The U.S. is coming closer to pulling its weight on new capacity: in 2008, it installed 8.3 GW, while Germany installed 1.7 GW, or about a 5x factor. Not quite the 30x factor of land area, but hey, 1.7 million of those sq km are in Alaska, which is kind of inconvenient for electric transmission (same reason Canada's wind power is fairly low, despite massive land area).
For example, there was a bunch of excited speculation about when China's GDP would surpass the U.S.'s, despite the fact that that would still leave China nowhere near the U.S. on a per-capita GDP basis.
A huge proportion of censorship of, for example, literary works, was done by elected governments---Ireland's government, which has been democratic since its independence, is pretty notorious for their treatment of James Joyce's works, for example. That doesn't make it not government censorship.
Just because various countries are sovereign states doesn't mean it's inappropriate to criticize them if they're run by hard-right authoritarians (or hard-left authoritarians, for that matter).
With how often Facebook changes their API and makes subtle, app-breaking changes to how existing stuff operates, you won't really be able to rely on anything in a published book.
The 965 chipset's MTRRs are stillbroken, after a dozen firmware revisions, but nobody at Intel cares because it only breaks 64-bit Linux with >=4gb of RAM.
While the local channels are likely to face technical problems by a delay, the networks themselves seem to be part of the group lobbying for a delay, hoping the switchover can take place after their seasons end.
The stuff he's talking about is fairly generic data collection and analysis software, not really physics-specific. LabView is a semi-standard piece of software for interfacing with laboratory machines, setting up experiments and reading in the data and storing it somewhere. Igor is analysis and graphing software. Inventor is a 3d modeling/CAD package. Eagle is an ECAD (CAD for circuits basically) package with some automated layout capabilities.
Some of these are easier to replace than others. Data analysis can generally be done by a mixture of R and Octave, as you note; and for just graphing/plotting/figures, there's a billion open-source alternatives, including a ton of Python libraries these days.
LabView is going to be trickier to replace, due to interoperability: there's a lot of machines out there that interface with LabView out of the box, exposing APIs that LabView can use to control experiments, and sending back data in formats LabView can deal with. I'm not aware of even a significant attempt to replace it, really. (This is partly because what it does is really boring.)
I don't know what open source alternatives for CAD are; that's a whole topic in itself probably.
I don't own any way of hooking up my phone to a computer (I don't think my phone even has a USB connection), I don't know or have any interest in learning how to configure Linux to talk to a USB-cell-phone modem, etc.
...mainly because I expect to see some Representative or Senator, with no hint of irony or sense of shame, offering an unrelated amendment to that bill
The creation of the coupon program was largely the result of industry pressure---they're worried that some number of people who use only over-the-air TV will disappear after their current equipment stops working, and never return, thereby harming viewership numbers. This was a particular concern of networks like Univision.
We're talking about a copyright license on the text of a scientific paper here. The normal status is "all rights reserved". Cc-by-nd is strictly more permissive than the default copyright. The restrictions on commercialization, public performance, etc., are all part of the copyright license on the text of the paper, nothing more. The ideas in the paper are protected, if at all, by patent law.
Evaluation committees don't look only at the number of lines in your CV, but at how frequently your papers are cited. Papers in closed-access journals are much less frequently cited than papers in open-access journals. At the very least, if you publish in closed-access journals and don't want to shoot yourself in the foot, you should put up a PDF on your website so potential citers can actually get your article rather than just shrugging and giving up on it (sometimes this is permitted, sometimes officially illegal but rarely enforced).
I agree it's one of many factors, but it's been the deciding factor for me several times---if there are multiple journals roughly in the same class, the open-access one is better for my career, because it results in actually publishing my paper rather than locking it away in a box.
People noticed some decades ago that failure analyses seemed to predict failures with much less frequency than failures actually occurred. Why? Because the failure analyses themselves have some frequency with which they're wrong (rest on incorrect assumptions, themselves contain errors, etc.). Furthermore, failures can come not only from poor design, but poor implementation, poor initial data (e.g. the site wasn't what you thought it was like), etc. If you want accurate estimates of the overall likelihood of failure, as you generally do, modern failure analysis tends to take an empirical, whole-system view, in which no component of the process is assumed a priori to have a zero error rate.
a lot is being pushed to the client
on
Less Is Moore
·
· Score: 1
You don't have to imagine futuristic AI to think of computationally intensive stuff that's in the process of being pushed to the client. Excel is getting machine-learning and data-mining algorithms added to it, for example, so in future versions you'll be able to not only plot your data, but plot kernel-smoothed versions, cluster it using various clustering algorithms, fill in incomplete labels using, say, SVMs, and so on. A lot of the stuff that statisticians and savvy business analysts do in R now will be things that are standard and expected for a much wider range of people in the future (and R itself is getting more uptake outside of statistics professors).
