You raise very good points, and it has been a pleasure to keep this thread going. Thanks for helping me dismember this dead horse:-) (I think we went past beating it about three posts ago). Hope to read more from you in the future. At this point, I think we've reached a good level of agreement. Thanks for the insight. --
You know, you miss the point in so many levels, it is hard to adress it, really.
I am Mexican. Mexico is a developing 3rd world country, and a lot of the more promising government projects are software industry development based, like this Project to give a boost to multimedia and animation industry in Jalisco (spanish only, sorry). Just what do you think we are? a bunch of freaking retards? There IS industry and education in here! It's not all about helping out the extremely poor, it's about helping society and industry develop so there is more cash flow, education and yes, more programs to help those in need. That said, open source initiatives DO help. A lot. And a bunch of the people involved in open source projects are from third world countries (incidentally, Miguel de Icaza is Mexican).
Please note, too, that foreign backing is not all there is to charity. A country has to develop its own strenghts and help itself. I think we're on our way there. Politics and economic situations are a complicated thing, and broad statements like the ones you make are usually off the mark. --
I'd have to wonder why humans still drive cars, fly airplanes and write algorithms.
Because computers still can't do it. If you think you can drive better than a computer you should also try to run a semaphore network or perform one million mathematical operations in less that a minute. Accept it: Creativity (and writing algorithms) may be the realm of people, but determinism is something a bunch of chips with software are much better at.
As a programmer, I assure you, I can write algorithms that show bias. Of course, you are correct that none of them think.
As another programmer, I assure you, that would greatly surprise me. Of course you can write a program that selects one answer over the other for arbitrary reasons! and you know what? That's exactly what it will do! It won't decide to go the other way just because it feels like it. You can even put in a random decision, and it will take a random decision! Don't you think an analysis of the algorithm would reveal the intended outcome? That was the original point of this discussion. I'm not advocating the rise of the machines. But I sure couldn't be trusted with an image and expected to come up with a totally unbiased, objective enhancement of it. Good or bad, an algorithm that doesn't try to replace the original content of the image and instead works with it will do a more reliable, repeatable job than I, or anyone else for that matter, can do (please note again that I'm not talking about a better enhancement, rather a more objective one! (which is what would be needed to bring this in front of a jury) --
programs often work better on one set of data than another... In this way they can be just as random as poor "human subjectivity".
No, they can't. An algorithm is not random. It is a process applied to data. This is just bullshit. Of course some algorithms will work better or worse on a particular set of data. If you can prove to a jury that the algorithm is not pulling an image out of its ass (which a human may), it is a valid point.
To think that a process is better just because it is wholly deterministic is silly
I'm not saying it is better in a general sense. I'm saying it is more trustworthy, repeatable and does not introduce external variables, which for this particular application, is better. An algorithm does not say "this line would look much better if I stretched it this way" whereas a human may. The algorithm does not care if the resulting image looks like shit.
Interesting. "proving" something with data? Either that wasn't science or verificationism must not be dead after all.
I think you're misusing the word, and Data Mining CAN prove that there exists a correlation or not in a set of data. For instance: a scientist may look at data for the foci of a disease in a country and see that there is roughly a correlation between locations of nuclear plants and the disease sprouts. He may then propose that there is a causal relationship. But after running the raw data through a data mining program, a much higher probability relationship may be found between temperature+water concentrations+disease (these three-tiered relationships are difficult for humans to find), showing that the former, seemingly correct deduction was just coincidental in nature.
Regardless, different data mining techniques would each give a different "score" would they not? What should that difference be called if not a "bias"?
Different data mining techniques are just different ways of analyzing a data set. They do not show statistically different results (they ALL agree on how much of each of the original data there is). Some of them are just better at finding certain kinds of patterns than others. It largely depends on the format of your raw data. If you have a point, make it. Processes don't think. They are not biased! is that clear? --
How does that work? you two go around pissing people off? =) --
Re:could be safer than rolercoaster
on
Robocoaster
·
· Score: 2
You forget some advantages, like the possibility of changing the ride's path whenever you want, simply by feeding the robot different commands. Try that with a rollercoaster.
