Re: [fonc] IBM eyes brain-like computing
On 10/29/2011 6:46 AM, karl ramberg wrote: On Sat, Oct 29, 2011 at 5:06 AM, BGBcr88...@gmail.com wrote: On 10/28/2011 2:27 PM, karl ramberg wrote: On Fri, Oct 28, 2011 at 6:36 PM, BGBcr88...@gmail.comwrote: On 10/28/2011 7:28 AM, K. K. Subramaniam wrote: On Thursday 27 Oct 2011 11:27:39 PM BGB wrote: most likely, processing power will stop increasing (WRT density and/or watts) once the respective physical limits are met (basically, it would no longer be possible to get more processing power in the same space or using less power within the confines of the laws of physics). The adoption of computing machines at large is driven primarily by three needs - power (portable), space/weight and speed. The last two are now solvable in the large but the third one is still stuck in the dark ages. I recollect a joke by Dr An Wang (founder of Wang Labs) in keynote during the 80s that goes something like this: A man struggled to lug two heavy suitcases into a bogie in a train that was just about to depart. A fellow passenger helped him in and they start a conversation. The man turns out to be a salesman from a company that made portable computers. He showed one that fit in a pocket to his fellow passenger. It does everything that a mainframe does and more and it costs only $100. Amazing! exclaimed the passenger as he held the marvel in his hands, Where can I get one?. You can have this piece, said the gracious gent, as thank you gift for helping me. Thank you very much. the passenger was thrilled beyond words as he gingerly explored the new gadget. Soon, the train reached the next station and the salesman stepped out. As the train departed, the passenger yelled at him. Hey! you forgot your suitcases!. Not really! the gent shouted back. Those are the batteries for your computer. ;-) .. Subbu yeah... this is probably a major issue at this point with hugely multi-core processors: if built, they would likely use lots of power and produce lots of heat. this is sort of also an issue with video cards, one gets a new/fancy nVidia card, which is then noted to have a few issues: it takes up two card slots (much of this apparently its heat-sink); it is long enough that it partially sticks into the hard-drive bays; it requires a 500W power supply; it requires 4 plugs from the power-supply; ... so, then one can joke that they have essentially installed a brick into their computer. nevermind it getting high framerates in games... however, they would have an advantage as well: people can still write their software in good old C/C++/Java/... it is likely that the existence of existing programming languages and methodologies will continue to be necessary of new computing technologies. also, likewise people will continue pushing to gradually drive-down the memory requirements, but for the most part the power use of devices has been largely dictated by what one can get from plugging a power-cord into the wall (vs either running off batteries, or OTOH, requiring one to plug in a 240V dryer/arc-welder/... style power cord). elsewhere, I designed a hypothetical ISA, partly combining ideas from ARM and x86-64, with a few unique ways of representing instructions (the idea being that they are aligned values of 1/2/4/8 bytes, rather than either more free-form byte-patterns or fixed-width instruction-words). or such... ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc This is also relevant regarding understanding how to make these computers work: http://www.infoq.com/presentations/We-Really-Dont-Know-How-To-Compute seems interesting, but is very much a pain trying to watch as my internet is slow and the player doesn't really seem to buffer up the video all that far when paused... but, yeah, eval and reflection are features I really like, although sadly one doesn't really have much of anything like this standard in C, meaning one has to put a lot of effort into making a lot of scripting and VM technology primarily simply to make up for the lack of things like 'eval' and 'apply'. this becomes at times a point of contention with many C++ developers, where they often believe that the greatness of C++ for everything more than makes up for its lack of reflection or dynamic features, and I hold that plain C has a lot of merit if-anything because it is more readily amendable to dynamic features (which can plug into the language from outside), which more or less makes up for the lack of syntax sugar in many areas... The notion I get from this presentation is that he is against C and static languages in general. It seems lambda calculus derived languages that are very dynamic and can self generate code is the way he thinks the exploration should take. I was not that far into the video at the point I posted, due mostly to slow internet, and the player not allowing the pause, let it buffer, and come back later strategy, generally needed for things
Re: [fonc] IBM eyes brain-like computing
On 10/28/2011 7:28 AM, K. K. Subramaniam wrote: On Thursday 27 Oct 2011 11:27:39 PM BGB wrote: most likely, processing power will stop increasing (WRT density and/or watts) once the respective physical limits are met (basically, it would no longer be possible to get more processing power in the same space or using less power within the confines of the laws of physics). The adoption of computing machines at large is driven primarily by three needs - power (portable), space/weight and speed. The last two are now solvable in the large but the third one is still stuck in the dark ages. I recollect a joke by Dr An Wang (founder of Wang Labs) in keynote during the 80s that goes something like this: A man struggled to lug two heavy suitcases into a bogie in a train that was just about to depart. A fellow passenger helped him in and they start a conversation. The man turns out to be a salesman from a company that made portable computers. He showed one that fit in a pocket to his fellow passenger. It does everything that a mainframe does and more and it costs only $100. Amazing! exclaimed the passenger as he held the marvel in his hands, Where can I get one?. You can have this piece, said the