Maybe Google will offer a server-side version of that sort of thing for Google Docs, but that would require a lot of processing power on their end.
It's just that, with FireFox, anything that isn't related to bare simple display of HTML pages, is usually tucked into separate plugins.
Firefox includes all sorts of "security" stuff turned on by default, some of it both pointless and really annoying, like the click-4-times annoyance when you want to visit any https site that doesn't have its SSL certificate signed by one of the worthless central authorities. Some of it is also useful, like popup blocking. "Clickjacking" prevention seems like it'd go in the same category of stuff, if Firefox were interested.
Perhaps you have't looked lately but DIMMs and el-cheapo graphics cards are both around "free" these days. As in, a 512-mb DIMM costs less than a sandwich. El-cheapo graphics cards that can run Aero are about $25, bringing our total to just under $35.
I agree that the lion's share of money generally goes to the lawyers (and sometimes the original class-representative litigants), but for the past decade or so in the U.S. after some reforms, most settlements are in cash, not vouchers or coupons. They are, however, usually not a large amount of cash. For example, I got a check for about $11 from some California state class-action lawsuit against record labels for price-fixing; the Grand Theft Auto settlement gave purchasers are a choice of a replacement disc or $5-35 cash depending on what damages they asserted and/or had documentation of; and a recent iPod settlement will pay out $15-25 cash.
It was Dell inducing me to purchase a computer using "Vista ready" in their advertising, not Microsoft. Now Microsoft might have duped them by having poor specs in their sticker program, but that's a matter between Dell and Microsoft---the matter for me is that Dell sold me a computer with a false advertising claim on it.
Having monopoly control over, say, food, means that there is no "lack of duress caused by the other party". The corporation has gotten itself ownership of all food, so now anyone who wants to eat must pay it.
My point is precisely that in monopoly situations such as that, it is not the case that "other options [are] available".
In the Dell example, I wish to buy a PC from Dell, which they would like to sell me. Microsoft is not involved in this transaction, as it does not involve their products. However, they exercise their monopoly power to forbid Dell from doing business me me, unless they force me to purchase a Microsoft product---i.e. force me to sign a separate contract with someone who was not party to the original proposed contract.
In short, I generally take Thomas Jefferson's view that the goal is individual liberty, and the means depends on the situation---which is why Jefferson was generally opposed both to powerful governments and to large corporations.
I don't generally use "searching the internet" as a way to discover absolute truth. Rather, I use it a a way to get a sense of what information is out there; a quick introduction to a subject; a tutorial overview of something; or pointers to further reading.
Before Wikipedia, this more often than not resulted in me being at some random guy's geocities site. Sometimes these were good, sometimes not. You would apply your mental "is this guy a total crank?" filter, cross-reference a few such sites, and often get a decent view of the subject.
I see Wikipedia as basically a much better version of that. You've forced the 10 people with WW2 geocities sites to all get together and bang out a consensus set of WW2 articles, with citations to where they got their info from. Some articles are still total crap, but I can apply my same mental "was this article written by a total crank?" filter---you can often tell from the style and tone whether the article was written by someone with a chip on their shoulder. From the rest, I can get a decent overview of some version of consensus on the subject and pointers to future reading.
There's a sweet spot between 1-to-4-threaded apps (which x86 does very well) and trivially parallelizable apps (which x86 clusters do very well), where you need a ton of CPUs/cores in the same box, with a whole shitload of RAM and fast interconnects. That's the only real niche where non-x86 stuff has survived, because at the moment you can't really get a 64-way x86 box with a few hundred gigs of RAM. So it's what's being pushed by POWER6, the remaining SPARC lines, and Cray (which has survived basically by transforming from a vector-CPU designer into a designer of interconnects).
To some extent this class of architectures has always had that strength---when I read old OS benchmark papers, I'm sometimes surprised by how, even though the CPUs are obviously dated (quoting e.g. 300 MHz clock speeds), the rest of the specs are vaguely modern (that 300 MHz UltraSPARC might have been in an 8-way configuration with a gig of RAM).
But they used to also be able to sell cut-down versions as more normal servers and workstations---a 1-CPU or 2-CPU UltraSPARC was a perfectly reasonable server or workstation. The x86 commodity hardware has totally destroyed that market, which kills economies of scale for these non-x86 architectures, since now they have only the much smaller highly-SMP market.
Slashdot Mirror
Unless your job is legal compliance or you own a lot of Microsoft stock, why bother with this?
Wouldn't having a degree from Ray Kurzweil's "Singularity University" make you seem kind of like an AI kook?
The U.S. is coming closer to pulling its weight on new capacity: in 2008, it installed 8.3 GW, while Germany installed 1.7 GW, or about a 5x factor. Not quite the 30x factor of land area, but hey, 1.7 million of those sq km are in Alaska, which is kind of inconvenient for electric transmission (same reason Canada's wind power is fairly low, despite massive land area).
For example, there was a bunch of excited speculation about when China's GDP would surpass the U.S.'s, despite the fact that that would still leave China nowhere near the U.S. on a per-capita GDP basis.
A huge proportion of censorship of, for example, literary works, was done by elected governments---Ireland's government, which has been democratic since its independence, is pretty notorious for their treatment of James Joyce's works, for example. That doesn't make it not government censorship.
Just because various countries are sovereign states doesn't mean it's inappropriate to criticize them if they're run by hard-right authoritarians (or hard-left authoritarians, for that matter).
With how often Facebook changes their API and makes subtle, app-breaking changes to how existing stuff operates, you won't really be able to rely on anything in a published book.
The 965 chipset's MTRRs are still broken, after a dozen firmware revisions, but nobody at Intel cares because it only breaks 64-bit Linux with >=4gb of RAM.
While the local channels are likely to face technical problems by a delay, the networks themselves seem to be part of the group lobbying for a delay, hoping the switchover can take place after their seasons end.
The stuff he's talking about is fairly generic data collection and analysis software, not really physics-specific. LabView is a semi-standard piece of software for interfacing with laboratory machines, setting up experiments and reading in the data and storing it somewhere. Igor is analysis and graphing software. Inventor is a 3d modeling/CAD package. Eagle is an ECAD (CAD for circuits basically) package with some automated layout capabilities.
Some of these are easier to replace than others. Data analysis can generally be done by a mixture of R and Octave, as you note; and for just graphing/plotting/figures, there's a billion open-source alternatives, including a ton of Python libraries these days.
LabView is going to be trickier to replace, due to interoperability: there's a lot of machines out there that interface with LabView out of the box, exposing APIs that LabView can use to control experiments, and sending back data in formats LabView can deal with. I'm not aware of even a significant attempt to replace it, really. (This is partly because what it does is really boring.)
I don't know what open source alternatives for CAD are; that's a whole topic in itself probably.
I don't own any way of hooking up my phone to a computer (I don't think my phone even has a USB connection), I don't know or have any interest in learning how to configure Linux to talk to a USB-cell-phone modem, etc.
...mainly because I expect to see some Representative or Senator, with no hint of irony or sense of shame, offering an unrelated amendment to that bill
The creation of the coupon program was largely the result of industry pressure---they're worried that some number of people who use only over-the-air TV will disappear after their current equipment stops working, and never return, thereby harming viewership numbers. This was a particular concern of networks like Univision.
We're talking about a copyright license on the text of a scientific paper here. The normal status is "all rights reserved". Cc-by-nd is strictly more permissive than the default copyright. The restrictions on commercialization, public performance, etc., are all part of the copyright license on the text of the paper, nothing more. The ideas in the paper are protected, if at all, by patent law.
Evaluation committees don't look only at the number of lines in your CV, but at how frequently your papers are cited. Papers in closed-access journals are much less frequently cited than papers in open-access journals. At the very least, if you publish in closed-access journals and don't want to shoot yourself in the foot, you should put up a PDF on your website so potential citers can actually get your article rather than just shrugging and giving up on it (sometimes this is permitted, sometimes officially illegal but rarely enforced).
I agree it's one of many factors, but it's been the deciding factor for me several times---if there are multiple journals roughly in the same class, the open-access one is better for my career, because it results in actually publishing my paper rather than locking it away in a box.
People noticed some decades ago that failure analyses seemed to predict failures with much less frequency than failures actually occurred. Why? Because the failure analyses themselves have some frequency with which they're wrong (rest on incorrect assumptions, themselves contain errors, etc.). Furthermore, failures can come not only from poor design, but poor implementation, poor initial data (e.g. the site wasn't what you thought it was like), etc. If you want accurate estimates of the overall likelihood of failure, as you generally do, modern failure analysis tends to take an empirical, whole-system view, in which no component of the process is assumed a priori to have a zero error rate.
