1) This was in 1974.. for how long a period was this considered a serious threat? One year? Two? Global warming has been considered a threat for more than 10 years.
2) They say
Just what causes the onset of major and minor ice ages remains a mystery. "Our knowledge of the mechanisms of climatic change is at least as fragmentary as our data," concedes the National Academy of Sciences report. "Not only are the basic scientific questions largely unanswered, but in many cases we do not yet know enough to pose the key questions."
This is completely different from today's situation. We have better theory and models and much more data. From wikipedia:
Although there was a cooling trend then, it should be realised that climate scientists were perfectly well aware that predictions based on this trend was not possible - because the trend was poorly studied and not understood (for example:[5]). However in the popular press the possibility of cooling was reported generally without the caveats present in the scientific reports.
All in all, I think the current evidence for global warming is quite strong. The evidence for cooling was not so strong although some of the models used then, such as aerosols reducing the amount of solar energy absorbed, are relavant now: The amount of aerosols present in the atmosphere has declined cosniderably since the 70s, thus forcing temperatures up even more.
Why does he say 'it is water vapor rather than caron dioxide that causes warming in europe'? The next sentence makes it clear that this is not the case, so the first sentence should have been omitted as it is misleading.
Meanwhile Linux, noted Russinovich, owes a great deal to the work of Andrew Tanenbaum, who created the Unix-like Minix operating system for educational purposes. Although Linux creator Linux Torvalds
readily admits that he based his work on Minix, both he and Tanenbaum refute claims that Torvalds borrowed more than he admitted.
The link to 'readily admits' points to another ZDNet article which says nothing of the kind. I take it that the AdT institute's FUD is spreading rapidly for some reason. People have to understand that just because someone spreads FUD, that does not turn an undisputed fact into a contested issue. Jesus.
I did read and I was not sure I understood your point, so I replied and you have now clarified. You had said:
> How it works has been described, but to employ that method oneself requires crawling the web the same way that Google does.
Valid, point. However, if (as the OP) you are talking about an entity that *could* crawl the entire web, then it is not applicable.
> There is no way to just _get_ a page's pagerank, as ranked by Google, without crawling the entire web yourself, or resorting to methods that are quite closely guarded secrets.
I was a bit intrigued by this sentence's meaning, especially the phrase 'closely guarded secrets'. That's why I said 'hardly a secret'.
How to crawl the web is hardly a secret, so I don't understand your point.
Google does use some interesting methods to crawl the web, but it is just a matter of managing resources, so that only sites that are updated frequently are crawled frequently.
In any case, if you want to have an algorithm that works, you can check out the Pagerank algorithm (without added safeguards for people that try to cheat, but those are just heuristics that are easy to add. You just have to fine tune them). There are other algorithms that you could use, which might be better than pagerank [depending on your purpose]. For example you could have an algorithm that weighs up incoming links, outgoing links and keywords in different ways, depending on what you think is most important.
However, you will have to fine tune the parameters of the algorithm for your particular application. If your application is searching the entire web and avoiding 'cheaters', then your parameters will be quite different from searching your database of documents.
This is the case for many algorithms that are well known. It is hardly a secret how neural networks work. However, even with source that implements one, you usually have to know a couple of facts related to their convergence, either from theory or from experimentation, in order to use them properly.
Which goes back to this: If google were to give you the source code for everything, how could you possibly expect to run it? If you would like to have something similar, you are encouraged to look at the Clever (IBM's similar algo) and Pagerank papers, implement the algorithms, play around with them in some small database (like your own website) and then try to apply them to the problem at hand until you get a satisfactory performance.
Even if it is closed source, you can hack it. People have been doing this for ages and I am sure that right now it is trivial to find any number of cracks that remove this Photoshop restriction.
Recently some guys managed to prove that there exists an infinite number of arithmetic progressions of prime numbers of any length. So, it is not only true for p, p+2.. but true for (p, p+N), and also for (p, p+k,..., p+k*N)..
In setting out to prove that there are an infinite number of arithmetic progressions of prime numbers with four terms, two mathematicians appear to have proved the result for prime progressions of all lengths.
Recent publications (I think in Science) blur significantly the distinction between actual and imagined visual input. I don't remember the names of the areas involved, but the results indicated that the part of the visual cortex that was initially thought to be only activated by the retina, showed visual like activity when subjects where dreaming.
Ideas deployment is usually incremental - and the ideas that are newly deployed are pretty obvious.