gracious gent, as thank you gift for helping me. Thank you very much. the passenger was thrilled beyond words as he gingerly explored the new gadget. Soon, the train reached the next station and the salesman stepped out. As the train departed, the passenger yelled at him. Hey! you forgot your suitcases!. Not really! the gent shouted back. Those are the batteries for your computer. ;-) .. Subbu yeah... this is probably a major issue at this point with hugely multi-core processors: if built, they would likely use lots of power and produce lots of heat. this is sort of also an issue with video cards, one gets a new/fancy nVidia card, which is then noted to have a few issues: it takes up two card slots (much of this apparently its heat-sink); it is long enough that it partially sticks into the hard-drive bays; it requires a 500W power supply; it requires 4 plugs from the power-supply; ... so, then one can joke that they have essentially installed a brick into their computer. nevermind it getting high framerates in games... however, they would have an advantage as well: people can still write their software in good old C/C++/Java/... it is likely that the existence of existing programming languages and methodologies will continue to be necessary of new computing technologies. also, likewise people will continue pushing to gradually drive-down the memory requirements, but for the most part the power use of devices has been largely dictated by what one can get from plugging a power-cord into the wall (vs either running off batteries, or OTOH, requiring one to plug in a 240V dryer/arc-welder/... style power cord). elsewhere, I designed a hypothetical ISA, partly combining ideas from ARM and x86-64, with a few unique ways of representing instructions (the idea being that they are aligned values of 1/2/4/8 bytes, rather than either more free-form byte-patterns or fixed-width instruction-words). or such... ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
On 10/28/2011 2:27 PM, karl ramberg wrote: On Fri, Oct 28, 2011 at 6:36 PM, BGBcr88...@gmail.com wrote: On 10/28/2011 7:28 AM, K. K. Subramaniam wrote: On Thursday 27 Oct 2011 11:27:39 PM BGB wrote: most likely, processing power will stop increasing (WRT density and/or watts) once the respective physical limits are met (basically, it would no longer be possible to get more processing power in the same space or using less power within the confines of the laws of physics). The adoption of computing machines at large is driven primarily by three needs - power (portable), space/weight and speed. The last two are now solvable in the large but the third one is still stuck in the dark ages. I recollect a joke by Dr An Wang (founder of Wang Labs) in keynote during the 80s that goes something like this: A man struggled to lug two heavy suitcases into a bogie in a train that was just about to depart. A fellow passenger helped him in and they start a conversation. The man turns out to be a salesman from a company that made portable computers. He showed one that fit in a pocket to his fellow passenger. It does everything that a mainframe does and more and it costs only $100. Amazing! exclaimed the passenger as he held the marvel in his hands, Where can I get one?. You can have this piece, said the gracious gent, as thank you gift for helping me. Thank you very much. the passenger was thrilled beyond words as he gingerly explored the new gadget. Soon, the train reached the next station and the salesman stepped out. As the train departed, the passenger yelled at him. Hey! you forgot your suitcases!. Not really! the gent shouted back. Those are the batteries for your computer. ;-) .. Subbu yeah... this is probably a major issue at this point with hugely multi-core processors: if built, they would likely use lots of power and produce lots of heat. this is sort of also an issue with video cards, one gets a new/fancy nVidia card, which is then noted to have a few issues: it takes up two card slots (much of this apparently its heat-sink); it is long enough that it partially sticks into the hard-drive bays; it requires a 500W power supply; it requires 4 plugs from the power-supply; ... so, then one can joke that they have essentially installed a brick into their computer. nevermind it getting high framerates in games... however, they would have an advantage as well: people can still write their software in good old C/C++/Java/... it is likely that the existence of existing programming languages and methodologies will continue to be necessary of new computing technologies. also, likewise people will continue pushing to gradually drive-down the memory requirements, but for the most part the power use of devices has been largely dictated by what one can get from plugging a power-cord into the wall (vs either running off batteries, or OTOH, requiring one to plug in a 240V dryer/arc-welder/... style power cord). elsewhere, I designed a hypothetical ISA, partly combining ideas from ARM and x86-64, with a few unique ways of representing instructions (the idea being that they are aligned values of 1/2/4/8 bytes, rather than either more free-form byte-patterns or fixed-width instruction-words). or such... ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc This is also relevant regarding understanding how to make these computers work: http://www.infoq.com/presentations/We-Really-Dont-Know-How-To-Compute seems interesting, but is very much a pain trying to watch as my internet is slow and the player doesn't really seem to buffer up the video all that far when paused... but, yeah, eval and reflection are features I really like, although sadly one doesn't really have much of anything like this standard in C, meaning one has to put a lot of effort into making a lot of scripting and VM technology primarily simply to make up for the lack of things like 'eval' and 'apply'. this becomes at times a point of contention with many C++ developers, where they often believe that the greatness of C++ for everything more than makes up for its lack of reflection or dynamic features, and I hold that plain C has a lot of merit if-anything because it is more readily amendable to dynamic features (which can plug into the language from outside), which more or less makes up for the lack of syntax sugar in many areas... although, granted, in my case, the language I eval is BGBScript and not C, but in many cases they are similar enough that the difference can be glossed over. I had considered, but never got around to, creating a language I was calling C-Aux, which would have taken this further, being cosmetically similar to and mostly (85-95% ?) source-compatible with C, but being far more dynamic (being designed to more readily allow quickly loading code from source, supporting eval, ...). essentially, in a practical sense C-Aux would
