Our chief insight is that while the space of images is effectively infinite, the space of semantically differentiable scenes is actually not that large. For many image completion tasks we are able to find similar scenes which contain image fragments that will convincingly complete the image. Our algorithm is entirely data-driven, requiring no annotations or labelling by the user.
What are the "semantics" here?
Is this like google images, where the nearby HTML text determines the classification of the image [i.e ASCII-text as meta-data for images]?
Or is this a great big neural net of wavelet data which classifies the images mathematically?
PS: I have the same question about that infamous Photosynth/Sea Dragon demonstration:
Throwing a bunch of rocks at a single bulldozer is a serial act.
The parallel problem is to get a fleet of 100 bulldozers or 1000 bulldozers or 10,000 bulldozers simultaneously attacking a pile of rocks so that:
A) The bulldozers aren't constantly colliding with one another, and
B) When the bulldozers back off to avoid colliding with one another, they aren't all just sitting around twiddling their thumbs, needlessly burning diesel fuel [not to mention "prevailing" union wages & time value of the loan which was used to purchase the bulldozers], while waiting in endlessly long lines until it's time for their turn [finally!] to take a whack at the pile of rocks, and so that
C) The inefficiencies of B) aren't so great that it's actually counterproductive to have introduced the extra bulldozers in the first place.
Anyone who has been a developer or managed developers can tell you that an expert can accomplish as much as 10 average developers. However, companies typically pay only a 10-20% premium for an expert over the average programmer. Whether or not their title is Lead, Architect, Development Manager, Guru or whatever nomenclature the company uses. I am not saying that if your average developer is paid $50k/year that you should pony up $500k/year for an expert. The employer/employee relationship never works like that, but what employers don't seem to realize is that in the end paying more saves them more.
This guy should be ready to put his money where his mouth is: If there really is such a thing as the über-programmer, who is literally 10 times more efficient than his average [median] colleague, THEN BY ALL MEANS he ought to be paid 10 times as much in salary - maybe even more.
The very fact that such disparities in salary don't exist means that either the über-programmer does not exist, or else there is something so screwy about the internal politics of corporations that the suits in management won't stand for some dweeb hax0r making ten times their salaries.
But I think that if the über-programmer really does exist, then eventually the free market will figure that out, and compensate him accordingly.
Seriously though, who knows? Maybe enough people suck at searching to make this service worthwhile
Given that much of modern intellectual life has degenerated into seeing who can come up with the best Google searches [or PubMed searches, or arXiv searches, or whatever], how does hiring someone to do your searching for you differ from hiring someone to write your term papers for you?
The patent, 'Synchronized Parallel Processing with Shared Memory' was issued in October 1991...
The stuff in Claim 1 of the patent (5,056,000) is basic multi-processor stuff which certainly wasn't actually novel in 1989 (when the patent was filed). And the Cell doesn't seem to violate it anyway. It appears to be Claim 6 they are suing over. Claim 6 describes a particular way of partitioning processing power in a MIMD system, but again I doubt it was novel in 1989.
Even in a worst case scenario, the patent ought to expire next year.
Good grief, am I the only guy who was once a red-blooded American juvenile delinquent?
I can't believe that a bunch of self-proclaimed "engineers" would even debate how to proceed in this matter.
If there were any justice in the world, they would be forced to turn in their pocket protectors & their slide rules and slither home with their tails between their legs and their heads hung in shame.
Well, if that's true, and assuming that anti-virus software can become sophisticated enough to recognize microcode-flashing attempts on reboot, then here are two possible hax0r/crax0r countterattacks:
1) Target machines, like servers, whose uptimes are measured in months, or even years [and if M$FT forces reboots for critical updates every Tuesday, forcing their uptimes down into the 7-day range, then target Intel/AMD servers with longer up-times, like FreeBSD, Linux, OSX, etc].
2) Flash the motherboard BIOS with hax0red/crax0red code, so that it will flash the CPU microcode [on every reboot, if necessary]. Or heck, flash the video [or sound] card's BIOS with with hax0red/crax0red code, and let them corrupt the CPU [on every reboot, if necessary].
But either way, the very fact that you can flash this stuff from within the OS just scares the bejeebus out of me.
Uhh, you can flash the CPU Microcode from within a Microsoft OS?!? [Other than DOS?!?]
Wouldn't this pose like the Mother of All Possible Challenges to the Black Hat community - how to right a worm that could flash CPU's, thereby rendering nearly limitless power over every possible sandboxing or anti-virus countermeasure?
