DFW, 200k is high but I still know several people that are currently making it working out here (and with the things they have they DEFINITELY are not lying or they are dealing meth on the side...). DFW has a great market for software engineers and so many people overlook it.
Granted that may be true to an extent, but given time they will be forced to take the lower wage jobs or simply not work therefore this will only be true in the short term. Not only that, but software engineers getting that kind of salary is hardly a long term expectation. There has been a surge in the pay scale even after things like the dot com bubble bursting. I should know, I kind of work in the industry (and not in silicon valley where things are way out of whack).
Glut of Software Engineers? Where the hell are you pulling that from? Maybe Google has a glut of software engineers applying to them because they are a massive company in the industry, but your average or even above average software shop is starving for software engineers hence why they pay on average 60k+ to college grads and 150k to 200k to someone experienced. That is simple economics, because if there was a glut, then they wouldn't be able to command those kind of wages.
I'm no materials engineer/scientist, but I would think some of this could be overcome simply by width and orientation of the tin strips. Considering that TFA stated that this is only conducting on the edges of the "stanene" you could probably somewhat pyramid the tin layers with some type of insulator between them and the resulting magnetic field size could be accounted for at each layer such that they don't interfere with each other. Also by just making sure major strips were run perpendicular to each other the magnetic fields then would be unable to interfere with each others respective operation.
Again, pure speculation on my part based on my very fundamental understanding of the physics, I could be way off base since I don't know all the more advanced properties and such.
No it can be 100% efficient within the system. By the first law of thermodynamics this is completely possible. In order to GET that energy into the system some must be removed from another system, but as long as it just stays on the "stanene" then it is implied to remain as electrical energy and not turn into heat/radiation/whatever.
For starters, job creation and economic investment. I know the area near the SSC and almost everyone there will tell you when it was under construction the entire area was projected to undergo a massive economic boom. It started to do it too until the project was cancelled and then things stagnated in the area pretty badly. It has only been in the past 5 or so years that there has been significant growth for that area because everything changed direction.
Dallas now has several major economic sectors, but scientific research is way behind for it. If the SSC had actually been completed, that area would probably be one of the largest scientific hubs in the world by now and we likely would have found the Higgs (and other things) a lot earlier. This in turn would have generated a much greater demand to come to the US for STEM type education and research and likely would have helped our post secondary and general secondary education system flourish as opposed to what is happening now with them struggling to keep up with rising costs (especially in Texas, where the secondary education is abysmal in most areas). If even a small ROI had been shown that probably would have helped the scientific community in the US in general with some of these larger projects and we would likely being growing that sector much better than now (funny, that is now what half the country has talked we NEED to do in order to continue competing on the global market...).
Other industries would have also been able to grow out of the construction and maintenance efforts and Dallas would probably have seen some amazing growth from that alone. Not to mention the indirect benefits of growing the field causing other improvements in the long run. Big Science is like IT, very necessary and can have tremendous benefits, but since you can't easily quantify all those benefits on paper people will shoot it down (kind of ironic considering that is the entire point of most hard science fields, quantification and making it easier to understand).
I find time estimations to be by far the most difficult thing. Especially when sales promises the customer the damn sky and says we will have delivered by Monday... Our sales guys even readily acknowledge now they often give unreasonable timelines to the customer and the developers bail their asses out.
Good IDE with autocomplete typing gives pretty close to the best of both worlds. Only one time do you have to type out the long name and after that it is just autocomplete everything. I lean towards long names partially because of this and on large codebases this becomes necessary to help facilitate proper documentation.
Honeslty, a lot of the best documentation I have seen goes hand in hand with just having a good design. When the program is well designed most variable names, function names, objects, etc. are meaningful and if you adhere to good practices such as single use, limits on functions line size etc. The code explains a lot of itself. It does NOT eliminate the need for comments (at my company we use less of them, mostly they are to explain design decision or crazy areas of the code that are not intuitive or were hacked in because of a deadline) or general documents like design docs and functional specs.
Comments and other documentation can be extremely useful and not too time consuming to create as long as they are done right.
Actually there have been significant improvements in the manufacturing process and the general solar cell efficiency that may make that statement incorrect. I haven't looked at the math in some time, but they have made significant advances in solar cell production such that I don't think they actually are a net loss in energy now.