Re: [fonc] IBM eyes brain-like computing
On Wed, Oct 26, 2011 at 12:53:24AM -0700, BGB wrote: from what I read, IBM was using a digital crossbar. It sounds like Kwabena Boahen (Carver Mead's school) is on the right track http://web.cecs.pdx.edu/~strom/onr_workshop/boahen.pdf the group seems to be still publishing http://www.stanford.edu/group/brainsinsilicon/pubs.html I suspect something more generic would be needed. I don't see how generic will do long-term any than for bootstrap (above co-evolution) reasons. more generic is more likely to be able to do something interesting. More generic hardware (not optimized for a particular model) also means it's less efficient. On the other hand, we don't have a particular model to optimize for, so right now generic is the way to go. ARM or similar could work (as could a large number of 386-like cores). I had considered something like GPGPU, but depending on the type of neural-net, there could be issues with mapping it efficiently to existing GPUs. You could wire it up in a 3d grid, with offsets to mostly local other grid sites and stream through memory. Refresh rate could be 10 Hz or more, depending on how complex the model, and how much random-access like memory accesses you're producing. also, the strong-areas for GPUs are not necessarily the same as the requirements for implementing neural nets. again, it could work, but it If you can do plenty of parallel ops on short integers (8-16 bit) then it seems to match well, provided you can feed the shaders. Mapping things to memory is tricky, as otherwise you'll starve the shaders and thus underutilize the hardware. is just less certain it is ideal. the second part of the question is: assuming one can transition to a purely biology-like model, is this a good tradeoff?... if one gets rid of a few of the limitations of computers but gains some of the limitations of biology, this may not be an ideal solution. Biology had never had the issue to deal with high-performance numerics, I'm sure if it had it wouldn't do too shabbily. You can always go hybrid e.g. if you want to do proofs or cryptography. biology also doesn't have software distribution, ability to make backups, ... Even hybrid analog systems can be converted into a large blob of binary data and loaded back. The loss of precision due to digitization is negligible, as the system will be noisy/low precision (6 bit) to start with. ideally, a silicon neural-net strategy would also readily allow the installation of new software and creation of backups. You can always digitize and serialize your data, put it through a pipe and reinstantialize it in equivalent piece of hardware on the other end. Consider crystalline computation (Margolus/Toffoli) http://people.csail.mit.edu/nhm/cc.pdf which maps very well to 3d lattice of sites in future's molecular hardware. If you can halt the execution, or clone a static shadow copy of the dynamic process you can serialize and extract from the faces of the crystal at leisure. the most likely strategy here IMO is to leverage what existing OS's can already do, essentially treating any neural-net processors as another piece of hardware as far as the OS is concerned. I'm not sure you need an OS. Many aspects of operation can be mapped to hardware. Consider a 10 GBit/s short reach fiber, the length of fiber through these keystrokes are passing is an optical FIFO conveniently containing a standard MTU (1500 Bytes) frame/packet. Same thing does vacuum to a sufficiently fast line of sight laser link. With the right header layout, you can see how you could make a purely photonic cut-through router or switch, with zero software. Same thing with operating Toffoli's hypothetical computronium crystal, you have stop and go, read and write, and that's it. I/O could be mapped directly to crystal faces. this would probably mean using neural-net processors along-side traditional CPU cores (rather than in place of them). better would be to try for a strategy where the merits of both can be gained, and as many limitations as possible can be avoided. most likely, this would be via a hybrid model. Absolutely. Hybrid at many scales, down to analog computation for neurons. yeah. analog is an idea I had not personally considered. I guess a mystery here is how effectively semiconductor logic (in integrated circuits) can work with analog signals. the main alternative is, of course, 8 or 16-bit digital signals. 8 or 16 bit ALUs are ridiculously complicated, power-hungry and slow if compared with analog computation in transistors or memristor. Of course the precision is low, particularly if you're working at nanoscale. A few atoms displaced shift the parameters of the device visibly, but then, you're engaging a redundant mode of computation anyway. The brain does not just tolerate noise, some aspects of it are noise-driven. I had more imagined as hybrids of neural-nets and traditional software. granted,
Re: [fonc] IBM eyes brain-like computing
On Thu, Oct 27, 2011 at 1:10 PM, Steve Dekorte st...@dekorte.com wrote: BGB cr88...@gmail.com wrote: Leitl wrote: John Zabroski wrote: Kurzweil addresses that. As far as I know Kurzweil hasn't presented anything technical or even detailed. Armwaving is cheap enough. yep, one can follow a polynomial curve to pretty much anything... actually getting there is another matter. I wonder what the curve from the early Roman civilization looked liked and how that compared to the actual data from the Dark Ages. Good point! Folks in the Dark Ages still built the Chartres Cathedral, while Rome was forced to build the Parthenon. ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