Kinda the Helen of Troy [or Anne Boleyn] of Hax0r/Crax0r Wet Dreams?
Seriously - how can you hope to maintain the illusion of Virtual Machines if you can write to the microcode?
The SR-71 is easily the baddest mofo of any item in either the Smithsonian's downtown Air & Space or Air & Space II in the big hangar out by the airport [which is where the SR-71 sits, right smack in the middle of the floor, dominating everything else around it].
Badder than the Wright Bros' biplane, badder than Lindbergh's Spirit of St Louis, badder than Apollo 11, badder than the Space Shuttle.
Just one great big Samuel Jackson Pulp Fiction Bad Mofo of an airplane.
"Science" [to the extent that it hasn't already devolved into Yet Another Pagan Religion], is just another tiny sub-specialty within the broader discipline of Applied Mathematics.
And let's face it, unless the kid is one of these Korean super-geniuses, with an IQ of 300, who taught himself multi-variable calculus in his spare when he wasn't suckling on his mama's teat, then he [or she] just isn't going to know enough mathematics to tackle anything of any true "scientific" interest until his mid- to- late teens [and that's assuming home schooling; add another five years or more if he's in the publik skewlz].
At the age of four, I'd be concentrating on simple math games [lots and lots of counting], and maybe some pre-geometry, like playing with blocks, and legos, and Transformers, and, of course, drawing [you can never have enough blank drawing paper], but, much, much, much more important than that, at such a ripe young age, would be language skills.
Lots and lots and lots of ABC's, and basic spelling, and phonics out the wazoo, and handwriting, and sing-song games.
Actually, singing is very interesting in that it combines both Left Brain-ish stuff [song lyrics] with Right Brain-ish stuff [musical melodies & harmonies].
In other words: This is a little child we're talking about. Have fun. Make learning fun.
Life's too short to be miserable all the damned time.
In case of big data structures (in some of the cases I've had to deal with: hundreds of thousands of objects, built gradually and modified heavily during several hours) finding the exact sequence that causes the inconsistency can be a nightmare.
Did you mean to type "hundreds OR thousands of objects"?
I'm kinda dubious that even M$FT Vista has hundreds OF thousands of objects [i.e. hundreds OF thousands of instantiations of classes] when it boots up. Or that Vista even ships with hundreds OF thousands of classes capable of being instantiated.
Even the idea of "thousands" of objects makes the hair on the back of my neck stand up on end...
Or maybe you use the word "objects" to mean something other than "instantiations of classes"?
AMD's cool & quiet tech will shut down individual cores when you are not using them. I believe this is all new for the Barcelona. It idles down cores when you are not using them fully. It shuts off parts of cores that you aren't using (eg the FPU if you are only using integer instructions).
According to the last picture [imageID=9] in the Image Gallery, different cores on the same chipset can run at different voltages and different MHz's:
GP:Dell has agreed to buy Suse Linux Enterprise Server certificates from Microsoft
P:SUSE has become MS Linux. Thanks a lot, Novell.
That's the way business works in the real world.
You subcontract to a subcontractor who subcontracts to a sub-subcontractor who subcontracts to a sub-sub-subcontractor who subcontracts to... until eventually the circle is completed and someone subcontracts the final assembly to a different division within your very own company.
If you really want your precious little GPLv2'd operating system to come down out of the ivory tower and make a run for it in the real world, then you're going to have to get your hands dirty with this thing called "The Free Market".
The upside being that when it's all said and done, you get to enjoy this thing called "Profit" as a result.
The sign of an incompetent department/person is one who doesn't want to be benchmarked.
But the act of benchmarking takes TIME. And ENERGY.
And the last thing that people want to part with - people who are doing useful, productive, mission critical work - is time. Or energy.
Bean counters & suits have plenty of time and energy to devote to nonsense like this. People who actually do something productive with their days do not.
Consider classical computer coding.
Say that writing the code to satisfy some architectural requirement takes X units of time and Y units of energy.
Now suppose that someone has decided [perhaps correctly] that all code must be commented. Guess what? X just became 2X, and Y just became 2Y.
Now suppose that some PHB comes along and says that all work must be "benchmarked". Well guess what? That means that both the code writing and the code commenting must be "benchmarked", ergo the original X becomes 4X [and the original Y becomes 4Y].
And then suppose some little smart ass in the back of the room says, "Oh yeah, well if everything must be benchmarked, then that means that the benchmarking must be benchmarked!"
La voila - the original X units of time are now 8X, and the original Y units of energy are now 8X [and then when the benchmarking of the benchmarking must be benchmarked, we're at 16X & 16Y, & cetera ad infinitum...]