Less obvious ideas are discussed in the public domain long before they are introduced as products. I like to give ADSL as an example - the original idea was that you have to split your band into many (ideally infinite) subbands to send information. The problem was that until the development of fast DSPs it was impossible to do it - and it was certainly impossible to do it when the idea was first mentioned.. since it was the days of analog filtering.
Nevertheless, there are numerous patents on ADSL, which are about very specific parts of the overall ITU-T standard rather than the idea of splitting the band in subbands to send information..
Furthermore, an individual has little or no protection against predatory companies that want to steal his ideas even with the patent system in place. The reason that a patent costs much more than an individual can possibly afford. What people do, is they give the idea to a patent attorney/firm, who agrees to pay for the upfront cost in return for a percentage on any profit made from the patent.
Sure, as long as your A/D converter has a known transfer function. Most of the time the transistors in the A/D are not matched perfectly, so some quanta might have a different size from others. There is no way to correct for that.
CCDs are great, but they still have a bandwidth limit. You cannot expose them to any degree of light and expect them to store it properly. The same happens with audio A/D converters, they do actually have a limited range.. and thus exhibit some distortion. Of course the solution is to scale the input using high-quality low-noise analogue equipment to the range where the digital equipment shows no distortion.
I do not agree with other posters below saying that analogue media has 'higher bandwidth'. This is simply not true. However, it is true that digital has a sharp bandwith/dynamic range cut-off, while analogue's cut-offs are much smoother.
>SCO tried going off on yet another FUD tangent >(presumeably since as of today, their legal case > should cease to exist) by pointing the finger at > various Linux-related companies, asking why, if so > confident in SCO's lack of a case, they didn't > offer to indemnify their clients.
Yeah, and today McBride said that since they are idemnifying their costumers, they must know there is a problem with Linux:
"We believe Novell's indemnification announcement is significant for a couple of reasons. By announcing the program they are acknowledging the problems with Linux. Through the restrictions and the limitations on the program, they are showing their unwillingness to bet very much on their position."
hehe - as you can see, it is possible to spin either action the way SCO likes.
How many samples do you usually have to take to make a good approximation? Imagine the case where you have a highly reflective small object, like for example a watch that shines reflected light onto a small circle somewhere far away. Hm.. I guess one could just sample a few times only on all objects, but in a complex scene something like that can be completely lost.
Also, it all must get quite a bit more complicated when the medium itself is diffracting. In this case you'd need to sample on all possible directions and ranges around your point. Not that any other methods could give you a better solution in that case, mind you.
Perhaps, but what radiosity achieves is to model the diffracted light rays from other surfaces. For example, suppose you have a red-coloured surface next to a white one. The red colour normally spills on to the white surface because some of the light diffracted from the red surface reaches the white one. This is quite hard to model accurately with ray-tracing because for every point that you trace back to a surface you must not only calculate the effect of the light coming directly from the light source itself, but from all other points in all other surfaces, because they all transmit live passively. So, it can't work.
I took some nice shots this summer. I gave the negatives to a shop. When I got the pictures, the colour was all wrong. Why? Because what they do is they *scan* the negatives and then print them out. However, someone had changed the colour balance in their image processing software. They had also put the digital scans on a CD and I could definitely see the difference with the printed material.
So, my question is: What do photo shops do these days? Do they all scan the negatives and print them out? How can you tell the difference? At least I hope they scan them at a relatively high resolution. Which brings me to an interesting point: Can I scan the negatives my self? How do I turn change them to true colour? (I guess just inverting the colour will not do... it must depend on the film, right?)
Or what have they ever had known about any kind of technology? I know around half a dozen people that are economists, one of them a uni professor, and none of them exhibit any understanding of technology. Here is my question to the prof:
If it costs X to produce the main branch of the code, how much does it cost to fork it N times? The upper limit would be NX, but actually it should be much less. Furthermore, what is the utility of the main branch? It is true that the utility of the main branch, or any fork, might be the same for just *one* customer, but what when there are many customers which want different things?
Furthermore, what about closed source software? With closed-source, each client will have a completely customised version of the software. If one of the forks for one client gets a fix/upgrade, the fork for another client will not necessarily get it. Plus, it is much harder for to migrate. (If something is open-source, it would be easy to write a migration application).
Hm.. I remember I had used to have stability problems with my nvidia card and the VIA K133 motherboard with an Athlon CPU some time ago, which required to use mem=nopentium as a kernel argument. Do I still have to do this in the newest kernel releases?