On Oct 27, 2011, at 1:32 PM, John Zabroski johnzabro...@gmail.com wrote: Steve Dekorte wrote: I wonder what the curve from the early Roman civilization looked liked and how that compared to the actual data from the Dark Ages. Good point! Folks in the Dark Ages still built the Chartres Cathedral, while Rome was forced to build the Parthenon. After a millennium, yes. Does kurzweil's theory allow for millennium wide technology depressions?___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
On Thu, Oct 27, 2011 at 1:32 PM, John Zabroski johnzabro...@gmail.comwrote: On Thu, Oct 27, 2011 at 1:10 PM, Steve Dekorte st...@dekorte.com wrote: BGB cr88...@gmail.com wrote: Leitl wrote: John Zabroski wrote: Kurzweil addresses that. As far as I know Kurzweil hasn't presented anything technical or even detailed. Armwaving is cheap enough. yep, one can follow a polynomial curve to pretty much anything... actually getting there is another matter. I wonder what the curve from the early Roman civilization looked liked and how that compared to the actual data from the Dark Ages. Good point! Folks in the Dark Ages still built the Chartres Cathedral, while Rome was forced to build the Parthenon. Oops, I meant Athens was forced to build the Parthenon. (At the time, many in Athens thought the Parthenon was disgusting due to its extra columns and other features required for its abnormal size.) ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
On 10/27/2011 10:10 AM, Steve Dekorte wrote: BGBcr88...@gmail.com wrote: Leitl wrote: John Zabroski wrote: Kurzweil addresses that. As far as I know Kurzweil hasn't presented anything technical or even detailed. Armwaving is cheap enough. yep, one can follow a polynomial curve to pretty much anything... actually getting there is another matter. I wonder what the curve from the early Roman civilization looked liked and how that compared to the actual data from the Dark Ages. probably: sharp rise... plateau... collapse... dark ages then begin. a lot was forgotten for a while, but then in the following centuries much of what was lost was recovered, and then the original roman empire was surpassed. now, things are rising at the moment, and may either: continue indefinitely; hit a plateau and stabilize; hit a plateau and then follow a downward trend. most likely, processing power will stop increasing (WRT density and/or watts) once the respective physical limits are met (basically, it would no longer be possible to get more processing power in the same space or using less power within the confines of the laws of physics). granted, I suspect there may still be a ways to go (it is possible that such a computer might not even necessarily be matter as currently understood). then again, the limits of what is practical could come a bit sooner. a fairly conservative estimate would be if people hit the limits of what could be practically done with silicon, and then called it good enough. otherwise, people may migrate to other possibilities, such as graphene or photonics, or maybe build anyonic systems, or similar. ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
John Zabroski wrote: Steve Dekorte wrote: After a millennium, yes. Does kurzweil's theory allow for millennium wide technology depressions? Interesting question! What technology depressions can you think of and are referring to? Basically, government intervention crushing the economy. There are no shortage of examples. One is how east Germany was effectively frozen in time under communism. Another is the effective destruction of technological civilization in Rome and Egypt: http://www.dekorte.com/blog/blog.cgi?do=itemid=4571 ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
On Thu, Oct 27, 2011 at 8:35 PM, David Goehrig d...@nexttolast.com wrote: probably: sharp rise... plateau... collapse... dark ages then begin. As probably the only Late Ancient / Early Medievalist on this list, I feel a need to correct this myth of the Dark Ages (which can be squarely blamed on Edward Gibbon, and his personal issues with organized religion). As we managed to work beyond a certain cultural bias brought on by Imperialistic 19th century powers manipulating our perspective of the Roman world for political gain, most historians who now study this era see it as an incredibly vibrant period of political, technological, and cognitive change. Most languages spoken in Europe today are a direct result of a massive growth in technical language developed by people who married Classical thought with new Germanic and Asiatic influences. Critical mathematical advances occurred laying the groundwork for what would become symbolic logic and algebra. If you focus on the then more populist and wealthy east, there is a straight continuity. In the west, there is actually pretty radical change which gave birth to the political structures that created the modern era (which Classicists view as everything after 1066). While outside of Ireland, nearly all knowledge of Greek was lost, those concepts were translated into vulgate resulting in a vast democratization of thought. (seeds of the reformation) From a pure info technology standpoint, there is no plateau, merely a paradigm shift which enabled new sources of intellectual growth. Just like we saw with the advent of digital computing. My friend who is full of weird facts tells me, and I quote: It's not that [the Dark Ages] were dismal and miserable, it's that a lot of the rest of the word (especially the Middle East) surpassed Europe very quickly in a lot of areas including philosophy, medicine, and mathematics. Avicenna alone probably exceeded the combined output of Europe for a few hundred years... ...Dark Ages really reflects a very culturally Eurocentric view. ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