Ah yes, that is the one. Hopefully some proprietary device development might start focusing on specific functionality to give it some traction, though I still kind of think mechanical buttons are going to win out for the time being. NVidia actually had some Shield demos out at Quakecon this year, maybe something like that with more touchscreen focus could help them out (Nintendo, I am looking at you with the damn DS/3DS already...).
Yea, see you should probably have looked at more context if you planned to insult what he was talking about. Even without hearing the rest of the keynote, the Kinect comment illustrates exactly the problem. Mechanical buttons are still far superior for user actions than trying to virtualize the buttons via touchscreen or whatever bullshit apple is trying to pull. The Kinect is the same idea as their stupid magic mouse only a bit worse, where you have to make some exact action gesture combination as opposed to a mechanical button "one trick pony" that has two states: pressed or not pressed. That mouse does not have a binary state because of the software interpretation layer trying to translate it to a binary state when it has more cases than that.
That isn't to say that these could become better controls, and Carmack himself even said with better accuracy, less latency, and taking fuzzy cases into account this could become fairly viable, but right now it is still a damn joke.
I think the bigger complaint here is how easy it would be for someone that has no real clue how to perform a real attack in such a way gaining access to the computer easily (as many have stated due to the computer being left unattended and unlocked for a short period of time). For me, I constantly lock my computer while I am not sitting at it anyway, and usually if someone asks to use it I am right beside them (though not 100% of the time). The passwords I have stored in Chrome are mostly non-essential accounts/passwords though I should probably double check it when I get home and purge anything that might not be.
I think a master password not tied to the OS login credentials would be pretty easy for them to add on and abate most of the fears people have with the password management system in chrome. They could still allow the browser proper access for the passwords through secure means while keeping snoopers at bay (just like what Firefox already implmented). The thing that surprises me here is NO ONE seemed to noticed this for so long. I stumbled onto to this quite a while back (I want to say over a year, but I think its more like 6+ months) and thought it was a poor implementation, but because I practice other physical access security measures better (BIOS password, optical drive behind main drive in boot priority, lock the computer when not at it etc.) I really didn't worry about it.
Well that is the theory, you use some linguistics tricks in such a way that the bot would need to nearly Turing complete to effectively defeat the system. It is a bit of a challenge though as the system itself would probably become very nearly Turing complete to accomplish this goal, but there are probably some short cuts one could take to help with that challenge. Now it would be quite entertaining to point the Turk at the Turk though if money were not a concern...
Valid point, but there are probably ways around that to make it more accessible though damned if I can think of one right now. I really have already expanded a hell of a lot more on this idea than I intended to in this discussion, my point was that it is a possible replacement, not that it definitely would work. It might, it might not, but there are probably alternatives to the current system, people just have to research them and put some work into it.
One word: anagrams. Now that "solution" is broken entirely and the bot now needs to do heuristic natural language processing to figure out which word it is supposed to be, much less what words are possible as it may not be in the bots dictionary which at best can run lookups in O(nlogn) but more likely will run them in at least O(n^2) probably worse. You can also use homonyms to force additional computational efforts on the bots part.
String processing just to do word matching with known good words is quite complex and often times will run in exponential time. For correct answers you are looking at running in essentially O(n^n) where n = the number of letters, for EVERY SINGLE WORD which is geometric growth in computation time.
Everyone seems to be missing the point with trying to defeat this argument using processing logic. The CURRENT system is defeated using processing logic and short of using flat out encryption levels ALL systems will eventually be beaten by a bot with unlimited time and processing power. Essentially you have to create something that is difficult for the bot to effectively break and have 100% accuracy. Even the blind squirrel finds a nut so to speak, but natural language processing is not 100% certainty or even close in many cases.
What would make it more entertaining is when the computers "overhear" each other and now you have a logic block for two completely different problems being entered simultaneously. Sad part is, I have seen some people that probably wouldn't notice, would check the code in, and resolve the issue they were working on.
Interesting this comes up. Over the weekend at Quakecon John Carmack addressed the evolution of controllers and his thoughts on the subject in his keynote. My favorite was when he was mocking Apple having the one button mouse and then one upped that by saying the kinect was "like a 0-button mouse."