Slashdot Mirror
1) This was in 1974.. for how long a period was this considered a serious threat? One year? Two? Global warming has been considered a threat for more than 10 years.
2) They say
This is completely different from today's situation. We have better theory and models and much more data. From wikipedia:
All in all, I think the current evidence for global warming is quite strong. The evidence for cooling was not so strong although some of the models used then, such as aerosols reducing the amount of solar energy absorbed, are relavant now: The amount of aerosols present in the atmosphere has declined cosniderably since the 70s, thus forcing temperatures up even more.
Check also
http://www.wmconnolley.org.uk/sci/iceage/wcc-1979
I'd like to know which ones were those.
Why does he say 'it is water vapor rather than caron dioxide that causes warming in europe'? The next sentence makes it clear that this is not the case, so the first sentence should have been omitted as it is misleading.
Interestingly, all the report's paragraphs are marked as UNCLASSIFIED so I cannot see why they were redacted in the first place.
My opinion is that this is strikingly similar to the case of the DMCA and printer cartridge 'encryption'
considering that most people have never even learnt how to bloody whisper on the blody phone, I cannot expect them to keep bloody quiet!
In every school I ever went to, boys and girls were equally BAD.
> replace them with legal codecs...
How can a codec be illegal?
The article says, and I quote:
The link to 'readily admits' points to another ZDNet article which says nothing of the kind. I take it that the AdT institute's FUD is spreading rapidly for some reason. People have to understand that just because someone spreads FUD, that does not turn an undisputed fact into a contested issue. Jesus.
I did read and I was not sure I understood your point, so I replied and you have now clarified. You had said:
> How it works has been described, but to employ that method oneself requires crawling the web the same way that Google does.
Valid, point. However, if (as the OP) you are talking about an entity that *could* crawl the entire web, then it is not applicable.
> There is no way to just _get_ a page's pagerank, as ranked by Google, without crawling the entire web yourself, or resorting to methods that are quite closely guarded secrets.
I was a bit intrigued by this sentence's meaning, especially the phrase 'closely guarded secrets'. That's why I said 'hardly a secret'.
How to crawl the web is hardly a secret, so I don't understand your point.
Google does use some interesting methods to crawl the web, but it is just a matter of managing resources, so that only sites that are updated frequently are crawled frequently.
In any case, if you want to have an algorithm that works, you can check out the Pagerank algorithm (without added safeguards for people that try to cheat, but those are just heuristics that are easy to add. You just have to fine tune them). There are other algorithms that you could use, which might be better than pagerank [depending on your purpose]. For example you could have an algorithm that weighs up incoming links, outgoing links and keywords in different ways, depending on what you think is most important.
However, you will have to fine tune the parameters of the algorithm for your particular application. If your application is searching the entire web and avoiding 'cheaters', then your parameters will be quite different from searching your database of documents.
This is the case for many algorithms that are well known. It is hardly a secret how neural networks work. However, even with source that implements one, you usually have to know a couple of facts related to their convergence, either from theory or from experimentation, in order to use them properly.
Which goes back to this: If google were to give you the source code for everything, how could you possibly expect to run it? If you would like to have something similar, you are encouraged to look at the Clever (IBM's similar algo) and Pagerank papers, implement the algorithms, play around with them in some small database (like your own website) and then try to apply them to the problem at hand until you get a satisfactory performance.
Pagerank has been described before. It's not a secret. Possibly they are using a fine-tuned version of it, but still...
Even if it is closed source, you can hack it. People have been doing this for ages and I am sure that right now it is trivial to find any number of cracks that remove this Photoshop restriction.
A summary of the article appeared in science. The research article is currently under review. but there is a preprint available on arXiv, and also a nice image that shows the result graphically.
Recent publications (I think in Science) blur significantly the distinction between actual and imagined visual input. I don't remember the names of the areas involved, but the results indicated that the part of the visual cortex that was initially thought to be only activated by the retina, showed visual like activity when subjects where dreaming.
Ideas deployment is usually incremental - and the ideas that are newly deployed are pretty obvious.
Less obvious ideas are discussed in the public domain long before they are introduced as products. I like to give ADSL as an example - the original idea was that you have to split your band into many (ideally infinite) subbands to send information. The problem was that until the development of fast DSPs it was impossible to do it - and it was certainly impossible to do it when the idea was first mentioned.. since it was the days of analog filtering.