You have a point about the group experience, but this thing could certainly be used in conjunction with a spherical screen and digitally generated animation to make one hell of an entertaining ride. All in all, I think the flexibility of the framework makes up for the differences in experience. Maybe, as you say, it won't end up competing with roller coasters, but I think it will make great rides with the right add-ons. --
This is offtopic, but I took a look at your posting history and noticed that you started at 0 and just recently started getting Score:1 on your posts. How was that possible? I thought every logged in user had a default 1 score and 0 was for AC's. Could it be because you used "Troll" in your username? or was there a change in the Slashdot code to modify this behavior? anyone? --
In the book "Airframe" by Michael Crichton there is one part where they have really blurred footage of an accident aboard a plane and the main character requests an enhanced version of it. The technician suggests some techniques that would be based on taking educated guesses, but she refuses and tells him to only perform totally computer based enhancements to it. They end up running the frames at high speed and taking "snapshots" of the composite result, which greatly enhances the details.
So, what's that got to do with anything? The use of enhancing techniques CAN be abused. By selectively enhancing a section of a print against another one you can manipulate the outcome. It is very different to increase visibility on the whole print than to darken some parts, blur others. If there is a human element, it is not to be trusted. Allow me to explain (and sorry for the long rant):
Some time ago I attended a Data Mining Seminar. Many people here will be familiar with the techniques that are used in that area. What's the point of using a computer algorithm to find patterns when we are so good at it naturally, you ask? Why, exactly that. Humans are extremely good at finding patterns. We even tend to see them where they're not, and ignore the extra evidence that may point in different directions. A computer works on the data and does not have prejudice for/against it. Before the AC's jump and say that algorithms can be manipulated too, let me just say that they can be audited for soundness and logical mistakes/mathematically analized, etc. They are much more reliable than a human. There have been numerous instances when a correlation between factors was suspected, and data mining was used to prove/disprove the correlation or give a score to it.
So what's the right way to go here, you ask? I think that an image you can get from running an algorithm against the source is valid and can be considered an objective derivative of the original. Tools that allow to selectively retouch pictures should be out of the question. There are many techniques that can be applied to images to enhance them and that are wholly based on what's already there. --
I took a look at the source code for this page to see what he was talking about and I know the guy sent his very own file nicely formatted, but you guys should at least have removed the <html><head> and <body> tags from his document. It is extremely bad form to insert a whole html document inside another one. I don't think this page renders well in most browsers. Maybe the Slashdot Editors can update the story removing that from the source code? --
Thanks for the insight. That was really helpful. Hope your recovery process goes great. I can understand your point about the routines. I used to be a smoker. The hardest part of not smoking anymore (after the physical need goes away) tends to be dissasociating certain activities (partying, having some coffee, being outside on the cold) with smoking. Well, thanks for a great post:-) --
Yep. Are you going to listen to the heroin addict about the virtues of his vice? Never trust an addict.
Well, you can hear all about the non-virtues of the vice by the people not involved in it, right? If you want to actually know why people do it, you have to hear the opinion of the people doing it. After all, there must be more to heroine that shitting your pants constantly and screwing your life, or people wouldn't be so keen on using it, right? (have you seen trainspotting?). Same thing with all vices. True, the opinion of the addict is biased. But the opinion of the non-addict is also uninformed. --
Good point. I think this would be akin to road billboard ads trying to pass themselves as road signs. That would surely get them in trouble real fast. Maybe computerdom should be viewed in the same light? --
"Put your hand on a hot stove for a minute, and it seems like an hour. Sit next to a pretty girl for an hour and it seems like a minute. Now that's relativity!"
On the same vein: the passing of time is relative depending on which side of the bathroom door you are. --
Icing on the mornings cake: I got up on time, drove to work, posted a story, and then forgot to press *save* on the goddamn web form. So for hte next 2 hours I keep deleting submission after submission about Mozilla 1.2.1 thinking "geezus, are people blind?" and not realizing that no, I am in fact stupid. Of course, why so many people submit a bug fix release of a web browser is beyond me. Some stories I'd rather not post, but sheer volume of submissions really makes it impractical to ignore them --
A gaming console works on a TV, which roughly has a resolution of 640x483. My really old Matrox G200 (gathering dust now) could deliver decent performance for a wide host of games at that resolution.