My main argument would probably be, when you have a physical key/button with tactile feedback you can much more easily ensure an intentional action on the part of the user. Whereas touchscreens you are much more prone to fat fingering the wrong key (although this happens sometimes on keyboards too to be fair, just not nearly as often in my experience). A promising tech I remember reading about I think on slashdot was the touchscreen that could bend in such a way as to give tactile feedback at any point of the screen for a variety of sizes and shapes. Arguments could still be made to have solid keys/buttons still, but brings the two a little closer together at least.
I agree that security through obscurity is not smart practice, but this is only a very basic form of the idea. As I stated, it would require a lot more research into language processing and implementation of a lot of tricks. The two I mentioned in another comment, intentional randomized typos that humans will gloss over and word substitutions for synonyms and slang (possibly even some phrases). Those are only a couple that I can think of right now. I admit, it would require more research and design (that is how the current system came to be as well) and may not even be feasible in the long run, but I am saying there are alternatives to the current system.
There are actually a couple of other basic language techniques I can think of off the top of my head that could be used against the spammers.
For starters, you can use a language processing trick that essentially allows humans to gloss over a lot of typos. In general when a person sees a word as long as the word starts and ends with the correct letters and contains the correct ones (in some cases just close to the correct letters), we will process them like they are in order when they are completely out of order without even thinking about it since we process words and not individual letters. This would add some additional complexity to machine processing because now not only does the bot need to tokenize and understand the sentence, but it has to rearrange half the letters in the words to get the correct meaning from the sentence. A human would barely notice and it could be somewhat randomized within the phrase.
Another thing would be word substitution for slang or synonyms that a machine would then have to reference, but as long as the substitutions are not overly complicated the human user would again have no issues. This could also result in some jargon filled goofy sentences that a person can figure out pretty simply, but a machine would get confused as hell.
Those are just off the top of my head, with some additional research it could possibly be made more viable. I am not saying this would be easy to implement (or necessarily even feasible, you do raise valid points/concerns), there is a reason we use the current image distortion system, but I think there are some alternatives to what is currently being used.
No, there are more to consider than just that corner case in the summary alone. Besides that, the discussion evolved into how bad CAPTCHAs are in the general case too which is what I was talking about.
Pretty much this. I am actually a software engineer and hold a degree in Computer Science, I know a thing or two about the processing power required for this and just because it can be done on the large scale of a supercomputer like Watson, doesn't mean your average spammer can even come close to natural language processing of that level.
Not to be offensive or insulting, but that was some pretty weak satire. Especially when 95% of the internet (including slashdot) posts some very piss poor comments and arguments about things and they are totally serious.
Are off limits and would be banned by this campaign. They are looking for test that can be solved by blind people.
This isn't necessarily an end all solution I am suggesting, but definitely an improvement. This is no more excluding than the current system and helps in all cases but the one you suggest. Why would you shoot down an improvement if it gets all but one corner case while the current system barely even works in the general case?
That is a pretty simplified example, but you can find these all over the place and they are fairly easy to create.
How are these easy to create? To create these questions a program would need to start out with an answer. In this case white. It would then need to randomly generate a unique question that has never been asked before whose answer is white. This question would need to be easy for humans to solve. But it would need to be impossible to solve by the program that just generated the question. Ohh and you need to generate millions of these question answer pairs everyday. It also needs to be done quickly. People don't want to wait longer than a second for the CAPTCHA to load.
Well, these are easy for a person to create, and while I admit a dictionary style would be difficult to maintain and be attack-able if the bots could get the dictionary, I would think in theory some of this could be generated with proper algorithms as this is the essence of encryption (albeit encryption is math based witchcraft to your average user). At the very least some measure of formula could be used to create some of this I would think, but I would have to do some significant work to figure that out (work I am not getting payed to do at that, i.e. spare time).
But would a computer be able to easily ask questions that itself cannot answer, but a human can?
That is very nearly the basis of thought for encryption so I would say it is fairly possible. It is not trivial, but using some language processing techniques you could probably come up with some fairly simple questions and answers on a regular basis. Now, creating them on the fly might be more difficult I admit, but with some effort may be possible.
Slashdot Mirror
DFW, 200k is high but I still know several people that are currently making it working out here (and with the things they have they DEFINITELY are not lying or they are dealing meth on the side...). DFW has a great market for software engineers and so many people overlook it.
Granted that may be true to an extent, but given time they will be forced to take the lower wage jobs or simply not work therefore this will only be true in the short term. Not only that, but software engineers getting that kind of salary is hardly a long term expectation. There has been a surge in the pay scale even after things like the dot com bubble bursting. I should know, I kind of work in the industry (and not in silicon valley where things are way out of whack).
Glut of Software Engineers? Where the hell are you pulling that from? Maybe Google has a glut of software engineers applying to them because they are a massive company in the industry, but your average or even above average software shop is starving for software engineers hence why they pay on average 60k+ to college grads and 150k to 200k to someone experienced. That is simple economics, because if there was a glut, then they wouldn't be able to command those kind of wages.
I'm no materials engineer/scientist, but I would think some of this could be overcome simply by width and orientation of the tin strips. Considering that TFA stated that this is only conducting on the edges of the "stanene" you could probably somewhat pyramid the tin layers with some type of insulator between them and the resulting magnetic field size could be accounted for at each layer such that they don't interfere with each other. Also by just making sure major strips were run perpendicular to each other the magnetic fields then would be unable to interfere with each others respective operation.
Again, pure speculation on my part based on my very fundamental understanding of the physics, I could be way off base since I don't know all the more advanced properties and such.
No it can be 100% efficient within the system. By the first law of thermodynamics this is completely possible. In order to GET that energy into the system some must be removed from another system, but as long as it just stays on the "stanene" then it is implied to remain as electrical energy and not turn into heat/radiation/whatever.
For starters, job creation and economic investment. I know the area near the SSC and almost everyone there will tell you when it was under construction the entire area was projected to undergo a massive economic boom. It started to do it too until the project was cancelled and then things stagnated in the area pretty badly. It has only been in the past 5 or so years that there has been significant growth for that area because everything changed direction.
Dallas now has several major economic sectors, but scientific research is way behind for it. If the SSC had actually been completed, that area would probably be one of the largest scientific hubs in the world by now and we likely would have found the Higgs (and other things) a lot earlier. This in turn would have generated a much greater demand to come to the US for STEM type education and research and likely would have helped our post secondary and general secondary education system flourish as opposed to what is happening now with them struggling to keep up with rising costs (especially in Texas, where the secondary education is abysmal in most areas). If even a small ROI had been shown that probably would have helped the scientific community in the US in general with some of these larger projects and we would likely being growing that sector much better than now (funny, that is now what half the country has talked we NEED to do in order to continue competing on the global market...).
Other industries would have also been able to grow out of the construction and maintenance efforts and Dallas would probably have seen some amazing growth from that alone. Not to mention the indirect benefits of growing the field causing other improvements in the long run. Big Science is like IT, very necessary and can have tremendous benefits, but since you can't easily quantify all those benefits on paper people will shoot it down (kind of ironic considering that is the entire point of most hard science fields, quantification and making it easier to understand).
I find time estimations to be by far the most difficult thing. Especially when sales promises the customer the damn sky and says we will have delivered by Monday... Our sales guys even readily acknowledge now they often give unreasonable timelines to the customer and the developers bail their asses out.
Good IDE with autocomplete typing gives pretty close to the best of both worlds. Only one time do you have to type out the long name and after that it is just autocomplete everything. I lean towards long names partially because of this and on large codebases this becomes necessary to help facilitate proper documentation.
Honeslty, a lot of the best documentation I have seen goes hand in hand with just having a good design. When the program is well designed most variable names, function names, objects, etc. are meaningful and if you adhere to good practices such as single use, limits on functions line size etc. The code explains a lot of itself. It does NOT eliminate the need for comments (at my company we use less of them, mostly they are to explain design decision or crazy areas of the code that are not intuitive or were hacked in because of a deadline) or general documents like design docs and functional specs.
Comments and other documentation can be extremely useful and not too time consuming to create as long as they are done right.
Actually there have been significant improvements in the manufacturing process and the general solar cell efficiency that may make that statement incorrect. I haven't looked at the math in some time, but they have made significant advances in solar cell production such that I don't think they actually are a net loss in energy now.
Ah yes, that is the one. Hopefully some proprietary device development might start focusing on specific functionality to give it some traction, though I still kind of think mechanical buttons are going to win out for the time being. NVidia actually had some Shield demos out at Quakecon this year, maybe something like that with more touchscreen focus could help them out (Nintendo, I am looking at you with the damn DS/3DS already...).
Yea, see you should probably have looked at more context if you planned to insult what he was talking about. Even without hearing the rest of the keynote, the Kinect comment illustrates exactly the problem. Mechanical buttons are still far superior for user actions than trying to virtualize the buttons via touchscreen or whatever bullshit apple is trying to pull. The Kinect is the same idea as their stupid magic mouse only a bit worse, where you have to make some exact action gesture combination as opposed to a mechanical button "one trick pony" that has two states: pressed or not pressed. That mouse does not have a binary state because of the software interpretation layer trying to translate it to a binary state when it has more cases than that.
That isn't to say that these could become better controls, and Carmack himself even said with better accuracy, less latency, and taking fuzzy cases into account this could become fairly viable, but right now it is still a damn joke.
I think the bigger complaint here is how easy it would be for someone that has no real clue how to perform a real attack in such a way gaining access to the computer easily (as many have stated due to the computer being left unattended and unlocked for a short period of time). For me, I constantly lock my computer while I am not sitting at it anyway, and usually if someone asks to use it I am right beside them (though not 100% of the time). The passwords I have stored in Chrome are mostly non-essential accounts/passwords though I should probably double check it when I get home and purge anything that might not be.
I think a master password not tied to the OS login credentials would be pretty easy for them to add on and abate most of the fears people have with the password management system in chrome. They could still allow the browser proper access for the passwords through secure means while keeping snoopers at bay (just like what Firefox already implmented). The thing that surprises me here is NO ONE seemed to noticed this for so long. I stumbled onto to this quite a while back (I want to say over a year, but I think its more like 6+ months) and thought it was a poor implementation, but because I practice other physical access security measures better (BIOS password, optical drive behind main drive in boot priority, lock the computer when not at it etc.) I really didn't worry about it.
Well that is the theory, you use some linguistics tricks in such a way that the bot would need to nearly Turing complete to effectively defeat the system. It is a bit of a challenge though as the system itself would probably become very nearly Turing complete to accomplish this goal, but there are probably some short cuts one could take to help with that challenge. Now it would be quite entertaining to point the Turk at the Turk though if money were not a concern...
Valid point, but there are probably ways around that to make it more accessible though damned if I can think of one right now. I really have already expanded a hell of a lot more on this idea than I intended to in this discussion, my point was that it is a possible replacement, not that it definitely would work. It might, it might not, but there are probably alternatives to the current system, people just have to research them and put some work into it.
One word: anagrams. Now that "solution" is broken entirely and the bot now needs to do heuristic natural language processing to figure out which word it is supposed to be, much less what words are possible as it may not be in the bots dictionary which at best can run lookups in O(nlogn) but more likely will run them in at least O(n^2) probably worse. You can also use homonyms to force additional computational efforts on the bots part.
String processing just to do word matching with known good words is quite complex and often times will run in exponential time. For correct answers you are looking at running in essentially O(n^n) where n = the number of letters, for EVERY SINGLE WORD which is geometric growth in computation time.
Everyone seems to be missing the point with trying to defeat this argument using processing logic. The CURRENT system is defeated using processing logic and short of using flat out encryption levels ALL systems will eventually be beaten by a bot with unlimited time and processing power. Essentially you have to create something that is difficult for the bot to effectively break and have 100% accuracy. Even the blind squirrel finds a nut so to speak, but natural language processing is not 100% certainty or even close in many cases.
What would make it more entertaining is when the computers "overhear" each other and now you have a logic block for two completely different problems being entered simultaneously. Sad part is, I have seen some people that probably wouldn't notice, would check the code in, and resolve the issue they were working on.
Interesting this comes up. Over the weekend at Quakecon John Carmack addressed the evolution of controllers and his thoughts on the subject in his keynote. My favorite was when he was mocking Apple having the one button mouse and then one upped that by saying the kinect was "like a 0-button mouse."
My main argument would probably be, when you have a physical key/button with tactile feedback you can much more easily ensure an intentional action on the part of the user. Whereas touchscreens you are much more prone to fat fingering the wrong key (although this happens sometimes on keyboards too to be fair, just not nearly as often in my experience). A promising tech I remember reading about I think on slashdot was the touchscreen that could bend in such a way as to give tactile feedback at any point of the screen for a variety of sizes and shapes. Arguments could still be made to have solid keys/buttons still, but brings the two a little closer together at least.
I agree that security through obscurity is not smart practice, but this is only a very basic form of the idea. As I stated, it would require a lot more research into language processing and implementation of a lot of tricks. The two I mentioned in another comment, intentional randomized typos that humans will gloss over and word substitutions for synonyms and slang (possibly even some phrases). Those are only a couple that I can think of right now. I admit, it would require more research and design (that is how the current system came to be as well) and may not even be feasible in the long run, but I am saying there are alternatives to the current system.
There are actually a couple of other basic language techniques I can think of off the top of my head that could be used against the spammers.
For starters, you can use a language processing trick that essentially allows humans to gloss over a lot of typos. In general when a person sees a word as long as the word starts and ends with the correct letters and contains the correct ones (in some cases just close to the correct letters), we will process them like they are in order when they are completely out of order without even thinking about it since we process words and not individual letters. This would add some additional complexity to machine processing because now not only does the bot need to tokenize and understand the sentence, but it has to rearrange half the letters in the words to get the correct meaning from the sentence. A human would barely notice and it could be somewhat randomized within the phrase.
Another thing would be word substitution for slang or synonyms that a machine would then have to reference, but as long as the substitutions are not overly complicated the human user would again have no issues. This could also result in some jargon filled goofy sentences that a person can figure out pretty simply, but a machine would get confused as hell.
Those are just off the top of my head, with some additional research it could possibly be made more viable. I am not saying this would be easy to implement (or necessarily even feasible, you do raise valid points/concerns), there is a reason we use the current image distortion system, but I think there are some alternatives to what is currently being used.
From TFA:
and dyslexic people and older ones
No, there are more to consider than just that corner case in the summary alone. Besides that, the discussion evolved into how bad CAPTCHAs are in the general case too which is what I was talking about.
Pretty much this. I am actually a software engineer and hold a degree in Computer Science, I know a thing or two about the processing power required for this and just because it can be done on the large scale of a supercomputer like Watson, doesn't mean your average spammer can even come close to natural language processing of that level.
Not to be offensive or insulting, but that was some pretty weak satire. Especially when 95% of the internet (including slashdot) posts some very piss poor comments and arguments about things and they are totally serious.
Simple things like image processing
Are off limits and would be banned by this campaign. They are looking for test that can be solved by blind people.
This isn't necessarily an end all solution I am suggesting, but definitely an improvement. This is no more excluding than the current system and helps in all cases but the one you suggest. Why would you shoot down an improvement if it gets all but one corner case while the current system barely even works in the general case?
That is a pretty simplified example, but you can find these all over the place and they are fairly easy to create.
How are these easy to create? To create these questions a program would need to start out with an answer. In this case white.
It would then need to randomly generate a unique question that has never been asked before whose answer is white.
This question would need to be easy for humans to solve. But it would need to be impossible to solve by the program that just generated the question.
Ohh and you need to generate millions of these question answer pairs everyday. It also needs to be done quickly. People don't want to wait longer than a second for the CAPTCHA to load.
Well, these are easy for a person to create, and while I admit a dictionary style would be difficult to maintain and be attack-able if the bots could get the dictionary, I would think in theory some of this could be generated with proper algorithms as this is the essence of encryption (albeit encryption is math based witchcraft to your average user). At the very least some measure of formula could be used to create some of this I would think, but I would have to do some significant work to figure that out (work I am not getting payed to do at that, i.e. spare time).
But would a computer be able to easily ask questions that itself cannot answer, but a human can?
That is very nearly the basis of thought for encryption so I would say it is fairly possible. It is not trivial, but using some language processing techniques you could probably come up with some fairly simple questions and answers on a regular basis. Now, creating them on the fly might be more difficult I admit, but with some effort may be possible.