You don't have to imagine futuristic AI to think of computationally intensive stuff that's in the process of being pushed to the client. Excel is getting machine-learning and data-mining algorithms added to it, for example, so in future versions you'll be able to not only plot your data, but plot kernel-smoothed versions, cluster it using various clustering algorithms, fill in incomplete labels using, say, SVMs, and so on. A lot of the stuff that statisticians and savvy business analysts do in R now will be things that are standard and expected for a much wider range of people in the future (and R itself is getting more uptake outside of statistics professors).
Maybe Google will offer a server-side version of that sort of thing for Google Docs, but that would require a lot of processing power on their end.
Firefox includes all sorts of "security" stuff turned on by default, some of it both pointless and really annoying, like the click-4-times annoyance when you want to visit any https site that doesn't have its SSL certificate signed by one of the worthless central authorities. Some of it is also useful, like popup blocking. "Clickjacking" prevention seems like it'd go in the same category of stuff, if Firefox were interested.
When you put things on YouTube, you aren't getting paid
They are, actually---Google splits the ad revenue from ads shown on video pages with content providers.
Perhaps you have't looked lately but DIMMs and el-cheapo graphics cards are both around "free" these days. As in, a 512-mb DIMM costs less than a sandwich. El-cheapo graphics cards that can run Aero are about $25, bringing our total to just under $35.
I agree that the lion's share of money generally goes to the lawyers (and sometimes the original class-representative litigants), but for the past decade or so in the U.S. after some reforms, most settlements are in cash, not vouchers or coupons. They are, however, usually not a large amount of cash. For example, I got a check for about $11 from some California state class-action lawsuit against record labels for price-fixing; the Grand Theft Auto settlement gave purchasers are a choice of a replacement disc or $5-35 cash depending on what damages they asserted and/or had documentation of; and a recent iPod settlement will pay out $15-25 cash.
It was Dell inducing me to purchase a computer using "Vista ready" in their advertising, not Microsoft. Now Microsoft might have duped them by having poor specs in their sticker program, but that's a matter between Dell and Microsoft---the matter for me is that Dell sold me a computer with a false advertising claim on it.
Having monopoly control over, say, food, means that there is no "lack of duress caused by the other party". The corporation has gotten itself ownership of all food, so now anyone who wants to eat must pay it.
My point is precisely that in monopoly situations such as that, it is not the case that "other options [are] available".
In the Dell example, I wish to buy a PC from Dell, which they would like to sell me. Microsoft is not involved in this transaction, as it does not involve their products. However, they exercise their monopoly power to forbid Dell from doing business me me, unless they force me to purchase a Microsoft product---i.e. force me to sign a separate contract with someone who was not party to the original proposed contract.
In short, I generally take Thomas Jefferson's view that the goal is individual liberty, and the means depends on the situation---which is why Jefferson was generally opposed both to powerful governments and to large corporations.
I don't generally use "searching the internet" as a way to discover absolute truth. Rather, I use it a a way to get a sense of what information is out there; a quick introduction to a subject; a tutorial overview of something; or pointers to further reading.
Before Wikipedia, this more often than not resulted in me being at some random guy's geocities site. Sometimes these were good, sometimes not. You would apply your mental "is this guy a total crank?" filter, cross-reference a few such sites, and often get a decent view of the subject.
I see Wikipedia as basically a much better version of that. You've forced the 10 people with WW2 geocities sites to all get together and bang out a consensus set of WW2 articles, with citations to where they got their info from. Some articles are still total crap, but I can apply my same mental "was this article written by a total crank?" filter---you can often tell from the style and tone whether the article was written by someone with a chip on their shoulder. From the rest, I can get a decent overview of some version of consensus on the subject and pointers to future reading.
There's a sweet spot between 1-to-4-threaded apps (which x86 does very well) and trivially parallelizable apps (which x86 clusters do very well), where you need a ton of CPUs/cores in the same box, with a whole shitload of RAM and fast interconnects. That's the only real niche where non-x86 stuff has survived, because at the moment you can't really get a 64-way x86 box with a few hundred gigs of RAM. So it's what's being pushed by POWER6, the remaining SPARC lines, and Cray (which has survived basically by transforming from a vector-CPU designer into a designer of interconnects).
To some extent this class of architectures has always had that strength---when I read old OS benchmark papers, I'm sometimes surprised by how, even though the CPUs are obviously dated (quoting e.g. 300 MHz clock speeds), the rest of the specs are vaguely modern (that 300 MHz UltraSPARC might have been in an 8-way configuration with a gig of RAM).
But they used to also be able to sell cut-down versions as more normal servers and workstations---a 1-CPU or 2-CPU UltraSPARC was a perfectly reasonable server or workstation. The x86 commodity hardware has totally destroyed that market, which kills economies of scale for these non-x86 architectures, since now they have only the much smaller highly-SMP market.