A PC system tweaked for gaming can deliver astounding frame rates at 1600x1200. That is the main difference. Try to hook up your 17" monitor to a console (can't be done, but would be nice) and you would soon see the limitations of a console. --
And why would they need 25 BILLIONS to do something like that? that's the whole point of the "fable". Where does all this money go? Does NASA have full disclosure of expenses? Why are independent organizations able to achieve pretty impressive results with a fraction of the cost of NASA? Furthermore, why is it accepted that this agency should spend BILLIONS (that's thousands of millions of dollars for the ones that use the more common measurements) in order to get results they were able to get 20+ years ago? I'm not in the US, but if I were I'd sure be pissed off about this. Where's the space travel package in my travel agency yet? --
Hydra would be a nice name too. And it is sort of a dragon (many headed monster/serpent)... Kind of goes along with the naming scheme (Chimera, Mozilla..) --
I haven't used Photoshop much, but I believe that the way to draw primitives is the same as the one you describe in the Gimp. Paint Shop Pro, which added vector layers in version 6, does have primitives (rectangles, ellipses and lines). Antialiased vectors, too. After a while you get used to doing things the Gimp way. My only complaint is that vector shapes do make it much faster to prototype designs for websites and such (with big, flat spaces and basic figures in them).
The makers of Gimp don't seem to be interested in adding this functionality, though, so we'll have to turn to Zope or Kontour for that functionality --
Very true. Some time ago I switched from using Paint Shop Pro (Photoshop-like interface) to the Gimp full time. It was daunting at first, but if you know what you're doing with one program, it doesn't take much time to find out how to do in the other. After a while, I had a hard time using PSP again, and kept going for the menu-over-image approach constantly (not getting anywhere).
Layers and composite color schemes are some of the most difficult things to master in these programs and a mayor reason why people find them hard to use ("I pushed the pencil button and this thing doesn't draw anything!!" while painting over an invisible layer is a common complaint). AutoCad and 3d Studio are very hard to learn too.
Open source interfaces tend to be designed for functionality rather than mass appeal, which is not a bad thing. Things like blender, vi and the Gimp can be a pain in the ass the first time you try them, but over time you really find out the reasoning behind the interface functionality (there usually are very good reasons to do things one way or the other, not just accumulated cruft like a lot of commercial interfaces). --
Dreamweaver is better, produces more legible code, and does not have such a strong tendency to optimize things in non-standard ways (Microsoft's treatment of HTML responds to Explorer's ability to display it, not to W3C Standards). That said, I highly prefer to use Homesite (the text editor component of Dreamweaver). It's everything an HTML text editor should be, and more. There's just something about squeezing a table or two out of that layout and really optimizing the code... --
I'd guess that the point when you can't distinguish an artificial/alternate intelligence from a human one, that pretty much classifies as human, regardless of it being electronic, biological or otherwise, and should be granted the same rights as humans. This is a fine line to cross. If we start determining what classifies as human being based on anthropomorphism / genetical constitution or other factors we'll end up with justified, legal racism and apartheid. I don't care much about tissue. Let scientists do what they want to a bunch of cells. But if they come up with something intelligent, that something should have the right to be regarded as a someone --
Slashdot Mirror
You raise very good points, and it has been a pleasure to keep this thread going. Thanks for helping me dismember this dead horse :-) (I think we went past beating it about three posts ago). Hope to read more from you in the future. At this point, I think we've reached a good level of agreement. Thanks for the insight.
--
I am Mexican. Mexico is a developing 3rd world country, and a lot of the more promising government projects are software industry development based, like this Project to give a boost to multimedia and animation industry in Jalisco (spanish only, sorry). Just what do you think we are? a bunch of freaking retards? There IS industry and education in here! It's not all about helping out the extremely poor, it's about helping society and industry develop so there is more cash flow, education and yes, more programs to help those in need. That said, open source initiatives DO help. A lot. And a bunch of the people involved in open source projects are from third world countries (incidentally, Miguel de Icaza is Mexican).
Please note, too, that foreign backing is not all there is to charity. A country has to develop its own strenghts and help itself. I think we're on our way there. Politics and economic situations are a complicated thing, and broad statements like the ones you make are usually off the mark.
--
Because computers still can't do it. If you think you can drive better than a computer you should also try to run a semaphore network or perform one million mathematical operations in less that a minute. Accept it: Creativity (and writing algorithms) may be the realm of people, but determinism is something a bunch of chips with software are much better at.
As a programmer, I assure you, I can write algorithms that show bias. Of course, you are correct that none of them think.