Nevertheless, there are numerous patents on ADSL, which are about very specific parts of the overall ITU-T standard rather than the idea of splitting the band in subbands to send information..
Furthermore, an individual has little or no protection against predatory companies that want to steal his ideas even with the patent system in place. The reason that a patent costs much more than an individual can possibly afford. What people do, is they give the idea to a patent attorney/firm, who agrees to pay for the upfront cost in return for a percentage on any profit made from the patent.
In what way is Microsoft Office an invention? What is new/original/inventive about it?
Sure, as long as your A/D converter has a known transfer function. Most of the time the transistors in the A/D are not matched perfectly, so some quanta might have a different size from others. There is no way to correct for that.
CCDs are great, but they still have a bandwidth limit. You cannot expose them to any degree of light and expect them to store it properly. The same happens with audio A/D converters, they do actually have a limited range.. and thus exhibit some distortion. Of course the solution is to scale the input using high-quality low-noise analogue equipment to the range where the digital equipment shows no distortion.
I do not agree with other posters below saying that analogue media has 'higher bandwidth'. This is simply not true. However, it is true that digital has a sharp bandwith/dynamic range cut-off, while analogue's cut-offs are much smoother.
>SCO tried going off on yet another FUD tangent >(presumeably since as of today, their legal case
> should cease to exist) by pointing the finger at
> various Linux-related companies, asking why, if so
> confident in SCO's lack of a case, they didn't
> offer to indemnify their clients.
Yeah, and today McBride said that since they are idemnifying their costumers, they must know there is a problem with Linux:
"We believe Novell's indemnification announcement is significant for a couple of reasons. By announcing the program they are acknowledging the problems with Linux. Through the restrictions and the limitations on the program, they are showing their unwillingness to bet very much on their position."
hehe - as you can see, it is possible to spin either action the way SCO likes.
How many samples do you usually have to take to make a good approximation? Imagine the case where you have a highly reflective small object, like for example a watch that shines reflected light onto a small circle somewhere far away. Hm.. I guess one could just sample a few times only on all objects, but in a complex scene something like that can be completely lost.
Also, it all must get quite a bit more complicated when the medium itself is diffracting. In this case you'd need to sample on all possible directions and ranges around your point. Not that any other methods could give you a better solution in that case, mind you.
Perhaps, but what radiosity achieves is to model the diffracted light rays from other surfaces. For example, suppose you have a red-coloured surface next to a white one. The red colour normally spills on to the white surface because some of the light diffracted from the red surface reaches the white one. This is quite hard to model accurately with ray-tracing because for every point that you trace back to a surface you must not only calculate the effect of the light coming directly from the light source itself, but from all other points in all other surfaces, because they all transmit live passively. So, it can't work.
I took some nice shots this summer. I gave the negatives to a shop. When I got the pictures, the colour was all wrong. Why? Because what they do is they *scan* the negatives and then print them out. However, someone had changed the colour balance in their image processing software. They had also put the digital scans on a CD and I could definitely see the difference with the printed material.
So, my question is: What do photo shops do these days? Do they all scan the negatives and print them out? How can you tell the difference? At least I hope they scan them at a relatively high resolution. Which brings me to an interesting point: Can I scan the negatives my self? How do I turn change them to true colour? (I guess just inverting the colour will not do... it must depend on the film, right?)
Or what have they ever had known about any kind of technology? I know around half a dozen people that are economists, one of them a uni professor, and none of them exhibit any understanding of technology. Here is my question to the prof:
If it costs X to produce the main branch of the code, how much does it cost to fork it N times? The upper limit would be NX, but actually it should be much less. Furthermore, what is the utility of the main branch? It is true that the utility of the main branch, or any fork, might be the same for just *one* customer, but what when there are many customers which want different things?
Furthermore, what about closed source software? With closed-source, each client will have a completely customised version of the software. If one of the forks for one client gets a fix/upgrade, the fork for another client will not necessarily get it. Plus, it is much harder for to migrate. (If something is open-source, it would be easy to write a migration application).
Hm.. I remember I had used to have stability problems with my nvidia card and the VIA K133 motherboard with an Athlon CPU some time ago, which required to use mem=nopentium as a kernel argument. Do I still have to do this in the newest kernel releases?
I.ve burnt my motherboard (K7) because the fan had stopped. The CPU (Athlon 900Mhz) survived unscathed.