On 10/27/2011 5:35 PM, David Goehrig wrote: probably: sharp rise... plateau... collapse... dark ages then begin. As probably the only Late Ancient / Early Medievalist on this list, I feel a need to correct this myth of the Dark Ages (which can be squarely blamed on Edward Gibbon, and his personal issues with organized religion). As we managed to work beyond a certain cultural bias brought on by Imperialistic 19th century powers manipulating our perspective of the Roman world for political gain, most historians who now study this era see it as an incredibly vibrant period of political, technological, and cognitive change. Most languages spoken in Europe today are a direct result of a massive growth in technical language developed by people who married Classical thought with new Germanic and Asiatic influences. Critical mathematical advances occurred laying the groundwork for what would become symbolic logic and algebra. If you focus on the then more populist and wealthy east, there is a straight continuity. In the west, there is actually pretty radical change which gave birth to the political structures that created the modern era (which Classicists view as everything after 1066). While outside of Ireland, nearly all knowledge of Greek was lost, those concepts were translated into vulgate resulting in a vast democratization of thought. (seeds of the reformation) dunno, I think the usual idea (not necessarily accurate, but common perception) was that, following the fall of Rome, was a time where everything generally sucked, where disease was everywhere, where people almost invariably had mud on their clothes and faces, where sewage rained from buildings, ... then there was philosophers, founding for-fathers, and industrialization and so on, and then all of the stuff going on in the 20th century, and then one is more-or-less at the present (which was generally much better, apart from WWII and the 60s, where the 60s was sort of a more recent dark-age filled with hippies and similar...). I guess elsewhere (back in the dark ages) people were building cathedrals, and Muslims were off doing philosophy and developing Algebra and similar (before the crusaders went and brought a bunch of stuff back to Europe). but, in Rome there was concrete and some instances of reinforced concrete, which didn't come back into being until recently. also apparently (on a documentary I saw involving Terry Jones) they (more or less) had hamburgers in Rome, but these were not rediscovered until much later... (but apparently they didn't have french-fries or soda, or use cheese slices, making the modern form potentially more advanced...). probably in any case, people had surpassed Rome probably by the time industry was coming around, since AFAIK Rome had never industrialized (it is a question if Rome could have industrialized had it not collapsed first... say could the industrial revolution happened 1000 years earlier?...). just in my personal perception, the modern era probably began sometime between 1985 and 1995 (or, somewhere between MacOS and Win95), or maybe with the release of Quake in 1996. or, at least, this is when the world I as know it more or less took its current form (but I am left feeling old, as I am old enough to remember the end of the era that was the early 90s, of the gradual death of MS-DOS and 5.25 inch floppies). then a few times I have been left thinking about how terrible the modern times are, but then I am left to think about the 1960s and 70s and left to think they probably would have been much worse (or, at least, a strange-looking world I can't particularly relate to, and filled with people with overly promiscuous lifestyles and using lots of drugs and similar...). granted, others who have lived through these decades might decide to disagree with me, which is fair enough. also, apparently, it being younger people who watched Jeopardy and Wheel of Fortune, rather than these being primarily the domain of older people, with the people having always having watched these shows, rather than the show preference having been a result of the aging process... also, people who were not old-looking back in the 90s (on TV and similar) being much older looking now (and often born back in the 60s). it is almost surreal sometimes. and similar... From a pure info technology standpoint, there is no plateau, merely a paradigm shift which enabled new sources of intellectual growth. Just like we saw with the advent of digital computing. this is an interesting perspective... I had not heard this position claimed before (almost always, it was the Romans were smart, but collapsed, and there was a period of intense suck until either 'the Renaissance' or the 'Age of Enlightenment'..., or according to others, the Protestant Reformation, since the idea is that the Catholics have fallen into a state of heresy or apostasy). however, granted,
Re: [fonc] IBM eyes brain-like computing
On Tue, Oct 25, 2011 at 10:17:24AM -0700, BGB wrote: I was not arguing about the limits of computing, rather, IBM's specific design. it doesn't really realistically emulate real neurons, rather it is a Real neurons have many features, many of them unknown, and do not map well into solid state as is. However, you can probably find a simplified model which is powerful and generic enough and maps well to solid state by co-evolving substrate and representation. from what I can gather a simplistic accumulate and fire model, with the neurons hard-wired into a grid. In principle you can use a hybrid model by using a crossbar for local connectivity which is analog, and a packet-switched signalling mesh for long-range interactions, similiarly as real neurons do it. The mesh can emulate total connectivity fine, and you can probably even finagle something which scales better than a crossbar locally. I suspect something more generic would be needed. I don't see how generic will do long-term any than for bootstrap (above co-evolution) reasons. another question is what can be done in the near term and on present hardware (future hardware may or may not exist, but any new hardware may take years to make it into typical end-user systems). Boxes with large number of ARM SoCs with embedded memory and signalling mesh have been sighted, arguably this is the way to go for large-scale. GPGPU approaches are also quite good, if you map your neurons to a 3d array and stream through memory sequentially. Exchanging interface state with adjacent nodes (which can be even on GBit Ethernet) is cheap enough. the second part of the question is: assuming one can transition to a purely biology-like model, is this a good tradeoff?... if one gets rid of a few of the limitations of computers but gains some of the limitations of biology, this may not be an ideal solution. Biology had never had the issue to deal with high-performance numerics, I'm sure if it had it wouldn't do too shabbily. You can always go hybrid e.g. if you want to do proofs or cryptography. better would be to try for a strategy where the merits of both can be gained, and as many limitations as possible can be avoided. most likely, this would be via a hybrid model. Absolutely. Hybrid at many scales, down to analog computation for neurons. or such... -- Eugen* Leitl a href=http://leitl.org;leitl/a http://leitl.org __ ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org 8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
Eugen Leitl wrote: On Wed, Oct 26, 2011 at 09:00:36AM -0400, John Zabroski wrote: Kurzweil addresses that. As far as I know Kurzweil hasn't presented anything technical or even detailed. Armwaving is cheap enough. Kurzweil addresses that. Do you have literature references for that? As for biology, it is an iterative approach. The biggest insights are coming from better medical imaging techniques that allow us to see inside living systems' organs and better understand how they work. Kurzweil sort of discusses this too, and relates it to how it helped him develop better and more scalable algorithms even when the underlying hardware did not change. When was the last time Kurzweil did design something? 1990? Prior to that? It looks like Kurzweil is mainly using an outside view perspective when making his predictions. Moore's law for instance comes from such a perspective: we observe a past trend (number of transistors doubling every 18 months), and conclude that this trend will very likely go on for a while. (Note that we use similar methods for the laws of physics: we let an apple fall a millions times, and we predict that it will fall again if we try one more time.) Yet Moore's law doesn't by itself increase the power of our computers. Researchers do, and their findings don't exactly come out of thin air. Yet we don't need to know how they work to be able to derive accurate predictions, such as Moore's law. Now we could debate the validity of the outside view when making predictions for the next few decades. I for one am not as confident as Kurzweil seems to be about the regularity of the exponential growth in technology, and what it tells us about a technological singularity. However, the outside view doesn't require its wielder to have as much expertise as the experts in the field. For instance,One doesn't need to have designed anything to notice that things are regularly being designed. Does Kurzweil have relevant expertise? Did he demonstrate understanding of the subjects he talks about? Did he earned authority? As long as he is using an outside view, those questions don't matter much. What matters is whether the outside view is a valid method or not, and how accurately does Kurzweil uses it (we could meta-recursively apply the outside view on Kurzweil predictions: look at his past predictions to predict the accuracy (and bias) of his present ones). Of course, if Kurzweil does use the inside view, then his expertise suddenly becomes much more relevant. Loup. ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
On 10/26/2011 6:06 AM, Eugen Leitl wrote: On Wed, Oct 26, 2011 at 09:00:36AM -0400, John Zabroski wrote: Kurzweil addresses that. As far as I know Kurzweil hasn't presented anything technical or even detailed. Armwaving is cheap enough. yep, one can follow a polynomial curve to pretty much anything... actually getting there is another matter. Kurzweil addresses that. Do you have literature references for that? especially considering the context: another question is what can be done in the near term and on present hardware (future hardware may or may not exist, but any new hardware may take years to make it into typical end-user systems). if it can't be done on present HW, it can't be done on present HW. much like I can't render a whole damn city-scape down to the level of the individual pencils in the desk drawers in real-time on a current video card, it just isn't going to work. the other issue is the time to market cycle... consider, for example, the Intel AVX extensions: it was detailed out a number of years ago; it only started coming out in real chips fairly recently; it will likely be another number of years until most consumer systems have chips with this feature. and this was something trivial (adding larger YMM registers, and adding instruction forms which took additional arguments). other current/past examples would be pixel/fragment shaders: most graphics cards now have them, but there are still some floating around that don't. so, it is unlikely that Kurzweil is going to be able to make the market-cycle go away anytime soon (IOW: within the next several years). the closest thing we have at the moment is software, where features can be deployed more readily, and end users can get updated versions more quickly, rather than having to wait many years for everything to cycle through. As for biology, it is an iterative approach. The biggest insights are coming from better medical imaging techniques that allow us to see inside living systems' organs and better understand how they work. Kurzweil sort of discusses this too, and relates it to how it helped him develop better and more scalable algorithms even when the underlying hardware did not change. When was the last time Kurzweil did design something? 1990? Prior to that? can't comment much on this. ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
I've always suspected that it comes from the ability to see around corners, which appears to be a rare ability. If someone keeps seeing things that other people say aren't there, eventually it will drive them a little crazy :-) An amazing example of this (I think) is contained in this video: http://www.randsinrepose.com/archives/2011/10/06/you_are_underestimating_the_future.html Paul. From: John Zabroski johnzabro...@gmail.com To: Fundamentals of New Computing fonc@vpri.org Sent: Tuesday, October 25, 2011 11:55:29 AM Subject: Re: [fonc] IBM eyes brain-like computing Brian, I recommend you pick up a copy of Ray Kurzweil's The Singularity Is Near. Ray is smarter than basically everyone, and although a tad bit crazy (teaching at MIT will do that to you :)), he is a legitimate genius. Basically, before arguing about the limits of computing, read Ray Kurzweil. Others have written similar stuff here and there, but nobody is as passionate and willing to argue about the subject as Ray. Cheers, Z-Bo On Fri, Oct 14, 2011 at 2:44 PM, BGB cr88...@gmail.com wrote: On 10/14/2011 9:29 AM, karl ramberg wrote: Interesting article : http://www.itnews.com.au/News/276700,ibm-eyes-brain-like-computing.aspx Not much details, but the what they envisions seems to be more of the character a autonomic system that can be quarried for answers, not programmed like today's computers. I have seen stuff about this several times, with some articles actively demeaning and belittling / trivializing the existing pre-programmed Von Veumann / stored-program style machines. but, one can ask, but why then are there these machines in the first place: largely it is because the human mind also falls on its face for tasks which computers can perform easily, such as performing large amounts of calculations (and being readily updated). also, IBM is exploring some lines of chips (neural-net processors, ...) which may well be able to do a few interesting things, but I predict, will fall far short of their present claims. it is likely that the road forwards will not be a one or the other scenario, but will likely result in hybrid systems combining the strengths of both. for example, powerful neural-nets would be a nice addition, but I would not want to see them at the cost of programmability, ability to copy or install software, make backups, ... better IMO is if the neural nets could essentially exist in-computer as giant data-cubes under program control, which can be paused/resumed, or loaded from or stored to the HDD, ... also, programs using neural-nets would still remain as software in the traditional sense, and maybe neural-nets would be stored/copied/... as ordinary files. (for example, if a human-like mind could be represented as several TB worth of data-files...). granted, also debatable is how to best represent/process the neural-nets. IBM is exploring the use of hard-wired logic and crossbar arrays / memristors / ... also implied was that all of the neural state was stored in the chip itself in a non-volatile manner, and also (by implication from things read) not readily subject to being read/written externally. my own thoughts had been more along the lines of fine-grained GPUs, where the architecture would be vaguely similar to a GPU but probably with lots more cores and each likely only being a simple integer unit (or fixed-point), probably with some local cache memory. likely, these units would be specialized some for the task, with common calculations/... likely being handled in hardware. the more cheaper/immediate route would be, of course, to just do it on the GPU (lots of GPU power and OpenCL or similar). or maybe creating an OpenGL-like library dedicated mostly to running neural nets on the GPU (with both built-in neuron types, and maybe also neuronal shaders, sort of like fragment shaders or similar). maybe called OpenNNL or something... although potentially not as powerful (in terms of neurons/watt), I think my idea would have an advantage that it would allow more variety in neuron behavior, which could likely be necessary for making this sort of thing actually work in a practical sense. however, I think the idea of memristors is also cool, but I would presume that their use would more likely be as a type of RAM / NVRAM / SSD-like technology, and not in conflict with the existing technology and architecture. or such... ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
I was not arguing about the limits of computing, rather, IBM's specific design. it doesn't really realistically emulate real neurons, rather it is a from what I can gather a simplistic accumulate and fire model, with the neurons hard-wired into a grid. I suspect something more generic would be needed. another question is what can be done in the near term and on present hardware (future hardware may or may not exist, but any new hardware may take years to make it into typical end-user systems). the second part of the question is: assuming one can transition to a purely biology-like model, is this a good tradeoff?... if one gets rid of a few of the limitations of computers but gains some of the limitations of biology, this may not be an ideal solution. better would be to try for a strategy where the merits of both can be gained, and as many limitations as possible can be avoided. most likely, this would be via a hybrid model. or such... On 10/25/2011 9:07 AM, Paul Homer wrote: I've always suspected that it comes from the ability to see around corners, which appears to be a rare ability. If someone keeps seeing things that other people say aren't there, eventually it will drive them a little crazy :-) An amazing example of this (I think) is contained in this video: http://www.randsinrepose.com/archives/2011/10/06/you_are_underestimating_the_future.html Paul. *From:* John Zabroski johnzabro...@gmail.com *To:* Fundamentals of New Computing fonc@vpri.org *Sent:* Tuesday, October 25, 2011 11:55:29 AM *Subject:* Re: [fonc] IBM eyes brain-like computing Brian, I recommend you pick up a copy of Ray Kurzweil's The Singularity Is Near. Ray is smarter than basically everyone, and although a tad bit crazy (teaching at MIT will do that to you :)), he is a legitimate genius. Basically, before arguing about the limits of computing, read Ray Kurzweil. Others have written similar stuff here and there, but nobody is as passionate and willing to argue about the subject as Ray. Cheers, Z-Bo On Fri, Oct 14, 2011 at 2:44 PM, BGB cr88...@gmail.com mailto:cr88...@gmail.com wrote: On 10/14/2011 9:29 AM, karl ramberg wrote: Interesting article : http://www.itnews.com.au/News/276700,ibm-eyes-brain-like-computing.aspx Not much details, but the what they envisions seems to be more of the character a autonomic system that can be quarried for answers, not programmed like today's computers. I have seen stuff about this several times, with some articles actively demeaning and belittling / trivializing the existing pre-programmed Von Veumann / stored-program style machines. but, one can ask, but why then are there these machines in the first place: largely it is because the human mind also falls on its face for tasks which computers can perform easily, such as performing large amounts of calculations (and being readily updated). also, IBM is exploring some lines of chips (neural-net processors, ...) which may well be able to do a few interesting things, but I predict, will fall far short of their present claims. it is likely that the road forwards will not be a one or the other scenario, but will likely result in hybrid systems combining the strengths of both. for example, powerful neural-nets would be a nice addition, but I would not want to see them at the cost of programmability, ability to copy or install software, make backups, ... better IMO is if the neural nets could essentially exist in-computer as giant data-cubes under program control, which can be paused/resumed, or loaded from or stored to the HDD, ... also, programs using neural-nets would still remain as software in the traditional sense, and maybe neural-nets would be stored/copied/... as ordinary files. (for example, if a human-like mind could be represented as several TB worth of data-files...). granted, also debatable is how to best represent/process the neural-nets. IBM is exploring the use of hard-wired logic and crossbar arrays / memristors / ... also implied was that all of the neural state was stored in the chip itself in a non-volatile manner, and also (by implication from things read) not readily subject to being read/written externally. my own thoughts had been more along the lines of fine-grained GPUs, where the architecture would be vaguely similar to a GPU but probably with lots more cores and each likely only being a simple integer unit (or fixed-point
[fonc] IBM eyes brain-like computing
Interesting article : http://www.itnews.com.au/News/276700,ibm-eyes-brain-like-computing.aspx Not much details, but the what they envisions seems to be more of the character a autonomic system that can be quarried for answers, not programmed like today's computers. Karl ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
Re: [fonc] IBM eyes brain-like computing
On 10/14/2011 9:29 AM, karl ramberg wrote: Interesting article : http://www.itnews.com.au/News/276700,ibm-eyes-brain-like-computing.aspx Not much details, but the what they envisions seems to be more of the character a autonomic system that can be quarried for answers, not programmed like today's computers. I have seen stuff about this several times, with some articles actively demeaning and belittling / trivializing the existing pre-programmed Von Veumann / stored-program style machines. but, one can ask, but why then are there these machines in the first place: largely it is because the human mind also falls on its face for tasks which computers can perform easily, such as performing large amounts of calculations (and being readily updated). also, IBM is exploring some lines of chips (neural-net processors, ...) which may well be able to do a few interesting things, but I predict, will fall far short of their present claims. it is likely that the road forwards will not be a one or the other scenario, but will likely result in hybrid systems combining the strengths of both. for example, powerful neural-nets would be a nice addition, but I would not want to see them at the cost of programmability, ability to copy or install software, make backups, ... better IMO is if the neural nets could essentially exist in-computer as giant data-cubes under program control, which can be paused/resumed, or loaded from or stored to the HDD, ... also, programs using neural-nets would still remain as software in the traditional sense, and maybe neural-nets would be stored/copied/... as ordinary files. (for example, if a human-like mind could be represented as several TB worth of data-files...). granted, also debatable is how to best represent/process the neural-nets. IBM is exploring the use of hard-wired logic and crossbar arrays / memristors / ... also implied was that all of the neural state was stored in the chip itself in a non-volatile manner, and also (by implication from things read) not readily subject to being read/written externally. my own thoughts had been more along the lines of fine-grained GPUs, where the architecture would be vaguely similar to a GPU but probably with lots more cores and each likely only being a simple integer unit (or fixed-point), probably with some local cache memory. likely, these units would be specialized some for the task, with common calculations/... likely being handled in hardware. the more cheaper/immediate route would be, of course, to just do it on the GPU (lots of GPU power and OpenCL or similar). or maybe creating an OpenGL-like library dedicated mostly to running neural nets on the GPU (with both built-in neuron types, and maybe also neuronal shaders, sort of like fragment shaders or similar). maybe called OpenNNL or something... although potentially not as powerful (in terms of neurons/watt), I think my idea would have an advantage that it would allow more variety in neuron behavior, which could likely be necessary for making this sort of thing actually work in a practical sense. however, I think the idea of memristors is also cool, but I would presume that their use would more likely be as a type of RAM / NVRAM / SSD-like technology, and not in conflict with the existing technology and architecture. or such... ___ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc