You're half-right. Both punishment and training need to occur. Many large organizations simply do not have the kind of culture of accountability that is required to protect sensitive information "for real." Bureaucrats (by which I mean any member of a large organization, even soldiers and sailors) do not take information protection seriously until they begin to take seriously the idea that lapses will result in someone being fired and/or sent to prison.
NAVOCEANO is located in Bay St. Louis, MS. If you think someone from a "renowned bastion" is going to spend their days hangin' out in southern Mississippi...
it looks like this company is using MuscleFish to do the matching - MuscleFish was a audio query-by-content startup with actual products, which they bought out.
in fact, then, what this suggests is that there was no market for what you are calling Songle, so the technology wound up being perverted for use in DRM enforcement...
what is truly sad about this is that DEC Systems Research Center (aka DEC SRC, aka Compaq SRC, now part of HP Labs and basically moribund) built a
pocket-sized HDD jukebox in 1998.
needless to say, none of SRC's parent companies pushed it hard enough. (Compaq did sell PJBs commercially for a while.) so now HP gets to license the iPod...
first, Moore's Law is about density. processor speed improvements are just a historical corollary.
second, everyone's roadmap leads to a multicore design (an inevitable development as increasing transistor count gets easier relative to shrinking).
third, improving "throughput" performance is not the same as improving serial performance. you have to find and then exploit parallelism, and neither is easy outside of data centers (which is what the US-IV and friends are aimed at). personally, i have lots of problems that are bottlenecked on serial performance.
it's not clear that throughput computing is going to shift things dramatically in sun's direction.
(and this is from someone who likes sun gear. this week i tried to propose a V440 as a VLM server for my department and lost out to a pile of 8GB dual G5s on price/performance. the difference in serial performance at this point is just ridiculous.)
If you read the EU proposal (click through the Wired article and then to the first link), you would in fact see that they go to great lengths to distance themselves from business-method patents or any other patents that do not have technical content.
Contrary to the hysterical claims you read in/., Europe is not free of software patents now. Anyone who bothers to click through to read the EU proposal here will read the following observation:
The patenting of computer-implemented inventions is not new. Indeed, patents involving use of software have been applied for and granted since the earliest days of the European patent system and it is now estimated that 15% of all applications for patents received by the EPO relate to computer-implemented inventions.
amen on the SIMD comment. not being a HPC guy, i hadn't realized until recently that intel had absorbed the kuck & associates team (the folks who used to do vectorizing compilers for supers, back when people cared about supers). it is just too cool to see this just pop up when you type "make":
Sound_to_Pitch.c(211) : (col. 7) remark: LOOP WAS VECTORIZED.
If you're academically inclined, it's worth noting that the top grad schools accept a heavy proportion of undergrads from the top grad schools (the IITs being an exception, and we're not talking "Illinois Tech" here), and the top labs and universities mostly interview applicants from the top grad schools. To some degree this is as much about familiarity with the letter-writers as ostensible academic quality.
So while USNews' ranking isn't very relevant, that doesn't mean that you can ignore the fact that CMU/MIT/Berkeley/Stanford are at the top of the list in CS.
A Berkeley engineering professor who also works on energy policy went so far as to call this paper a failure of the peer review process on today's Science Friday segment on NPR - not because of the science of their models per se, but because the assumptions and parameters they used are so unrealistic as to invalidate the conclusions. (Interestingly, even though he's in the Cal nuclear engineering department, he also professed a strong belief that a nuclear powerplant buildout doesn't make any economic sense...)
(Streaming audio can be found if you dig around here.)
SF movies typically don't count as prior art...
on
Reading Lips In Software
·
· Score: 4, Informative
I don't know if they have any patents, we all know some prior "art" from 2001, er.. 1968.
patents are supposed to be on inventions, not ideas. (very) generally speaking, you have to demonstrate you know how to do something for it to count as prior art. actually building something counts, as does a patent application (since the patent application has to explain how the invention works at a reasonable level of detail, for an admittedly arguable legal definition of reasonable).
ianal, but the last i heard, a mention in a science fiction book or movie wouldn't typically be considered prior art. a person skilled in the art can't tell from 2001 how to make a computer read lips.
There have been a lot of things that have used position or motion sensors with handhelds. The granddaddy of them all was George Fitzmaurice's Chameleon system, which used Ascension Birds trackers to capture 6DOF motion - this was about ten years ago. Later systems include DEC WRL's Rock 'n' Scroll and the Xerox PARC "tilty interface" work, which both exploited the fact that accelerometers had become cheap.
The nice thing about the Peephole work is not that it invented the handheld-as-peephole idea (Fitzmaurice did that). The nice parts are in the details...
The basic idea of using audio similarity to "grep" short sounds out of audio streams (as opposed to using ASR and text-matching) is quite old - some classic papers based on dynamic timewarping date back to 1977, and HMMs became popular for this application about ten years after that. Papers on this kind of thing appear in conferences like ICASSP - look for keywords like "keyword spotting" or "wordspotting." The phone company wanted to do this for obvious reasons.
Note that I'm not saying the GATech technology used by this company is derivative - I haven't looked at the specifics of this approach.
This is interesting why? The "chosen" contains (1) MIT PDOS and two schools (NYU and UCB) where MIT PDOS alumni have recently been hired, (2) a network shop (ICSI/ACIRI) and (3) a security shop (Rice). Like many such "picks," it reflects human connections and a fit with someone's agenda more than some abstract notion of organizational merit.
i've filed several hardware and software patent applications, and read
through hundreds of pages of issued patents. i think what most folks
are missing is the following:
the claims are a key part of the patent. the rest of the
disclosure is extremely important as supporting material, but the
claims form the structure of what is being patented. they initially
form a tree of ideas from which nodes can be deleted, pruning the
coverage into subgraphs (usually a forest) of the idea space.
when it comes to software patents, there's no (business)
incentive to write narrow independent claims. your average patent
attorney seems to tell you, "write the claims as broadly as you
possibly can, making sure that your actual invention (the thing
you really understand to be your idea) is described in a dependent
claim somewhere." so if you think you're patenting a novel method of
toasting bread, you may wind up effectively patenting the concept of
toasting bread.
if the USPTO lets it slide (which they often do not,
but sometimes they do), you totally win. the fact that there's prior
art only helps someone who's really willing to fight in court instead
of (1) settling or (2) licensing or (3) entering into a
cross-licensing agreement.
if they do catch you in the application
process, you just rejig the application or file a continuation in a
way that preserves the original filing date - no harm, no foul.
i think this is what happened to these oracle guys - they have (and
describe as an embodiment) a very specific method but they let the
attorney talk them into independent claims that would be completely
obvious to anyone using a system that supports snapshots.
does this mean the patent is invalid? no, it means that some of
the claims (and therefore some of the applicability of the patent
coverage) can be challenged. since the specific method is entirely
defensible as non-obvious, they have certainly protected their actual
invention. it's just this ridiculous top-level claim that gets hosed.
does this happen to everyone? no. for example, most IBM patents
i've seen have perfectly reasonable and specific claims. it's only
when you have crappy, unethical patent counsel that things like this
occur.
the same could be said of computer science research as a whole. the "ground" got broken long ago; not much fundamental has happened since the period you describe. the action has been in application and deployment. the web and internet offer very little that wasn't fully envisaged by the mid-80s. (remember ted nelson's xanadu? bob taylor's original vision of "distributed network computing"?) java offers little that wasn't in smalltalk or one of the lisp object systems. today's ai works because of moore's law.
faulting parc (or any other lab) because lightning hasn't struck every decade is silly.
anyone who thinks that kind of lightning will strike in a startup, the symbol of innovation in "this" era, misunderstands the nature of real innovation and basic research.
observe that "the future of xerox" and "the future of western society" (or "computing industry" or whatever) are not necessarily the same thing. i think it's pretty plain what xerox's executives care more about right now and what the spokesperson meant.
well, let's start with basics. postgresql is process-per-user - the engine isn't multithreaded. this is a scalability issue for apps that produce a large number of connections. (it's less of an issue when you have someone managing the number of connections for you, like a web/app server.) that alone leaves oracle and db2 plenty of market-space that will never be challenged...
the original article headline is inflammatory and, as you say, completely unrealistic.
the main thing that's not quite clear from the first paragraph is that postgres shared essentially no code with ingres. (a few hundred lines were ripped off for some utility code.)
it's hard to plugin core engine functionality. postgres has always had a set of plugin features, with a relatively limited scope - types, user-defined functions, etc. (that was actually the original point of the project). many commercial databases have comparable interfaces now. the predator project at cornell goes significantly farther in that direction. the real problem is that major plugins are a huge pain to write. an example: almost nobody outside of berkeley ever managed to write an access method plugin (think b-tree replacement) for postgres. it was just too hard to understand the lock manager, buffer manager, etc. etc.
... and you'll void your warranty at best and break the machine at worst (the copiers are designed not to work if the watermark circuitry is disabled).
not a problem for el-cheapo devices with a 90 day warranty. more problematic for anything used in a business, or that isn't on clearance at fry's, or that will likely require service during its warranty period.
Slashdot Mirror
The whole VIDEA story (including the organizer's side) is here.
You're half-right. Both punishment and training need to occur. Many large organizations simply do not have the kind of culture of accountability that is required to protect sensitive information "for real." Bureaucrats (by which I mean any member of a large organization, even soldiers and sailors) do not take information protection seriously until they begin to take seriously the idea that lapses will result in someone being fired and/or sent to prison.
NAVOCEANO is located in Bay St. Louis, MS. If you think someone from a "renowned bastion" is going to spend their days hangin' out in southern Mississippi...
Interesting! Certainly Fujitsu acknowledges its roots in Fuji Electric, so I see your point. However, the relationship still appears to be one of accidental name resemblance - the company that currently operates as Fuji Electric denies being related to Fuji (Photo) Film...
Fuji Xerox is a joint venture between Fuji Film and Xerox. Whereas Fujitsu is a computer company, analogous in a number of ways to IBM.
it looks like this company is using MuscleFish to do the matching - MuscleFish was a audio query-by-content startup with actual products, which they bought out.
in fact, then, what this suggests is that there was no market for what you are calling Songle, so the technology wound up being perverted for use in DRM enforcement...
needless to say, none of SRC's parent companies pushed it hard enough. (Compaq did sell PJBs commercially for a while.) so now HP gets to license the iPod...
second, everyone's roadmap leads to a multicore design (an inevitable development as increasing transistor count gets easier relative to shrinking).
third, improving "throughput" performance is not the same as improving serial performance. you have to find and then exploit parallelism, and neither is easy outside of data centers (which is what the US-IV and friends are aimed at). personally, i have lots of problems that are bottlenecked on serial performance.
it's not clear that throughput computing is going to shift things dramatically in sun's direction.
(and this is from someone who likes sun gear. this week i tried to propose a V440 as a VLM server for my department and lost out to a pile of 8GB dual G5s on price/performance. the difference in serial performance at this point is just ridiculous.)
If you read the EU proposal (click through the Wired article and then to the first link), you would in fact see that they go to great lengths to distance themselves from business-method patents or any other patents that do not have technical content.
Business-method patents are not the issue here.
amen on the SIMD comment. not being a HPC guy, i hadn't realized until recently that intel had absorbed the kuck & associates team (the folks who used to do vectorizing compilers for supers, back when people cared about supers). it is just too cool to see this just pop up when you type "make":
Sound_to_Pitch.c(211) : (col. 7) remark: LOOP WAS VECTORIZED.
If you're academically inclined, it's worth noting that the top grad schools accept a heavy proportion of undergrads from the top grad schools (the IITs being an exception, and we're not talking "Illinois Tech" here), and the top labs and universities mostly interview applicants from the top grad schools. To some degree this is as much about familiarity with the letter-writers as ostensible academic quality.
So while USNews' ranking isn't very relevant, that doesn't mean that you can ignore the fact that CMU/MIT/Berkeley/Stanford are at the top of the list in CS.
(Streaming audio can be found if you dig around here.)
patents are supposed to be on inventions, not ideas. (very) generally speaking, you have to demonstrate you know how to do something for it to count as prior art. actually building something counts, as does a patent application (since the patent application has to explain how the invention works at a reasonable level of detail, for an admittedly arguable legal definition of reasonable).
ianal, but the last i heard, a mention in a science fiction book or movie wouldn't typically be considered prior art. a person skilled in the art can't tell from 2001 how to make a computer read lips.
There have been a lot of things that have used position or motion sensors with handhelds. The granddaddy of them all was George Fitzmaurice's Chameleon system, which used Ascension Birds trackers to capture 6DOF motion - this was about ten years ago. Later systems include DEC WRL's Rock 'n' Scroll and the Xerox PARC "tilty interface" work, which both exploited the fact that accelerometers had become cheap.
The nice thing about the Peephole work is not that it invented the handheld-as-peephole idea (Fitzmaurice did that). The nice parts are in the details...
The basic idea of using audio similarity to "grep" short sounds out of audio streams (as opposed to using ASR and text-matching) is quite old - some classic papers based on dynamic timewarping date back to 1977, and HMMs became popular for this application about ten years after that. Papers on this kind of thing appear in conferences like ICASSP - look for keywords like "keyword spotting" or "wordspotting." The phone company wanted to do this for obvious reasons.
Note that I'm not saying the GATech technology used by this company is derivative - I haven't looked at the specifics of this approach.
This is interesting why? The "chosen" contains (1) MIT PDOS and two schools (NYU and UCB) where MIT PDOS alumni have recently been hired, (2) a network shop (ICSI/ACIRI) and (3) a security shop (Rice). Like many such "picks," it reflects human connections and a fit with someone's agenda more than some abstract notion of organizational merit.
- the claims are a key part of the patent. the rest of the
disclosure is extremely important as supporting material, but the
claims form the structure of what is being patented. they initially
form a tree of ideas from which nodes can be deleted, pruning the
coverage into subgraphs (usually a forest) of the idea space.
- when it comes to software patents, there's no (business)
incentive to write narrow independent claims. your average patent
attorney seems to tell you, "write the claims as broadly as you
possibly can, making sure that your actual invention (the thing
you really understand to be your idea) is described in a dependent
claim somewhere." so if you think you're patenting a novel method of
toasting bread, you may wind up effectively patenting the concept of
toasting bread.
- if the USPTO lets it slide (which they often do not,
but sometimes they do), you totally win. the fact that there's prior
art only helps someone who's really willing to fight in court instead
of (1) settling or (2) licensing or (3) entering into a
cross-licensing agreement.
- if they do catch you in the application
process, you just rejig the application or file a continuation in a
way that preserves the original filing date - no harm, no foul.
i think this is what happened to these oracle guys - they have (and describe as an embodiment) a very specific method but they let the attorney talk them into independent claims that would be completely obvious to anyone using a system that supports snapshots.does this mean the patent is invalid? no, it means that some of the claims (and therefore some of the applicability of the patent coverage) can be challenged. since the specific method is entirely defensible as non-obvious, they have certainly protected their actual invention. it's just this ridiculous top-level claim that gets hosed.
does this happen to everyone? no. for example, most IBM patents i've seen have perfectly reasonable and specific claims. it's only when you have crappy, unethical patent counsel that things like this occur.
anyone who thinks that kind of lightning will strike in a startup, the symbol of innovation in "this" era, misunderstands the nature of real innovation and basic research.
presumably "wjr" is william j. rucklidge, who used to be at parc before going to a startup early this year.
observe that "the future of xerox" and "the future of western society" (or "computing industry" or whatever) are not necessarily the same thing. i think it's pretty plain what xerox's executives care more about right now and what the spokesperson meant.
the original article headline is inflammatory and, as you say, completely unrealistic.
the main thing that's not quite clear from the first paragraph is that postgres shared essentially no code with ingres. (a few hundred lines were ripped off for some utility code.)
it's hard to plugin core engine functionality. postgres has always had a set of plugin features, with a relatively limited scope - types, user-defined functions, etc. (that was actually the original point of the project). many commercial databases have comparable interfaces now. the predator project at cornell goes significantly farther in that direction. the real problem is that major plugins are a huge pain to write. an example: almost nobody outside of berkeley ever managed to write an access method plugin (think b-tree replacement) for postgres. it was just too hard to understand the lock manager, buffer manager, etc. etc.
... and you'll void your warranty at best and break the machine at worst (the copiers are designed not to work if the watermark circuitry is disabled).
not a problem for el-cheapo devices with a 90 day warranty. more problematic for anything used in a business, or that isn't on clearance at fry's, or that will likely require service during its warranty period.