As another programmer, I assure you, that would greatly surprise me. Of course you can write a program that selects one answer over the other for arbitrary reasons! and you know what? That's exactly what it will do! It won't decide to go the other way just because it feels like it. You can even put in a random decision, and it will take a random decision! Don't you think an analysis of the algorithm would reveal the intended outcome? That was the original point of this discussion. I'm not advocating the rise of the machines. But I sure couldn't be trusted with an image and expected to come up with a totally unbiased, objective enhancement of it. Good or bad, an algorithm that doesn't try to replace the original content of the image and instead works with it will do a more reliable, repeatable job than I, or anyone else for that matter, can do (please note again that I'm not talking about a better enhancement, rather a more objective one! (which is what would be needed to bring this in front of a jury)
--
No, they can't. An algorithm is not random. It is a process applied to data. This is just bullshit. Of course some algorithms will work better or worse on a particular set of data. If you can prove to a jury that the algorithm is not pulling an image out of its ass (which a human may), it is a valid point.
To think that a process is better just because it is wholly deterministic is silly
I'm not saying it is better in a general sense. I'm saying it is more trustworthy, repeatable and does not introduce external variables, which for this particular application, is better. An algorithm does not say "this line would look much better if I stretched it this way" whereas a human may. The algorithm does not care if the resulting image looks like shit.
Interesting. "proving" something with data? Either that wasn't science or verificationism must not be dead after all.
I think you're misusing the word, and Data Mining CAN prove that there exists a correlation or not in a set of data. For instance: a scientist may look at data for the foci of a disease in a country and see that there is roughly a correlation between locations of nuclear plants and the disease sprouts. He may then propose that there is a causal relationship. But after running the raw data through a data mining program, a much higher probability relationship may be found between temperature+water concentrations+disease (these three-tiered relationships are difficult for humans to find), showing that the former, seemingly correct deduction was just coincidental in nature.
Regardless, different data mining techniques would each give a different "score" would they not? What should that difference be called if not a "bias"?
Different data mining techniques are just different ways of analyzing a data set. They do not show statistically different results (they ALL agree on how much of each of the original data there is). Some of them are just better at finding certain kinds of patterns than others. It largely depends on the format of your raw data. If you have a point, make it. Processes don't think. They are not biased! is that clear?
--
How does that work? you two go around pissing people off? =)
--
You have a point about the group experience, but this thing could certainly be used in conjunction with a spherical screen and digitally generated animation to make one hell of an entertaining ride. All in all, I think the flexibility of the framework makes up for the differences in experience. Maybe, as you say, it won't end up competing with roller coasters, but I think it will make great rides with the right add-ons.
--
This is offtopic, but I took a look at your posting history and noticed that you started at 0 and just recently started getting Score:1 on your posts. How was that possible? I thought every logged in user had a default 1 score and 0 was for AC's. Could it be because you used "Troll" in your username? or was there a change in the Slashdot code to modify this behavior? anyone?
--
So, what's that got to do with anything? The use of enhancing techniques CAN be abused. By selectively enhancing a section of a print against another one you can manipulate the outcome. It is very different to increase visibility on the whole print than to darken some parts, blur others. If there is a human element, it is not to be trusted. Allow me to explain (and sorry for the long rant):
Some time ago I attended a Data Mining Seminar. Many people here will be familiar with the techniques that are used in that area. What's the point of using a computer algorithm to find patterns when we are so good at it naturally, you ask? Why, exactly that. Humans are extremely good at finding patterns. We even tend to see them where they're not, and ignore the extra evidence that may point in different directions. A computer works on the data and does not have prejudice for/against it. Before the AC's jump and say that algorithms can be manipulated too, let me just say that they can be audited for soundness and logical mistakes/mathematically analized, etc. They are much more reliable than a human. There have been numerous instances when a correlation between factors was suspected, and data mining was used to prove/disprove the correlation or give a score to it.
So what's the right way to go here, you ask? I think that an image you can get from running an algorithm against the source is valid and can be considered an objective derivative of the original. Tools that allow to selectively retouch pictures should be out of the question. There are many techniques that can be applied to images to enhance them and that are wholly based on what's already there.
--
I took a look at the source code for this page to see what he was talking about and I know the guy sent his very own file nicely formatted, but you guys should at least have removed the <html> <head> and <body> tags from his document. It is extremely bad form to insert a whole html document inside another one. I don't think this page renders well in most browsers. Maybe the Slashdot Editors can update the story removing that from the source code?
--
Thanks for the insight. That was really helpful. Hope your recovery process goes great. I can understand your point about the routines. I used to be a smoker. The hardest part of not smoking anymore (after the physical need goes away) tends to be dissasociating certain activities (partying, having some coffee, being outside on the cold) with smoking. Well, thanks for a great post :-)
--
Considering NineNine has posted 1575 comments on Slashdot, you may have a point =)
--
Well, you can hear all about the non-virtues of the vice by the people not involved in it, right? If you want to actually know why people do it, you have to hear the opinion of the people doing it. After all, there must be more to heroine that shitting your pants constantly and screwing your life, or people wouldn't be so keen on using it, right? (have you seen trainspotting?). Same thing with all vices. True, the opinion of the addict is biased. But the opinion of the non-addict is also uninformed.
--
Good point. I think this would be akin to road billboard ads trying to pass themselves as road signs. That would surely get them in trouble real fast. Maybe computerdom should be viewed in the same light?
--
On the same vein: the passing of time is relative depending on which side of the bathroom door you are.
--
Icing on the mornings cake: I got up on time, drove to work, posted a story, and then forgot to press *save* on the goddamn web form. So for hte next 2 hours I keep deleting submission after submission about Mozilla 1.2.1 thinking "geezus, are people blind?" and not realizing that no, I am in fact stupid. Of course, why so many people submit a bug fix release of a web browser is beyond me. Some stories I'd rather not post, but sheer volume of submissions really makes it impractical to ignore them
--
It's a fairly common type of comment. There's this one for instance (by another guy, in the same vein). Not a literal copy though.
--
A PC system tweaked for gaming can deliver astounding frame rates at 1600x1200. That is the main difference. Try to hook up your 17" monitor to a console (can't be done, but would be nice) and you would soon see the limitations of a console.
--
And why would they need 25 BILLIONS to do something like that? that's the whole point of the "fable". Where does all this money go? Does NASA have full disclosure of expenses? Why are independent organizations able to achieve pretty impressive results with a fraction of the cost of NASA? Furthermore, why is it accepted that this agency should spend BILLIONS (that's thousands of millions of dollars for the ones that use the more common measurements) in order to get results they were able to get 20+ years ago? I'm not in the US, but if I were I'd sure be pissed off about this. Where's the space travel package in my travel agency yet?
--
Damn! you're right. I made a quick search for the name, but didn't come up with that reference... Oh well.
--
Hydra would be a nice name too. And it is sort of a dragon (many headed monster/serpent)... Kind of goes along with the naming scheme (Chimera, Mozilla..)
--
The makers of Gimp don't seem to be interested in adding this functionality, though, so we'll have to turn to Zope or Kontour for that functionality
--
Layers and composite color schemes are some of the most difficult things to master in these programs and a mayor reason why people find them hard to use ("I pushed the pencil button and this thing doesn't draw anything!!" while painting over an invisible layer is a common complaint). AutoCad and 3d Studio are very hard to learn too.
Open source interfaces tend to be designed for functionality rather than mass appeal, which is not a bad thing. Things like blender, vi and the Gimp can be a pain in the ass the first time you try them, but over time you really find out the reasoning behind the interface functionality (there usually are very good reasons to do things one way or the other, not just accumulated cruft like a lot of commercial interfaces).
--
Dreamweaver is better, produces more legible code, and does not have such a strong tendency to optimize things in non-standard ways (Microsoft's treatment of HTML responds to Explorer's ability to display it, not to W3C Standards). That said, I highly prefer to use Homesite (the text editor component of Dreamweaver). It's everything an HTML text editor should be, and more. There's just something about squeezing a table or two out of that layout and really optimizing the code...
--
I'd guess that the point when you can't distinguish an artificial/alternate intelligence from a human one, that pretty much classifies as human, regardless of it being electronic, biological or otherwise, and should be granted the same rights as humans. This is a fine line to cross. If we start determining what classifies as human being based on anthropomorphism / genetical constitution or other factors we'll end up with justified, legal racism and apartheid. I don't care much about tissue. Let scientists do what they want to a bunch of cells. But if they come up with something intelligent, that something should have the right to be regarded as a someone
--
I know. That was meant to be funny. I guess I'll better get some coffee now...
--