It's more than tweaking the system. It's first creating the software to even work with the data, and then coming up with an algorithm that does what would normally be the function of an intelligent person -- namely, making recommendations that are on a par with what a human site editor would have done, based on the behavior of ordinary (non-expert) users.
I am no fan of trivial patents, particularly things like one-click shopping, but I am confident that the work the patent descibes is non-trivially inventive.
Personally, I would be pretty interested in prior art. I sure tried to find it then, because I just wanted to generate really good recommendations -- I did not set out to reinvent the wheel.
Case-based reasoning is a far cry from what this patent is describing.
Specifically, CBR assumes a symbolic rich representation of prior cases, and a pre-built distance function for measuring the similarity of the current situation to prior cases.
Most of the AI in CBR is in 1) choosing the right representation, and 2) coming up with the right similarity metric.
In the technique that Linden et al have patented, they make no assumption about a representationally rich framework for describing events -- in other words, they didn't have to spend months coming up with the right representation of the data. Also, and more importantly, they probably did not pre-build a similarity metric to match against prior cases. Instead, they applied statistical techniques to come up with the right similarity metric.
There are likely other differences -- I don't think CBR circa 1999 was applied to commercial transaction data in order to generate sales recommendations. B
The nearest I know of is that both CBR and Memory Based Reasoning likely were tried by other companies to do credit card fraud detection, which itself is a transactional problem involving large amounts of data. Still, that's pretty far from "I see you're buying a Sony plasma TV, would you also like to buy a Denon DVD player?"
At the time the patent was filed, it was extremely uncommon for systems to make automatic recommendations based solely on the behavior of users. When I did my work at Alexa Internet (which was acquired by Amazon) in the late 90s, I had to solve a number of issues which had not been dealt with, both from an engineering perspective and from a quality of results perspective -- few companies, and no academic researchers that I am aware of -- had both the amount of data and the technical talent required to process it in order to test and refine recommendation systems based on transactional information.
My work in this area became Amazon's "customers who shopped for X also shopped for Y feature." Greg Linden, the first name on this patent, is now doing interesting recommendation work with his site Findory.
Jef P (of acme.com fame) probably runs one of his own servers, like thttpd. It's actually quite robust. If he's slashdotted, I suspect it'shis T1 being overwhelmed.
It depends on the cargo. I've heard that cargo ships leaving Coos Bay, OR for either Hawaii or Japan sometimes travel so slowly they take weeks to make the run.
Actually the 5th amendment probably wouldn't protect you in this case. If the police have a search warrant that includes the contents of those encrypted files you would have to turn over the key the same way you would have to turn over the key to your safe if they have a warrant.
I am interested in whether this is the case. If the police have a warrant to get files, is the accused required to tell them how to read the files? If so, this sounds like you can be compelled to testify against yourself.
Could someone with more legal expertise please explain what US laws or cases would require an accused to cough up an encryption key or password?
Assuming the parent and grandparent posts are correct, what prevents some outside company making a commercial tool that fits the DHS's needs is only that it is unlikely that a custom tool the DHW would want also has broad market potential.
In other words, for a company to make a commercial tool just for a DHS spec, it would be essentially coding on spec but without a contract. And given that the DHS would presumably be putting out a contract because they couldn't find the product in the market already, that implies that it is there aren't enough other interested custmers in this product.
Short answer: nothing is preventing someone from doing this except market risk and reality.
It depends. First, not all US universities are on a tenure system (and never mind that the university in the article is in Spain).
Second, most new professors are years away from tenure.
Now, even for tenured professors at US schools, tenure isn't quite the shield you describe. For instance, if the professor is in a profit-center department (otherwise known as biology, computer science, or one of the other funded areas), if the professor isn't pulling in sufficient grants, the university can put them in a broom-closet like space where it's unlikely that they'll have the lab facilities to put together successful proposals (in biology, for instance, you usually need to have done most of the research before putting in funding -- grants are that competitive and agencies that risk averse).
Now, let's say that being put in a broom closet isn't bad enough. The university can get rid of tenured professors by eliminating the department.
Here's an example (not of retribution against tenured professsors, but simply of how a department closing can lead to selective firing of tenured professors). In 1990, Brandeis University had a linguistics department with 6 faculty, and I believe all six had tenure. The university decided to close the department to save money (at the time, the school was eating its endowment, not just interest on the money).
The university then made offers to 3 of the 6 professors (including Ray Jackendoff) to join other departments.
(running an open access point) sounds great, right up to the point where some pervert uses your open wi-fi to download child porn...
Right now, there are a zillion anonymous proxies on wired connections. It's far more likely and convenient for J. Random Hacker to connect to one of these always-on proxies that are available from anywhere in the world than to get within 100' of your fiddly little access point.
If you're really worried about someone within 100' of your house doing something bad, sure, create a lead-lined bunker, put up chicken wire to keep the signals from leaking out, run encryption, etc.
Or, do what SFLan and hundreds of other metropolitan open wireless networks do -- stop worrying and share.
The Presidio had 10+ terabytes of storage as of 1999 or 2000. It's where the Internet Archive and now-Amazon subsidiary Alexa Internet started.
If I remember correctly, around 1999, they had approximately 2B web pages on disk in the most recent snapshot, with the previous two snapshots on disk and another n snapshots on tape.
California doesn't prohibit private ownership of diesel cars -- you can drive a beater diesel Mercedes or Volvo if you want -- but it does have high emissions standards on new privately owned diesel vehicles. That said, auto manufacturers are aware of this and I believe make California-spec diesel cars.
--Pat
Re:Actually, Microsoft is now paying for its
on
Gates on Google
·
· Score: 1
crovira writes:
There are almost no designers of PC chassis left. The differentiation comes with 'plastic panels' on the same box. Regardless of which panels you might buy, you're still stuck with the box underneath it all.
So true. I spent a few months searching for a cool PC case. There are very few cases that look like they were designed to be both beautiful and well-built. When it comes to rack mounted servers, they all look like they were designed by a high school shop class and then manufactured next to hell's own pig iron forge.
It's so bad that there are some people who buy old SGI rackmounts just to have decent-looking cases for their PC servers.
Apple, by contrast, has beautifully designed and built servers. the Xserves are great from an engineering and an aesthetic point of view.
All the PC case makers seem to be in a rush to be the cheapest; I'm surprised that at least one of them doesn't get out of that race and distinguish themselves by making something even half as interesting as Apple's cases and charge a small premium.
I am guessing that Google's bandwidth use is mostly from crawling. If this is correct, then they may have a fair amount of 'free' upload bandwidth that they aren't using. This application would use that bandwidth.
In return, they get real-time data on sites, indicating how relevant links are and also how frequently they should recrawl certain pages.
I'm not commenting on Microsoft's recent spending patterns, just pointing out that managing R&D costs in order to provide the illusion of smooth growth is a technique commonly used by large companies, and one that others have suggested that Microsoft uses.
(2) The alleged $1b cut in R&D expense was actually due to a $1b increase in R&D expense in the year before quarter.
This is the key element. Large companies often use R&D spending as a way to smooth out the numbers year to year, because most investors like a nice predictable growth curve rather than dramatic swings.
In flush years, a company will put more money into R&D not so much to have R&D as to smooth out the numbers. Then, in subsequent years when times are lean, they can cut back on R&D as needed to meet their numbers.
In this respect, R&D spending is like a capacitor. Companies take it on as an expense when times are good so that they can quickly eliminate an expense when times are bad. I know, it sounds insane, but it's such a common practice that finance majors probably learn about it in school.
Loco3KGT says "That's going to be one hell of a return on investment for him."
Let's say he bought the magazine in 1959 for $1.
It's taken him 46 years to turn $1 into $10000 (Intel's bounty).
His annual rate of return r is:
$1 * r ^ 46 = $10000
r ~= 22.5%
Not bad at all.
Even if we take into count the costs he bore to get this reward, he's still doing pretty well. Assume that his collection of magazines cost $1000 to buy and $100/year to store (10 sq ft at $10/sq ft / yr):
You do pay more for a dual processor machine, but it's not an obscene price. You can put together an 8GB RAM machine for well under $4k. Memory is the main cost in these systems.
As a data point, I built a machine around a Tyan S2882 motherboard with 8GB of RAM for just over $3000. This is a dual processor machine with 8 1GB modules. Half this cost is memory. This price includes the assembly cost, shipping, and tax.
Going to 16GB means using more expensive 2GB modules, meaning memory is an even greater percent of the system cost.
Going to 64B means going to a quad board with 2GB modules. You'll add perhaps $1000 for the upgrade to the quad CPU motherboard and additional CPUs, but the RAM is probably now 65% of the system cost.
I wouldn't be surprised if they do have in-building towers in some places. For instance, some SF cell providers have coverate in the BART system in stations that are three to five stories under ground (Powell St Station, for one).
I don't know if they have a tower or some sort of repeater, but I would be surprised if the signal was coming through fifty feet of dirt, power lines, reinforced concrete, and city water mains.
--Pat
Slashdot Mirror
--Pat / zippy@cs.brandeis.edu
I am no fan of trivial patents, particularly things like one-click shopping, but I am confident that the work the patent descibes is non-trivially inventive.
Personally, I would be pretty interested in prior art. I sure tried to find it then, because I just wanted to generate really good recommendations -- I did not set out to reinvent the wheel.
--Pat / zippy@cs.brandeis.edu / blog / pics
Specifically, CBR assumes a symbolic rich representation of prior cases, and a pre-built distance function for measuring the similarity of the current situation to prior cases.
Most of the AI in CBR is in 1) choosing the right representation, and 2) coming up with the right similarity metric.
In the technique that Linden et al have patented, they make no assumption about a representationally rich framework for describing events -- in other words, they didn't have to spend months coming up with the right representation of the data. Also, and more importantly, they probably did not pre-build a similarity metric to match against prior cases. Instead, they applied statistical techniques to come up with the right similarity metric.
There are likely other differences -- I don't think CBR circa 1999 was applied to commercial transaction data in order to generate sales recommendations. B
The nearest I know of is that both CBR and Memory Based Reasoning likely were tried by other companies to do credit card fraud detection, which itself is a transactional problem involving large amounts of data. Still, that's pretty far from "I see you're buying a Sony plasma TV, would you also like to buy a Denon DVD player?"
--Pat / zippy@cs.brandeis.edu / blog / pics
--Pat / zippy@cs.brandeis.edu / blog / pics
At the time the patent was filed, it was extremely uncommon for systems to make automatic recommendations based solely on the behavior of users. When I did my work at Alexa Internet (which was acquired by Amazon) in the late 90s, I had to solve a number of issues which had not been dealt with, both from an engineering perspective and from a quality of results perspective -- few companies, and no academic researchers that I am aware of -- had both the amount of data and the technical talent required to process it in order to test and refine recommendation systems based on transactional information.
My work in this area became Amazon's "customers who shopped for X also shopped for Y feature." Greg Linden, the first name on this patent, is now doing interesting recommendation work with his site Findory.
--Pat / zippy@cs.brandeis.edu / blog / pics.
That'd work really well for the pictures, but what about the videos?
Flip books!
--Pat / zippy@cs.brandeis.edu / pics / blog
Bury the evilness in the make file. No one will find it there.
--Pat
Best. Bad typesetting example. Ever.
--Pat "don't make me spell it out"
Jef P (of acme.com fame) probably runs one of his own servers, like thttpd. It's actually quite robust. If he's slashdotted, I suspect it'shis T1 being overwhelmed.
--Pat
It depends on the cargo. I've heard that cargo ships leaving Coos Bay, OR for either Hawaii or Japan sometimes travel so slowly they take weeks to make the run.
--Pat
Finally, I'll be able to understand all the Chinese and Russian spam in my Inbox!
--Pat
I am interested in whether this is the case. If the police have a warrant to get files, is the accused required to tell them how to read the files? If so, this sounds like you can be compelled to testify against yourself.
Could someone with more legal expertise please explain what US laws or cases would require an accused to cough up an encryption key or password?
--Pat
Assuming the parent and grandparent posts are correct, what prevents some outside company making a commercial tool that fits the DHS's needs is only that it is unlikely that a custom tool the DHW would want also has broad market potential.
In other words, for a company to make a commercial tool just for a DHS spec, it would be essentially coding on spec but without a contract. And given that the DHS would presumably be putting out a contract because they couldn't find the product in the market already, that implies that it is there aren't enough other interested custmers in this product.
Short answer: nothing is preventing someone from doing this except market risk and reality.
--Pat / zippy@cs.brandeis.edu
It depends. First, not all US universities are on a tenure system (and never mind that the university in the article is in Spain).
Second, most new professors are years away from tenure.
Now, even for tenured professors at US schools, tenure isn't quite the shield you describe. For instance, if the professor is in a profit-center department (otherwise known as biology, computer science, or one of the other funded areas), if the professor isn't pulling in sufficient grants, the university can put them in a broom-closet like space where it's unlikely that they'll have the lab facilities to put together successful proposals (in biology, for instance, you usually need to have done most of the research before putting in funding -- grants are that competitive and agencies that risk averse).
Now, let's say that being put in a broom closet isn't bad enough. The university can get rid of tenured professors by eliminating the department.
Here's an example (not of retribution against tenured professsors, but simply of how a department closing can lead to selective firing of tenured professors). In 1990, Brandeis University had a linguistics department with 6 faculty, and I believe all six had tenure. The university decided to close the department to save money (at the time, the school was eating its endowment, not just interest on the money).
The university then made offers to 3 of the 6 professors (including Ray Jackendoff) to join other departments.
Effectively, 3 tenured professors were fired.
--Pat
From the parent post:
...
(running an open access point) sounds great, right up to the point where some pervert uses your open wi-fi to download child porn
Right now, there are a zillion anonymous proxies on wired connections. It's far more likely and convenient for J. Random Hacker to connect to one of these always-on proxies that are available from anywhere in the world than to get within 100' of your fiddly little access point.
If you're really worried about someone within 100' of your house doing something bad, sure, create a lead-lined bunker, put up chicken wire to keep the signals from leaking out, run encryption, etc.
Or, do what SFLan and hundreds of other metropolitan open wireless networks do -- stop worrying and share.
--Pat
The Presidio had 10+ terabytes of storage as of 1999 or 2000. It's where the Internet Archive and now-Amazon subsidiary Alexa Internet started.
If I remember correctly, around 1999, they had approximately 2B web pages on disk in the most recent snapshot, with the previous two snapshots on disk and another n snapshots on tape.
--Pat / zippy@cs.brandeis.edu
The last quote has got to be German via Babelfish.
--Pat
--Pat
So true. I spent a few months searching for a cool PC case. There are very few cases that look like they were designed to be both beautiful and well-built. When it comes to rack mounted servers, they all look like they were designed by a high school shop class and then manufactured next to hell's own pig iron forge.
It's so bad that there are some people who buy old SGI rackmounts just to have decent-looking cases for their PC servers.
Apple, by contrast, has beautifully designed and built servers. the Xserves are great from an engineering and an aesthetic point of view.
All the PC case makers seem to be in a rush to be the cheapest; I'm surprised that at least one of them doesn't get out of that race and distinguish themselves by making something even half as interesting as Apple's cases and charge a small premium.
--Pat / zippy@cs.brandeis.edu / blog
I am guessing that Google's bandwidth use is mostly from crawling. If this is correct, then they may have a fair amount of 'free' upload bandwidth that they aren't using. This application would use that bandwidth.
In return, they get real-time data on sites, indicating how relevant links are and also how frequently they should recrawl certain pages.
--Pat / zippy@cs.brandeis.edu
I'm not commenting on Microsoft's recent spending patterns, just pointing out that managing R&D costs in order to provide the illusion of smooth growth is a technique commonly used by large companies, and one that others have suggested that Microsoft uses.
--Pat
This is the key element. Large companies often use R&D spending as a way to smooth out the numbers year to year, because most investors like a nice predictable growth curve rather than dramatic swings.
In flush years, a company will put more money into R&D not so much to have R&D as to smooth out the numbers. Then, in subsequent years when times are lean, they can cut back on R&D as needed to meet their numbers.
In this respect, R&D spending is like a capacitor. Companies take it on as an expense when times are good so that they can quickly eliminate an expense when times are bad. I know, it sounds insane, but it's such a common practice that finance majors probably learn about it in school.
--Pat / zippy@cs.brandeis.edu
Loco3KGT says "That's going to be one hell of a return on investment for him."
:
Let's say he bought the magazine in 1959 for $1.
It's taken him 46 years to turn $1 into $10000 (Intel's bounty).
His annual rate of return r is
$1 * r ^ 46 = $10000
r ~= 22.5%
Not bad at all.
Even if we take into count the costs he bore to get this reward, he's still doing pretty well. Assume that his collection of magazines cost $1000 to buy and $100/year to store (10 sq ft at $10/sq ft / yr):
$1 * r ^ 46 = $10000 - $1000 - 46 * $100
= $4400
r ~= 20%
--Pat
You do pay more for a dual processor machine, but it's not an obscene price. You can put together an 8GB RAM machine for well under $4k. Memory is the main cost in these systems.
As a data point, I built a machine around a Tyan S2882 motherboard with 8GB of RAM for just over $3000. This is a dual processor machine with 8 1GB modules. Half this cost is memory. This price includes the assembly cost, shipping, and tax.
Going to 16GB means using more expensive 2GB modules, meaning memory is an even greater percent of the system cost.
Going to 64B means going to a quad board with 2GB modules. You'll add perhaps $1000 for the upgrade to the quad CPU motherboard and additional CPUs, but the RAM is probably now 65% of the system cost.
--Pat / zippy@cs.brandeis.edu
I wouldn't be surprised if they do have in-building towers in some places. For instance, some SF cell providers have coverate in the BART system in stations that are three to five stories under ground (Powell St Station, for one). I don't know if they have a tower or some sort of repeater, but I would be surprised if the signal was coming through fifty feet of dirt, power lines, reinforced concrete, and city water mains. --Pat