Ever watch "Office Space"? Rememeber the TPS Reports?
At some point, you just have to draw the line, and say, "No Más!"
At the same time, this is a rather ingenious way of creating a virtual toll for roads. If the power gathered is then invested into a public transport system, then you'll end up having drivers subsidise public transport. The fuel savings with public transport may well offset the extra fuel burnt through the turbine induced drag.
Socialism: Punishing that which makes sense [and which people desire] in order to subsidize that which is nonsense [and which people loathe].
BTW, just in case you weren't aware, the NJ Turnpike is already a toll road, so you're talking about a "virtual" toll on top of an actual toll.
1) To be a cert authority, don't you need at least a medium-sized farm of supercomputers to mine very large prime numbers [<=, say, 2^4096] from the greater ether of non-primes? And ain't that gonna require some pretty serious investment $$$'s?
2) A little off-topic, but what happens in RSA if you cheat, and use non-primes as your keys? [Often the math will still work, but sometimes it won't - and what goes wrong if it doesn't?]
GP:Can't we just give the processes weapons and let them decide which follows?
P:That is actually the kind of question that my Operations Research professor (who also did a lot of work in CPU simulation and performance estimating) used to throw onto final exams as the "separate the B+ from the A" question. If your answer was interesting enough he would send you over to one of his Masters candidates to see if it could be taken any further. So I wouldn't count your suggestion out from the start!
Behold: The Mother of All Possible Comp Sci Flame Wars: The Darwinistically Selected Genetic Algorithms -vs- the Intelligently Designed Algorithms.
Bumper Stickers $4.95; T-Shirts $19.95:
$DEITY does not play dice with the Turing Machine.
Slashdot Mirror
Hundreds of frat brothers discover the party entertainment potential of lighting their urine streams on fire.
E Coli live in the bowels, not in the kidneys or the bladder.
Think lighting farts with a match.
Our chief insight is that while the space of images is effectively infinite, the space of semantically differentiable scenes is actually not that large. For many image completion tasks we are able to find similar scenes which contain image fragments that will convincingly complete the image. Our algorithm is entirely data-driven, requiring no annotations or labelling by the user.
What are the "semantics" here?
Is this like google images, where the nearby HTML text determines the classification of the image [i.e ASCII-text as meta-data for images]?
Or is this a great big neural net of wavelet data which classifies the images mathematically?
PS: I have the same question about that infamous Photosynth/Sea Dragon demonstration:
http://www.ted.com/index.php/talks/view/id/129
Throwing a bunch of rocks at a single bulldozer is a serial act.
The parallel problem is to get a fleet of 100 bulldozers or 1000 bulldozers or 10,000 bulldozers simultaneously attacking a pile of rocks so that:
A) The bulldozers aren't constantly colliding with one another, and
B) When the bulldozers back off to avoid colliding with one another, they aren't all just sitting around twiddling their thumbs, needlessly burning diesel fuel [not to mention "prevailing" union wages & time value of the loan which was used to purchase the bulldozers], while waiting in endlessly long lines until it's time for their turn [finally!] to take a whack at the pile of rocks, and so that
C) The inefficiencies of B) aren't so great that it's actually counterproductive to have introduced the extra bulldozers in the first place.
Anyone who has been a developer or managed developers can tell you that an expert can accomplish as much as 10 average developers. However, companies typically pay only a 10-20% premium for an expert over the average programmer. Whether or not their title is Lead, Architect, Development Manager, Guru or whatever nomenclature the company uses. I am not saying that if your average developer is paid $50k/year that you should pony up $500k/year for an expert. The employer/employee relationship never works like that, but what employers don't seem to realize is that in the end paying more saves them more.
This guy should be ready to put his money where his mouth is: If there really is such a thing as the über-programmer, who is literally 10 times more efficient than his average [median] colleague, THEN BY ALL MEANS he ought to be paid 10 times as much in salary - maybe even more.
The very fact that such disparities in salary don't exist means that either the über-programmer does not exist, or else there is something so screwy about the internal politics of corporations that the suits in management won't stand for some dweeb hax0r making ten times their salaries.
But I think that if the über-programmer really does exist, then eventually the free market will figure that out, and compensate him accordingly.
Seriously though, who knows? Maybe enough people suck at searching to make this service worthwhile
Given that much of modern intellectual life has degenerated into seeing who can come up with the best Google searches [or PubMed searches, or arXiv searches, or whatever], how does hiring someone to do your searching for you differ from hiring someone to write your term papers for you?
The patent, 'Synchronized Parallel Processing with Shared Memory' was issued in October 1991...
The stuff in Claim 1 of the patent (5,056,000) is basic multi-processor stuff which certainly wasn't actually novel in 1989 (when the patent was filed). And the Cell doesn't seem to violate it anyway. It appears to be Claim 6 they are suing over. Claim 6 describes a particular way of partitioning processing power in a MIMD system, but again I doubt it was novel in 1989.
Even in a worst case scenario, the patent ought to expire next year.
So much for Hungarian notation.
Good grief, am I the only guy who was once a red-blooded American juvenile delinquent?
I can't believe that a bunch of self-proclaimed "engineers" would even debate how to proceed in this matter.
If there were any justice in the world, they would be forced to turn in their pocket protectors & their slide rules and slither home with their tails between their legs and their heads hung in shame.
Well, if that's true, and assuming that anti-virus software can become sophisticated enough to recognize microcode-flashing attempts on reboot, then here are two possible hax0r/crax0r countterattacks:
1) Target machines, like servers, whose uptimes are measured in months, or even years [and if M$FT forces reboots for critical updates every Tuesday, forcing their uptimes down into the 7-day range, then target Intel/AMD servers with longer up-times, like FreeBSD, Linux, OSX, etc].
2) Flash the motherboard BIOS with hax0red/crax0red code, so that it will flash the CPU microcode [on every reboot, if necessary]. Or heck, flash the video [or sound] card's BIOS with with hax0red/crax0red code, and let them corrupt the CPU [on every reboot, if necessary].
But either way, the very fact that you can flash this stuff from within the OS just scares the bejeebus out of me.
Uhh, you can flash the CPU Microcode from within a Microsoft OS?!? [Other than DOS?!?]
Wouldn't this pose like the Mother of All Possible Challenges to the Black Hat community - how to right a worm that could flash CPU's, thereby rendering nearly limitless power over every possible sandboxing or anti-virus countermeasure?
Kinda the Helen of Troy [or Anne Boleyn] of Hax0r/Crax0r Wet Dreams?
Seriously - how can you hope to maintain the illusion of Virtual Machines if you can write to the microcode?
Seriously.
The SR-71 is easily the baddest mofo of any item in either the Smithsonian's downtown Air & Space or Air & Space II in the big hangar out by the airport [which is where the SR-71 sits, right smack in the middle of the floor, dominating everything else around it].
Badder than the Wright Bros' biplane, badder than Lindbergh's Spirit of St Louis, badder than Apollo 11, badder than the Space Shuttle.
Just one great big Samuel Jackson Pulp Fiction Bad Mofo of an airplane.
PS: This even applies to labor markets. In that case we call the piracy 'slavery', and the low end versions 'volunteers'.
Dude, your homepage says you're in San Diego! How can you possibly say this with a straight face?
Everybody knows who the pirates are in the labor market...
"Science" [to the extent that it hasn't already devolved into Yet Another Pagan Religion], is just another tiny sub-specialty within the broader discipline of Applied Mathematics.
And let's face it, unless the kid is one of these Korean super-geniuses, with an IQ of 300, who taught himself multi-variable calculus in his spare when he wasn't suckling on his mama's teat, then he [or she] just isn't going to know enough mathematics to tackle anything of any true "scientific" interest until his mid- to- late teens [and that's assuming home schooling; add another five years or more if he's in the publik skewlz].
At the age of four, I'd be concentrating on simple math games [lots and lots of counting], and maybe some pre-geometry, like playing with blocks, and legos, and Transformers, and, of course, drawing [you can never have enough blank drawing paper], but, much, much, much more important than that, at such a ripe young age, would be language skills.
Lots and lots and lots of ABC's, and basic spelling, and phonics out the wazoo, and handwriting, and sing-song games.
Actually, singing is very interesting in that it combines both Left Brain-ish stuff [song lyrics] with Right Brain-ish stuff [musical melodies & harmonies].
In other words: This is a little child we're talking about. Have fun. Make learning fun.
Life's too short to be miserable all the damned time.
making a computer understand Bubba Sixpack who can't type
Dude, your clichés are like so Eisenhower Administration.
This is Bush 43; get with the times.
Yeh wanna salsa weedat, señor?
In case of big data structures (in some of the cases I've had to deal with: hundreds of thousands of objects, built gradually and modified heavily during several hours) finding the exact sequence that causes the inconsistency can be a nightmare.
Did you mean to type "hundreds OR thousands of objects"?
I'm kinda dubious that even M$FT Vista has hundreds OF thousands of objects [i.e. hundreds OF thousands of instantiations of classes] when it boots up. Or that Vista even ships with hundreds OF thousands of classes capable of being instantiated.
Even the idea of "thousands" of objects makes the hair on the back of my neck stand up on end...
Or maybe you use the word "objects" to mean something other than "instantiations of classes"?
AMD's cool & quiet tech will shut down individual cores when you are not using them. I believe this is all new for the Barcelona. It idles down cores when you are not using them fully. It shuts off parts of cores that you aren't using (eg the FPU if you are only using integer instructions).
According to the last picture [imageID=9] in the Image Gallery, different cores on the same chipset can run at different voltages and different MHz's:
http://www.informationweek.com/galleries/showImag
NtOnMyPC.bat
WksOnMyn.bat
NTONMYPC.bat
Do not M$FT file systems tend to be case insensitive?
Little known fact: If you don't patent before you publish, you lose your patent rights.
Come work for me... in the entertainment sector...
You gotta link for that offer?
GP: Dell has agreed to buy Suse Linux Enterprise Server certificates from Microsoft
P: SUSE has become MS Linux. Thanks a lot, Novell.
That's the way business works in the real world.
You subcontract to a subcontractor who subcontracts to a sub-subcontractor who subcontracts to a sub-sub-subcontractor who subcontracts to... until eventually the circle is completed and someone subcontracts the final assembly to a different division within your very own company.
If you really want your precious little GPLv2'd operating system to come down out of the ivory tower and make a run for it in the real world, then you're going to have to get your hands dirty with this thing called "The Free Market".
The upside being that when it's all said and done, you get to enjoy this thing called "Profit" as a result.
The sign of an incompetent department/person is one who doesn't want to be benchmarked.
But the act of benchmarking takes TIME. And ENERGY.
And the last thing that people want to part with - people who are doing useful, productive, mission critical work - is time. Or energy.
Bean counters & suits have plenty of time and energy to devote to nonsense like this. People who actually do something productive with their days do not.
Consider classical computer coding.
Say that writing the code to satisfy some architectural requirement takes X units of time and Y units of energy.
Now suppose that someone has decided [perhaps correctly] that all code must be commented. Guess what? X just became 2X, and Y just became 2Y.
Now suppose that some PHB comes along and says that all work must be "benchmarked". Well guess what? That means that both the code writing and the code commenting must be "benchmarked", ergo the original X becomes 4X [and the original Y becomes 4Y].
And then suppose some little smart ass in the back of the room says, "Oh yeah, well if everything must be benchmarked, then that means that the benchmarking must be benchmarked!"
La voila - the original X units of time are now 8X, and the original Y units of energy are now 8X [and then when the benchmarking of the benchmarking must be benchmarked, we're at 16X & 16Y, & cetera ad infinitum...]
Ever watch "Office Space"? Rememeber the TPS Reports?
At some point, you just have to draw the line, and say, "No Más!"
At the same time, this is a rather ingenious way of creating a virtual toll for roads. If the power gathered is then invested into a public transport system, then you'll end up having drivers subsidise public transport. The fuel savings with public transport may well offset the extra fuel burnt through the turbine induced drag.
Socialism: Punishing that which makes sense [and which people desire] in order to subsidize that which is nonsense [and which people loathe].
BTW, just in case you weren't aware, the NJ Turnpike is already a toll road, so you're talking about a "virtual" toll on top of an actual toll.
[Not that you'd care...]
Two points.
1) To be a cert authority, don't you need at least a medium-sized farm of supercomputers to mine very large prime numbers [<=, say, 2^4096] from the greater ether of non-primes? And ain't that gonna require some pretty serious investment $$$'s?
2) A little off-topic, but what happens in RSA if you cheat, and use non-primes as your keys? [Often the math will still work, but sometimes it won't - and what goes wrong if it doesn't?]
GP: Can't we just give the processes weapons and let them decide which follows?
P: That is actually the kind of question that my Operations Research professor (who also did a lot of work in CPU simulation and performance estimating) used to throw onto final exams as the "separate the B+ from the A" question. If your answer was interesting enough he would send you over to one of his Masters candidates to see if it could be taken any further. So I wouldn't count your suggestion out from the start!
Behold: The Mother of All Possible Comp Sci Flame Wars: The Darwinistically Selected Genetic Algorithms -vs- the Intelligently Designed Algorithms.
Bumper Stickers $4.95; T-Shirts $19.95: