New Coke" debacle.
Right now it may only make the less complex games playable (maybe also enabling the full Vista aero UI without a graphics card), but given the pace of processor speed improvement that will change. Double the CPU clock and the FPS doubles, double or quadruple the number of cores and the FPS again goes up by the same factor...
I don't think it's be too long before most games have playable frame rates using this software renderer, even if not at the highest resolution, or with the highest quality settings enabled, but it'll be a process of continual improvement. In the meantime even if the most demanding games arn't playable under this, there's much other less demanding software for which it'll be immediately useful, and given that even the most-stripped down budget machines still have todays insanely fast processors (even if not the bleeding edge), it brings an ever increasing level of capability to the lowest level.
The curriculum may have gotten dumber, and the assessment bar set lower, but even without those additional factors the abandonment of the split grammar school vs comprehensive (& corresp. "O" level" vs CSE exams) has to be a large part of this.
Under the old system (I went to grammar school) the academically faster kids went to grammar schoool and the slower / less able ones went to secondary modern. The purpose of this system was to let each group advance at it's own pace, rather than having to gear the curriculum and pace to the average or less able students as would be needed without this seperation.
Of course eventually the politically correct crown got their way and it was deemed unfair to seperate kids based on ability, so now the brighter kids are doomed to the GCSE curriculum/pace of everyone else. If you're not even TRYING to educate the brighter kids to the best of their ability, is it at all surprising, or even newsworthy, that their level of education/ability is lower than when you were trying to?!
Of course this whole dumbing down appears to have gone further than the abolishment of ability streaming, and the usual response to slipping exam grades seems to be lowering the bar rather than increasing the quality of education.
From NVidia's CUDA site, most of their regular display cards support CUDA, just with less cores (hence less performance) than the Tesla card. The cores that CUDA uses are what used to be called the vertex shaders on your (NVidia) card. The CUDA API is designed so that your code doesn't know/specify how many cores are going to be used - you just code to the CUDA architecture and at runtime it distrubutes the workload to the available cores... so you can develop for a low end card (or they even have an emulator) then later pay for th hardware/performance you need.
Has anyone ever tried to do this - to clone a modern species using the egg of a different, but closely related, species?
I guess it's not too different to what Craig Venter is doing with his minimal bacteria - putting synthetic DNA into the cell of the original from which it was derived (by selectivey removing about 20% of its DNA). No word on whether he's actually bought the thing to life yet though.
Einstein has been cloned from a skidmark in a pair of his underpants kept by his family.
A skull-bong found in an MIT dormitary has been DNA identified as belonging to Isaac Newton. The DNA came from fingerprints on the bong, confirming rumors of Newton being a pothead.
A frozen turd found in Craig Venter's kitchen fridge, formally believed to have come from a pygmy marmoset, has been identified and carbon dated as the last movement of Wolfgang Armadeus Mozart.
If you believe in mind/body dualism then obviously you can't equate anything in the material domain with anything in the "spiritual" domain. Kurzweil isn't a moron so it's safe to assume he's not making that assertion.
The question is only meaningful (vs trivially false) if one assumes that by "soul" one means some NON-dualistic phenomena, so Kurzweil is simply choosing to assume that is what the questioner really means and answering it accordingly, even definining consciousness before doing so. You could hardly ask for a clearer or more concise answer to the question!
Well, IANAL, and in/. tradition I didn't RTFA, but what's the point of applying for patents in cases where you specifically know there's already prior art?
That's a rhetorical question - I don't even want to know the answer.
Yes, Obama raised - and spent - historic amounts of money, and this was well reported. I'm not sure why you think that Obama spending money on a campaign venue compares to the hypocrisy (or sheer spectacle, after she'd made herself into such an object of ridicule) of Palin wanting to present herself as a regular hockey mom then blowing $150,000 on clothes! What did you expect Obama to do with his campaign funds - NOT spend it on his campaign?!
The criticism of Obama deciding to change his mind on campaign funding is ridiculous. Once he realized that he could raise more money that way, would you really expect him (unless you were against him) to not do so beacuse it would be unfair to McCain's lack of comparable fundraising ability?! Would a decision to hobble his campaign, once it emerged there was a better alternative, indicate a sound executive decision to you?
One of the candidates (Obama) was a lot more newsworthy than the other, and the news coming out of both campaigns was decidedly different. You can't blame the press if MacCain campaign was all about a self-described "pitbull with lipstick" "hockey mom" with a $150K campaign wardrobe, a secessionist husband, and foreign policy experience that consisted of a geographical proximity to Russia. It's not the press making it up when McCain in rapid succession says he knows nothing about the economy, asserts that it is fundamentally sound, then suspends his campaign because he's so important in rescuing it (only to sit there silent in the meeting then return to his campaign). The press didn't make this stuff up - even the $150K clothes was somthing that was emphasized by disgruntled McCain insiders. The whole McCain campaign was about negativity - trying to shoot Obama down - while Obama's was much more positive - about change and hope and the future. You can't blame the press for reflecting the tone of the campaigns or reporting on their self-generated news (Joe the non-business-purchasing, non-plumber was even on the campaign trail with McCain), nor can you blame them for runnning more stories on the more newsworthy candidate. The press should be reporting on the news - they're not meant to be suppressing the differences and reporting both in equal column inches and in equally glowing terms ("Will Ameria elect historic first septuagenarian as president?", "Hitler sees bright future for germany!").
Slashdot Mirror
Good to know that about measels. I may have been confusing german measels (which we did have, along with chicken pox) with real measels. Obviously we didn't have MMR back then, but maybe we were vaccinated against M&M.
Interesting also to know that about DT(a)P. I don't know if my childhood doctor was an idiot, but at least he made house calls! ;-) As it happens I was later vaccinated against tetanus and diptheria as part of a school trip to the middle east.
Actually, given that one of the main symptons of autism is a lack of social and empathetic skills, I wouldn't totally discount that kids becoming less connected and spending so much time watching TV could conceivably have something to do with it. I would certainly anyway intend to limit the amount of TV my child watches, as most parents do.
Jenny McCarthy may not know what causes autism, but let's not forget that neither do doctors.
Tangentially related there was a documentary on TV last night about scientists discovering the mechanism of Gecko's adhesion to vertical surfaces (molecular Van de Waals forces). What was interesting - and hardly unique to this experiment - was that their initial experimental setup was unable to reproduce the adhesive effect leaving them scratching their heads, and it was only due to persistence and being sure that it was the experimental setup not the hypothesis that was wrong that caused them to persevere and eventually discover that you need to "back-up" the microscopic hair against a surface to get the "adhesive" effect.
The relevance? Maybe the experiments that have disproved any link between autism and vaccines are wrong, but scientists have abandoned them because they are so sure the connection is NOT there, rather than keeping at trying to prove it because of being sure that it IS the cause. One thing's for sure - I bet when the cuase is eventually discovered, there'll be some scientists somewhere having to admit they were wrong or at least surprised. In the meantime, don't trust a scientist who tells you that anything is impossible, because science can't prove impossibilities - it can only predict outcomes.
Since the cause of autism is unknown, but the incidence of it is up the last few decades, it seems quite likely that at least one of the assertions regarding what supposedly doesn't cause it may be wrong. Maybe the experiments that have "proved" MMR to be safe didn't reproduce the right conditions or test against the (unidentified) group most at risk of side effects... Maybe it's DTaP that's really the cause, maybe it's the cumulative effect of so many vaccines that does it (or maybe it is indeed unrelated to vaccines, and our source of ignorance lies elesewhere)..
In the meantime, until science has identified what is causing autism, it'd be better to have a bit more humility over what we don't know rather than asserting that for sure MMR/etc are safe. It may save you from a lawsuit to later be able to tell the parent of an autistic child ("but all the tests at the time said the vaccine was safe"), but that doesn't help the parent or the child at all. I'd say err on the side of caution and weight the risks. Don't just mindlessly vaccinate against everything possible, and use the most convenient multi-disease shots (MMR, DTaP).
I was born in the UK in 1961, and so grew up in the era where we weren't vaccinated against things like measels and chicken pox, and so of course we caught them, and we were fine. There may be rare side effects of these diseases, but the coincident rise in autism coupled with the rise in vaccination at least doesn't indicate autism as one of the side effects. As it happened I also almost died as an infant as a result of the DTP vaccine, and consequently wasn't given the 2nd shot of the series. I did subsequently catch whooping cough, and although it was unpleasant, it's sure better than being dead.
Since my wife is due with our first child in January, the topic of vaccinations has much personal relevance, and while I certainly intend to vaccinate our child in general, I will be looking for a pediatrician who isn't too blase about the potential risks given our lack of knowledge about what does autism.
First, imperative programming specifies the order of evaluation, whilst functional programming does not.
Yes, but there's nothing to stop you from building parallel processing libaries such as pthreads, OpenMP, CUDA or MapReduce using imperative languages - the semantics of your code are up to you. This seems to be a much more useful and practical way to take advantage of parallel hardware - keep the familiarity of the languages you know and then pick from a smorgasborg of parallel execution libraries as fits your needs. Relegating parallelism to libraries also has the advantage of freeing you from the limitations of the compiler... say that your Haskell compiler is able to generate "multi-threaded" code for multi-core processors, that's very nice, but what if you really want to spread your computation over multiple nodes in a cluster, or you want to use the power of your GPU-CPU video card to accelerate your application? Presumably you could write Haskell libraries to do this, but how is that much different from doing the same in C/C++ instead?
A crossbar switch in of itself isn't the solution to increased memory access bandwidth. What you need is to increase parallelism of memory access per CPU which means increasing the number of independent memory banks per CPU - one bank per N cores. You also then need (for global memory) to provide a way for each core group to access non-local memory memory banks (a shared bus would be sufficient).
e.g. something like this:
Cores 1 -4 = Memory bank 1 =+
=================+
Cores 5 -8 = Memory bank 2 =+
=================+
Cores 9-12 = Memory bank 3 =+
=================+
Cores 13-16 = Memory bank 4 =+
=================+
This type of memory/core architecture could easily map onto todays software. For example, have each process assigned to a single core group, such that pthread_create() assigns threads to cores only in that core group, and malloc() allocates memory only from the corresponding local memory bank. You'd presumably map part of the virtual address space onto the local memory bank and part onto the global address space (accessed via the shared bus/switch). The inter-process shared memory APIs (mmap, etc) would allocate memory from whereever they saw fit and then map it into the virtual address space accordingly.
If you wanted to allow a single process to create threads across multiple core groups (maybe controlled via processor affinity), then you could introduce something like pthread_malloc() to allocate local per-thread memory, have malloc() allocate per-process memory from the shared access space. You'd probably only want to have this cross-core-group thread allocation happen when explicity requested via the thread library so that existing binaries and recompiled applications would automatically only use the more efficient local memory, and only heavily threaded apps written to also use pthread_malloc() might ask to be spread across multiple core groups.
The idea here would be that each thread would use local memory as much as possible, thereby achieving maximum parallelism of memory access. Global memory access would result in less parallelism since you'd usually be competing for memory access with the local core group, as well as for global memory bus access. You could choose to implement a crossbar switch in place of the global memory bus, but that would be optional.
As both educators and researchers, we are amazed at the hype that the MapReduce proponents have spread about how it represents a paradigm shift in the development of scalable, data-intensive applications.
But MapReduce isn't about data-intensive applications... It's about scalable/massive primarily COMPUTE-intensive applications (which may or may not also be data-intensive - e.g. to compute the Madlebrot set you don't need any data at all - just massive compute power).
The problem MapReduce is solving is "My PC isn't fast enough to run my aplication, and I can't afford a supercomputer, so how can I make use of the aggregate COMPUTE (& I/O) speed of lots of cheap PCs to run my application faster?"
Writing parallel applications is hard, so what MapReduce does is provide one - limited, but yet quite widely applicable - easy to use way to parallelize compute intensive applications so that at run-time you can give them as many machines as you have available to run on and they'll automatically spread themselves over them (while incidently also providing for fault-tolerance and error recovery).
It's hardly any secret that CPU speed, even for single core processors, has been running ahead of memory bandwidth gains for years - that's why we have cache, and ever increasing amounts of it. It's also hardly any relevation to realize that if you're sharing your memory bandwidth between multiple cores then the bandwidth available per core is less than if you weren't sharing. Obviously you need to keep the amount of cache per core and the number of cores per machine (or, more precisely, per unit of memory sybsystem bandwidth) within reasonable bounds to keep it usable for general purpose aplications, else you'll end up in GPU-CPU (e.g. CUDA) territory where you're totally memory constrained and applicability is much less universal.
For cluster-based ("supercomputer") applications, partitioning between nodes is always going to be an issue in optimizing performance for a given architecture, and available memory bandwidth per node and per core is obviously a part of that equation. Moreover, even if CPU designers do add more cores per processor than is useful for some applications, no-one is forcing you to use them. The cost per CPU is going to remain approximately fixed, so extra cores per CPU essentially come for free. A library like pthreads, and different implementations of it (coroutine vs LWP based), gives you the flexibility over the mapping of threads to cores, and your overall across-node application partitioning gives you control over how much memory bandwidth per node you need.
Rather than dismissing Hawkins' HTM as just an ANN approach and moving on, I'd take a higher level view... what he's done is look at the undeniably regular, and perhaps surprisingly simple, structure of the neocortex, and try to figure out what the architecture is in abstract terms and what it's accomplishing. Of course it's true that in doing this he's making many simplifying assumptions, such as focusing on the Hebian function of neurons, but I think it's hard to argue that this is a reasonable starting point when one's goal is to understand the gross function of the connectivity pattern, and his results anyway rather justify the approach...
Hawkins' analysis of the cortex is that it's implementing an HTM, and this HTM level is the focus of both his book and of the Numeta software. It's only an ANN to the extent that the architecture is inspired by an ANN-like understanding of the cortical architecture.
In analyzing the cortex as an HTM, Hawkins has undeniably provided answers to many questions about how the cortex achieves things such as invarient representation, and marrying top down stored experience with bottom up sensory input to achieve perception. The Numenta SDK includes working examples of simple image recognition and speaker recognition. Does this view of the cortex as an HTM capture all it's nuances? Probably not, but I'm willing to bet that this is it's primary functionality and that further analysis will just add refinement to this model.
Of course the understanding of the neo-cortex as an HTM would fall totally flat if this didn't also explain cognition and intelligence, but you can't judge that until you've distilled the gazillion phenomenal manifestations of intelligence down to a minimalist abstract definition of what it actually is and therefore are in a position to attempt to bridge whatever divide you've left between mechanism (HTM) and fucntionality (intelligence). I can't recall how sucinctly Hawkins himself defines intelligence, or how well he explains it, but I can say that having thought about this myself - seriously - for a couple of decades, I would myself define intelligence in a single short sentence, and do indeed believe that an HTM in essence is all you need to achieve it.
If Hawkins has failed to explain things well enough to convince many people, then I think the failure is not of his architecture (not unique, but a useful summing up of ideas many others have had) but rather a failure of the audience to realize what it is that needs explaining! ;-)
Hawkin's isn't saying that complexity or computer power alone will get us to intelligence.
It's obvious that simulating a human brain, even at a high level of abstraction, will take a lot of compute power, memory, etc.
If you take, say, a neural net of a worm's brain and speed it up by a factor of a few million all you will have is a very fast worm.
However, if you take a architecture that is representative of our own brain (but say, with a cortex only 1/1,000,000 the size of ours, and only overall running at 1/1,000,000,000 of the aggregate speed), and you speed/scale THAT up accordingly, then you will in fact have something that has human like intelligence.
You are saying, and are claiming that Hawkin's is saying, that complexity/scaling won't lead to intelligence, but you are wrong. Scaling the RIGHT architecture will lead to it, scaling the wrong architecture will lead to super fast worms, etc.
Thanks for the reference - I just finished reading the Amazon reviews and ordered the book!
It sounds as if you're describing a feed-forward network. Things get much more interesting once you bring feedback paths into the picture. Try googling "Adaptive Resonance Theory (ART)" for one particular architecture, or consider your own grey noodle as the ultimate proof of concept of the power of neural nets!
The term AI is so nebulous that it doesn't really mean much of anything. It's more of a functional goal (computer-based human-like ability) than anything more concrete, and as anything that may fall under that general umbrella does become better understood or more concrete, then it tends to be no longer regarded as part of AI (e.g. machine learning, expert systems, speech recognition).
It's also worth noting that natural intelligence is also a rather nebulous concept - you'll find many definitions offered (e.e. an ability to learn and generalize), but almost any definition is doing to be met by some people with "well, that's PART of it...".
The larger problem is that the goal of AI really isn't artificial intelligence(!) - it's an artificial brain/mind. The emphasis on intelligence/cognition is perhaps why so many attempts to achieve AI have failed - because there's no theory of mind or overall brain architecture behind them. Artificial cognition itself is anyways arguably an already solved problem via general symbolic problem solvers like Allen Newell's SOAR, but the ability to manipulate and further refine knowledge isn't of much use if you can't aquire it in the first place. Pairing a cognition capability with an artificially aquired body of real-world knowledge, as done by Cyc, is no better because it is brittle and inflexible - an artifical brain needs to be able to distill it's own ever-changing categories and knowledge out of raw analog sensory input, and this means that symbolic approaches (fuzzy or not) are at best only a small part of the solution (more likley no part at all).
What you really need to do before getting into any technologies that may help implement "AI" is to define what it is that you want to implement. If you want to create an artificial brain, then you will need to figure out what a brain is and how to decompose it into functional blocks - an architecture. Biology texts may be of more use than "AI" ones, as even AI luminaries such as Marvin Minsky have produced laughably simple "theories of the mind". Hawkin's "On Intelligence" is a better start than many, but he also is really focusing on the problem of intelligence, but at least addressing the issue of how to mate this to the perceptual system.
It's only once you know what you want to build (i.e. the hard part of figuring out how the brain works) that the issue of how to implement it comes up, and what existing technologies may or may not be of use (Neural nets, HTMs, SVMs, etc). If you start to implement before having a fairly complete theory of the brain/mind, then your results will only be as good as your goals.
Sure most people will run whatever OS their new computer comes with (and there's not usually a choice), and one side effect of that is that Vista must be reducing sales of new hardware. Case in point is myself... I was casually browsing for black friday laptop deals to replace my wife's aging Toshiba Satellite (400MHz Celeron - still just about tolerable for web browsing), but was put off by the fact that all the ones I say came with Vista (and no mention of XP downgrade option, so who knows if they have XP drivers available).
Sure I'll eventually have to replace that laptop, and probably run whatever it comes with, but for the time being the fact that the new ones come with Vista didn't make me become a Vista user - it meant that I chose not to purchase a new laptop!
Maybe with a tanking economy and slowed sales, the hardware manufacturers will put pressure on Microsoft to fully revive Windows XP - rather like Coke reviving "Coke Classic" after the "New Coke" debacle.
Right now it may only make the less complex games playable (maybe also enabling the full Vista aero UI without a graphics card), but given the pace of processor speed improvement that will change. Double the CPU clock and the FPS doubles, double or quadruple the number of cores and the FPS again goes up by the same factor...
I don't think it's be too long before most games have playable frame rates using this software renderer, even if not at the highest resolution, or with the highest quality settings enabled, but it'll be a process of continual improvement. In the meantime even if the most demanding games arn't playable under this, there's much other less demanding software for which it'll be immediately useful, and given that even the most-stripped down budget machines still have todays insanely fast processors (even if not the bleeding edge), it brings an ever increasing level of capability to the lowest level.
The curriculum may have gotten dumber, and the assessment bar set lower, but even without those additional factors the abandonment of the split grammar school vs comprehensive (& corresp. "O" level" vs CSE exams) has to be a large part of this.
Under the old system (I went to grammar school) the academically faster kids went to grammar schoool and the slower / less able ones went to secondary modern. The purpose of this system was to let each group advance at it's own pace, rather than having to gear the curriculum and pace to the average or less able students as would be needed without this seperation.
Of course eventually the politically correct crown got their way and it was deemed unfair to seperate kids based on ability, so now the brighter kids are doomed to the GCSE curriculum/pace of everyone else. If you're not even TRYING to educate the brighter kids to the best of their ability, is it at all surprising, or even newsworthy, that their level of education/ability is lower than when you were trying to?!
Of course this whole dumbing down appears to have gone further than the abolishment of ability streaming, and the usual response to slipping exam grades seems to be lowering the bar rather than increasing the quality of education.
From NVidia's CUDA site, most of their regular display cards support CUDA, just with less cores (hence less performance) than the Tesla card. The cores that CUDA uses are what used to be called the vertex shaders on your (NVidia) card. The CUDA API is designed so that your code doesn't know/specify how many cores are going to be used - you just code to the CUDA architecture and at runtime it distrubutes the workload to the available cores... so you can develop for a low end card (or they even have an emulator) then later pay for th hardware/performance you need.
Has anyone ever tried to do this - to clone a modern species using the egg of a different, but closely related, species?
I guess it's not too different to what Craig Venter is doing with his minimal bacteria - putting synthetic DNA into the cell of the original from which it was derived (by selectivey removing about 20% of its DNA). No word on whether he's actually bought the thing to life yet though.
Einstein has been cloned from a skidmark in a pair of his underpants kept by his family.
A skull-bong found in an MIT dormitary has been DNA identified as belonging to Isaac Newton. The DNA came from fingerprints on the bong, confirming rumors of Newton being a pothead.
A frozen turd found in Craig Venter's kitchen fridge, formally believed to have come from a pygmy marmoset, has been identified and carbon dated as the last movement of Wolfgang Armadeus Mozart.
OK, so I lied. Mozart wasn't a scientist.
If you believe in mind/body dualism then obviously you can't equate anything in the material domain with anything in the "spiritual" domain. Kurzweil isn't a moron so it's safe to assume he's not making that assertion.
The question is only meaningful (vs trivially false) if one assumes that by "soul" one means some NON-dualistic phenomena, so Kurzweil is simply choosing to assume that is what the questioner really means and answering it accordingly, even definining consciousness before doing so. You could hardly ask for a clearer or more concise answer to the question!
Windows CE already runs on ARM, and I bet most of the people using these handhelds are barely even aware they are ARM based.
An ARM based netbook that ran either Ubunto or Win CE (vs Win XP) would be interesting.
I wouldn't be so sure about that - an ARM based netbook running Ubuntu (ow Windows CE) might make sense.
Well, IANAL, and in /. tradition I didn't RTFA, but what's the point of applying for patents in cases where you specifically know there's already prior art?
That's a rhetorical question - I don't even want to know the answer.
Yes, Obama raised - and spent - historic amounts of money, and this was well reported. I'm not sure why you think that Obama spending money on a campaign venue compares to the hypocrisy (or sheer spectacle, after she'd made herself into such an object of ridicule) of Palin wanting to present herself as a regular hockey mom then blowing $150,000 on clothes! What did you expect Obama to do with his campaign funds - NOT spend it on his campaign?!
The criticism of Obama deciding to change his mind on campaign funding is ridiculous. Once he realized that he could raise more money that way, would you really expect him (unless you were against him) to not do so beacuse it would be unfair to McCain's lack of comparable fundraising ability?! Would a decision to hobble his campaign, once it emerged there was a better alternative, indicate a sound executive decision to you?
One of the candidates (Obama) was a lot more newsworthy than the other, and the news coming out of both campaigns was decidedly different. You can't blame the press if MacCain campaign was all about a self-described "pitbull with lipstick" "hockey mom" with a $150K campaign wardrobe, a secessionist husband, and foreign policy experience that consisted of a geographical proximity to Russia. It's not the press making it up when McCain in rapid succession says he knows nothing about the economy, asserts that it is fundamentally sound, then suspends his campaign because he's so important in rescuing it (only to sit there silent in the meeting then return to his campaign). The press didn't make this stuff up - even the $150K clothes was somthing that was emphasized by disgruntled McCain insiders. The whole McCain campaign was about negativity - trying to shoot Obama down - while Obama's was much more positive - about change and hope and the future. You can't blame the press for reflecting the tone of the campaigns or reporting on their self-generated news (Joe the non-business-purchasing, non-plumber was even on the campaign trail with McCain), nor can you blame them for runnning more stories on the more newsworthy candidate. The press should be reporting on the news - they're not meant to be suppressing the differences and reporting both in equal column inches and in equally glowing terms ("Will Ameria elect historic first septuagenarian as president?", "Hitler sees bright future for germany!").