With the right perl modules, or even perhaps in shellscript with things like CURL, you could roll your own http/html regression tester. Won't handle javascript of course, and won't notice browser-dependant problems. You might be able to find a generic javascript/DOM library for unix to do the JS thing from your regression tester though.
enterprise Pronunciation Key (ntr-prz) n. An undertaking, especially one of some scope, complication, and risk. A business organization. Industrious, systematic activity, especially when directed toward profit: Private enterprise is basic to capitalism. Willingness to undertake new ventures; initiative: "Through want of enterprise and faith men are where they are, buying and selling, and spending their lives like serfs" (Henry David Thoreau).
I believe the definition in this context is the second one, "a business organization". To be more specific, in computing circles "enterprise" is meant to differentiate from "small business". It is meant to connotate large companies who need solutions that scale extremely well. Some of largest US Enterprises have over 100,000 employees, but a typical set of desktops managed by one central authority is probably in the 1-20k range.
In rough terms, c=s*r, meaning the cost will rise dramatically for high speed reliable storage versus low speed crap storage.
In addition, how the storage is designed (and how much more it can cost) depends a lot on data access patterns as well (read-mostly vs write-mostly, oltp vs dss vs datawarehouse vs...., etc).
Maxtor has 0.3TB IDE for $1/GB. If you built a huge array of IDE controllers for these, your disk cost for 50TB would be around $50k. If some vendor actually built a beast with the requisute number of IDE busses and whatnot, the chassis might run you another $100k. All in all, real cheap storage. But it would suck on performance and reliability, put out too much heat and noise probably, etc, etc.
Highly available disk arrays with extreme disk platter performance and large amounts of caching can easily run $20 million for 50TB, if not more. There are middle of the road solutions though, it doesn't have to be that expensive unless you're going all out for huge concurrency and speed in an OLTP environment that requires 99.999999% uptime.
Yeah, I was actually quite content and never had any space problems on my home boxes using 9 and 18 GB drives. Then recently, I built a new machine, and just for the hell of it because they didn't cust much, I threw in for two 40G IDE drives for the new box - basically one for linux and one for windows.
Since I knew it was far more than I need, I've made a conscious effort to be a litterbug just to see how much I can take up. I never delete anything, never uninstalled unused apps - I've run an object code cach (ccache) on the linux side of the box for months now with an unlimited cache (compiling gcc, glibc, mozilla, etc)... I've just been messy all over, and installed tons of software - several multi-CD games under windows for sure, and just about everything under the sun in linux.
To date I haven't reached the halfway mark on either drive, months later. The linux one is at about 19G used, the windows one around 10.
Is the article accurate? The settlement was just to "hide" the bundled software? There was no part indicating that the services offered to IE, Outlook, Media Player, etc by the OS have to be available to competitors, so that they can integrate and interoperate as seamlessly? No wonder 9 states dissented.
Good chess players think in terms of "blocks" on the board. They don't see individual moves and peices, they see larger conceptual board positions and sets of interrelated peices that confer advtanges to one side or the other, and they see "patterns" that are difficult to define in explicit terms. It was in one of Hofstadter's books this was brought up in examining how humans play chess.
I think you miss why it's harder to emulate PPC on x86 than it is to emulate x86 on PPC. PPC is RISC - lots of small instructions and lots of registers. x86 is CISC, fewer instructions that do more, less registers.
Therefore, it's (relatively) easy to decompose a complex x86 instruction into a sequence of simpler PPC instructions that get pipelined well. However, in the PPC on x86 emulator, the only way you could be even remotely efficient would be to do some intelligent analysis of the incoming opcodes as "groups" that resemble the function of a single x86 opcode and try to translate it that way, which would be hard to do in the fly in a tight runtime emulator core. If you stick to the standard quick-and-dirty translation, you end up failing to really utilize the x86's complex instructions well, as you're issuing one or more complex CISC instructions per simple RISC instruction, kinda (at least that's one way to think of it). On top of all that, the fact that the PPC has way more registers doesn't help either. The code compiled for PPC will be optimized to make use of all of them, with tight loops assuming register-speed access to more variables than the x86 core can put in registers, so you put them in memory and take yet another hit.
The above isn't really all that correct, but hopefully it conveys the general gist of things.
Yes, I've studied, and I agree with what you've said above. I think it does make a difference however. I think small amount of true randomness injected all over the place at various stages of every decision could turn an otherwise dull and boring network into something more human. I think this is possibly the basis for creativity in the form of minor changes to known concepts - and that it also allows "thinking outside the box" (to borrow a horrible catch phrase) - and perhaps most importantly, it probabilistically (with very good chances) prevents the brain from ever getting stuck the way a normal machine would (a major problem mentioned in MMT, don't remember if it was in GEB), which has a great deal to do with intelligence.
I am not a PurifyPlus User, so who knows - but based on the description at the article's link - it seems everything PurifyPlus provides is already available on linux for free. The types of profiling and coverage stuff they discuss can be done by gcc/gcov, and there's lots of shim libraries that hook malloc/free/new/delete/etc and check for leaks. Why is this so special?
On the contrary, I don't think a model of the brain as basically a computer augmented by a powerful RNG (or many RNGs) means that we're Different and Special. Quite the opposite, it implites that those human virtues like "intelligence", "insight", "creativity", etc.. are nothing special at all, and can be reproduced in a machine by figuring out the right neural net configuration and supplying truly random inputs in the right places.
Considering the simplicity of the desired operations, how hard would be to build one around a super-simple microcontroller yourself? I bet a PIC could do it just fine (or even a PIC packaged up as a Basic Stamp for even easier programming). You just need a driver chip for the keypad rows/columns that will interface as input to the controller, and a very simple LCD display.
It is sensationalist to say that the scientists were surprised - if they work in this field surely they've seen ersults similar in meaning before.
It is sensationalist to say the machines learned on their own. Perhaps you don't understand the scope of humans call "learning". Learning within a small fixed set of parameters set up by a human isn't learning. Notice that humans don't face a halting problem or break into infinite loops when we learn wholly new things (well, except maybe a few in asylums).
Thinking is pretty clearly defined in my head. Simple computer software making selections based on the performance of other simple computer software doesn't constitute thinking to me, it constitutes executing my explicit instructions.
Ok you can throw out my semantics on "Cheating", but the fact is that this design process they used is impractical in any real situation. It will always evolve to be too specialized. You can't assume that you can fix this by immersing it in diverse environments during training - the technology only becomes truly useful when it hypothetical goes beyond our ability to define parameters for it in the first place.
I'm of the opinion (not that I have any good basis for this, it's a layman's opinion) that the human brain is basically operating like a digital neural net with an incredibly good RNG available all over the place. I have a feeling it's really the randomness that makes us human and intelligent. That does put us beyond DFA and Turing. And I think quantum mechanics can indeed provide us this randomness.
Well, I disagree on evolutionary designs winning, but it's purely a matter of opinion at this point, and I respect yours. I listed only GEB precisely because I feel he makes very strong, subtle, and eloquent arguments for evolutionary AI in that book. I respect Hofstadter a *whole* lot, but I have my own opinions on these things.
Don't forget the old standby - the MIT LPF's rather huge paper from 1991 on the subject - it's still one of the better papers against patents around:
Against Software Patents
Fuck you, I'm not a fucktard or a christian. Based on my layman's understanding of AI research, which at least includes reading works like GEB, I believe genetic algorithms are a flawed approach.
The article is sensationalist and irresponsible, as it talks of how the genetic algorithm "surprised the scientists", and how nobody knows how the circuit "figured out" one trace could act as an antenna.
The problem is that the non-tech-savvy of the world will read this and actually be made to believe these are thinking machines which are truly learning on their own. It conjures up images of a Matrix future.
I'm quite sure the scientists didn't find the results all that stunning. They ran random mutations and "evolved" an oscillator from the interconnections of 10 transistors. The algorithm of course *failed* to generate an oscillator, and instead cheated by picking up a nearby radiowave.
Nothing in the circuit "figured out" about antennas and radio waves - it was just random luck, much as any result in such an experiment is.
Some might argue with calling the cheating oscillator a failure. I disagree - I think it's a wonderful example of how far AI research has to go yet. What they wanted was an oscillator, presumably one that would work (were this a circuit designing machine in the real world) elsewhere outside the lab. The algorithm was too dumb to realize it's design won't be portable past the lab table.
I really don't think random mutation with selection is going to be the answer, if there's even an answer to be had. Computers are for automating, humans using them as tools are for innovating.
Even if your "company" is just one individual who knows a lot about JBoss, $5k/year is cheap. If your full-time job is being a JBoss developer/consultant, you will be charging clients per hour out the wazoo like all consultants, and raking in enough to make this amount trivial.
I think these JBoss guys have really hit the nail on the head when it comes to making an open source business model work financially. Personally, I dislike java as anything but a client-side language for a thin GUI, so JBoss is not my cup of tea - but the model is impressive and I'm proud of them.
Google could put long term cookies to good use if they do it responsibly. They can't tell which search result you click due to direct linking (although they could see which search results you pull from their cache) - but they could definitely bias your query results based on your previous queries. Example:
UserA has a pattern of computer-related querys UserB has a pattern of photography-related querys Both search for "lens effects". UserA gets some information on computer graphics lens effects in software and algorithms ranked higher. UserB gets some info on how to use real camera lenses to achieve neat lens effects in photography ranked higher.
Search perlmonks.org for Tilly's article on the subject a while back. It appears that by most states' labor laws, if you are an exempt, salaried, full-time professional - the company does in fact own all of your output, even when you're not at work, and they don't need a special contract to get these rights. If you work as unix sysadmin, and you develop and patent a new lawn sprinkler on your own time on the weekends, they can take your patent away from you. They certainly in this light own your output during work hours, which means they very well can try to enforce that you don't do things like use hotmail.
IMHO, for the most part technical training is for people who "don't get it" and probably won't much better after th course, or neccesary for products that are so non-intuitive you shouldn't be buying them to begin with.
I take company-paid training all the time, usually in a nice city. I'd much rather spend a week hanging out in a hotel on the company's dime and listening to some bonehead moron read their product's user manual to me very slowly during the day for a few hours than sit at my desk at work.
I agree on the processor front. My most recent home box is an Athlon XP, because I liked the price/performance ratio versus intel, and I can deal with the issues for the money. However, I still lean towards Intel for real servers that I depend on, and there's solid reason why.
On a processor level, AMD has Intel beat in the overall - the processor is just as good if not better for less money depending what part of the leapfrog cycle we're in. The problem is in the chipsets. Intel's motherboard chipsets have a history of being extremely well supported in all OS's and very good quality. Every AMD-compatible chipset I've ever used has had problems. You name it: VIA, SiS, even AMD's own MP760 or whatever it was called. Invariably all the chipsets available for AMD processors have problems. Sometimes it's hardware bugs, sometimes it's instability. More often than not, it's just that driver support isn't really there in Linux or Windows, or IRQ tables are kinda hosed.
In the end, it's the chipsets that keep me buying Intel for production. If Intel ever released a chipset that supported an AMD processor, I'd deploy the two in a heartbeat, but fat chance of that happening soon.
Slashdot Mirror
With the right perl modules, or even perhaps in shellscript with things like CURL, you could roll your own http/html regression tester. Won't handle javascript of course, and won't notice browser-dependant problems. You might be able to find a generic javascript/DOM library for unix to do the JS thing from your regression tester though.
DUH
The pyramids were created by an alien civilization as a star gate. They're returning in 2038 to enslave us.
I believe the definition in this context is the second one, "a business organization". To be more specific, in computing circles "enterprise" is meant to differentiate from "small business". It is meant to connotate large companies who need solutions that scale extremely well. Some of largest US Enterprises have over 100,000 employees, but a typical set of desktops managed by one central authority is probably in the 1-20k range.
What was your problem with my comment?
40 PC's hardly constitutes an Enterprise. It would be an entirely different story if you had 5k+ desktop users to install to, support, and make happy.
The big factors in storage cost, breifly:
r) Reliability
s) Speed
c) Cost
In rough terms, c=s*r, meaning the cost will rise dramatically for high speed reliable storage versus low speed crap storage.
In addition, how the storage is designed (and how much more it can cost) depends a lot on data access patterns as well (read-mostly vs write-mostly, oltp vs dss vs datawarehouse vs
Maxtor has 0.3TB IDE for $1/GB. If you built a huge array of IDE controllers for these, your disk cost for 50TB would be around $50k. If some vendor actually built a beast with the requisute number of IDE busses and whatnot, the chassis might run you another $100k. All in all, real cheap storage. But it would suck on performance and reliability, put out too much heat and noise probably, etc, etc.
Highly available disk arrays with extreme disk platter performance and large amounts of caching can easily run $20 million for 50TB, if not more. There are middle of the road solutions though, it doesn't have to be that expensive unless you're going all out for huge concurrency and speed in an OLTP environment that requires 99.999999% uptime.
Yeah, I was actually quite content and never had any space problems on my home boxes using 9 and 18 GB drives. Then recently, I built a new machine, and just for the hell of it because they didn't cust much, I threw in for two 40G IDE drives for the new box - basically one for linux and one for windows.
Since I knew it was far more than I need, I've made a conscious effort to be a litterbug just to see how much I can take up. I never delete anything, never uninstalled unused apps - I've run an object code cach (ccache) on the linux side of the box for months now with an unlimited cache (compiling gcc, glibc, mozilla, etc)... I've just been messy all over, and installed tons of software - several multi-CD games under windows for sure, and just about everything under the sun in linux.
To date I haven't reached the halfway mark on either drive, months later. The linux one is at about 19G used, the windows one around 10.
Is the article accurate? The settlement was just to "hide" the bundled software? There was no part indicating that the services offered to IE, Outlook, Media Player, etc by the OS have to be available to competitors, so that they can integrate and interoperate as seamlessly? No wonder 9 states dissented.
Good chess players think in terms of "blocks" on the board. They don't see individual moves and peices, they see larger conceptual board positions and sets of interrelated peices that confer advtanges to one side or the other, and they see "patterns" that are difficult to define in explicit terms. It was in one of Hofstadter's books this was brought up in examining how humans play chess.
I think you miss why it's harder to emulate PPC on x86 than it is to emulate x86 on PPC. PPC is RISC - lots of small instructions and lots of registers. x86 is CISC, fewer instructions that do more, less registers.
Therefore, it's (relatively) easy to decompose a complex x86 instruction into a sequence of simpler PPC instructions that get pipelined well. However, in the PPC on x86 emulator, the only way you could be even remotely efficient would be to do some intelligent analysis of the incoming opcodes as "groups" that resemble the function of a single x86 opcode and try to translate it that way, which would be hard to do in the fly in a tight runtime emulator core. If you stick to the standard quick-and-dirty translation, you end up failing to really utilize the x86's complex instructions well, as you're issuing one or more complex CISC instructions per simple RISC instruction, kinda (at least that's one way to think of it). On top of all that, the fact that the PPC has way more registers doesn't help either. The code compiled for PPC will be optimized to make use of all of them, with tight loops assuming register-speed access to more variables than the x86 core can put in registers, so you put them in memory and take yet another hit.
The above isn't really all that correct, but hopefully it conveys the general gist of things.
Yes, I've studied, and I agree with what you've said above. I think it does make a difference however. I think small amount of true randomness injected all over the place at various stages of every decision could turn an otherwise dull and boring network into something more human. I think this is possibly the basis for creativity in the form of minor changes to known concepts - and that it also allows "thinking outside the box" (to borrow a horrible catch phrase) - and perhaps most importantly, it probabilistically (with very good chances) prevents the brain from ever getting stuck the way a normal machine would (a major problem mentioned in MMT, don't remember if it was in GEB), which has a great deal to do with intelligence.
Because there are no Chinese knockoffs that fit his needs anymore, because he's such a minority market to the manufacturers.
I am not a PurifyPlus User, so who knows - but based on the description at the article's link - it seems everything PurifyPlus provides is already available on linux for free. The types of profiling and coverage stuff they discuss can be done by gcc/gcov, and there's lots of shim libraries that hook malloc/free/new/delete/etc and check for leaks. Why is this so special?
On the contrary, I don't think a model of the brain as basically a computer augmented by a powerful RNG (or many RNGs) means that we're Different and Special. Quite the opposite, it implites that those human virtues like "intelligence", "insight", "creativity", etc.. are nothing special at all, and can be reproduced in a machine by figuring out the right neural net configuration and supplying truly random inputs in the right places.
Considering the simplicity of the desired operations, how hard would be to build one around a super-simple microcontroller yourself? I bet a PIC could do it just fine (or even a PIC packaged up as a Basic Stamp for even easier programming). You just need a driver chip for the keypad rows/columns that will interface as input to the controller, and a very simple LCD display.
Every statement you make is false.
It is sensationalist to say that the scientists were surprised - if they work in this field surely they've seen ersults similar in meaning before.
It is sensationalist to say the machines learned on their own. Perhaps you don't understand the scope of humans call "learning". Learning within a small fixed set of parameters set up by a human isn't learning. Notice that humans don't face a halting problem or break into infinite loops when we learn wholly new things (well, except maybe a few in asylums).
Thinking is pretty clearly defined in my head. Simple computer software making selections based on the performance of other simple computer software doesn't constitute thinking to me, it constitutes executing my explicit instructions.
Ok you can throw out my semantics on "Cheating", but the fact is that this design process they used is impractical in any real situation. It will always evolve to be too specialized. You can't assume that you can fix this by immersing it in diverse environments during training - the technology only becomes truly useful when it hypothetical goes beyond our ability to define parameters for it in the first place.
Oh yes, very powerful, fear the radio receiver.
I'm of the opinion (not that I have any good basis for this, it's a layman's opinion) that the human brain is basically operating like a digital neural net with an incredibly good RNG available all over the place. I have a feeling it's really the randomness that makes us human and intelligent. That does put us beyond DFA and Turing. And I think quantum mechanics can indeed provide us this randomness.
Well, I disagree on evolutionary designs winning, but it's purely a matter of opinion at this point, and I respect yours. I listed only GEB precisely because I feel he makes very strong, subtle, and eloquent arguments for evolutionary AI in that book. I respect Hofstadter a *whole* lot, but I have my own opinions on these things.
Don't forget the old standby - the MIT LPF's rather huge paper from 1991 on the subject - it's still one of the better papers against patents around: Against Software Patents
Fuck you, I'm not a fucktard or a christian. Based on my layman's understanding of AI research, which at least includes reading works like GEB, I believe genetic algorithms are a flawed approach.
The article is sensationalist and irresponsible, as it talks of how the genetic algorithm "surprised the scientists", and how nobody knows how the circuit "figured out" one trace could act as an antenna.
The problem is that the non-tech-savvy of the world will read this and actually be made to believe these are thinking machines which are truly learning on their own. It conjures up images of a Matrix future.
I'm quite sure the scientists didn't find the results all that stunning. They ran random mutations and "evolved" an oscillator from the interconnections of 10 transistors. The algorithm of course *failed* to generate an oscillator, and instead cheated by picking up a nearby radiowave.
Nothing in the circuit "figured out" about antennas and radio waves - it was just random luck, much as any result in such an experiment is.
Some might argue with calling the cheating oscillator a failure. I disagree - I think it's a wonderful example of how far AI research has to go yet. What they wanted was an oscillator, presumably one that would work (were this a circuit designing machine in the real world) elsewhere outside the lab. The algorithm was too dumb to realize it's design won't be portable past the lab table.
I really don't think random mutation with selection is going to be the answer, if there's even an answer to be had. Computers are for automating, humans using them as tools are for innovating.
Even if your "company" is just one individual who knows a lot about JBoss, $5k/year is cheap. If your full-time job is being a JBoss developer/consultant, you will be charging clients per hour out the wazoo like all consultants, and raking in enough to make this amount trivial.
I think these JBoss guys have really hit the nail on the head when it comes to making an open source business model work financially. Personally, I dislike java as anything but a client-side language for a thin GUI, so JBoss is not my cup of tea - but the model is impressive and I'm proud of them.
Google could put long term cookies to good use if they do it responsibly. They can't tell which search result you click due to direct linking (although they could see which search results you pull from their cache) - but they could definitely bias your query results based on your previous queries. Example:
UserA has a pattern of computer-related querys
UserB has a pattern of photography-related querys
Both search for "lens effects".
UserA gets some information on computer graphics lens effects in software and algorithms ranked higher.
UserB gets some info on how to use real camera lenses to achieve neat lens effects in photography ranked higher.
Search perlmonks.org for Tilly's article on the subject a while back. It appears that by most states' labor laws, if you are an exempt, salaried, full-time professional - the company does in fact own all of your output, even when you're not at work, and they don't need a special contract to get these rights. If you work as unix sysadmin, and you develop and patent a new lawn sprinkler on your own time on the weekends, they can take your patent away from you. They certainly in this light own your output during work hours, which means they very well can try to enforce that you don't do things like use hotmail.
IMHO, for the most part technical training is for people who "don't get it" and probably won't much better after th course, or neccesary for products that are so non-intuitive you shouldn't be buying them to begin with.
I take company-paid training all the time, usually in a nice city. I'd much rather spend a week hanging out in a hotel on the company's dime and listening to some bonehead moron read their product's user manual to me very slowly during the day for a few hours than sit at my desk at work.
I agree on the processor front. My most recent home box is an Athlon XP, because I liked the price/performance ratio versus intel, and I can deal with the issues for the money. However, I still lean towards Intel for real servers that I depend on, and there's solid reason why.
On a processor level, AMD has Intel beat in the overall - the processor is just as good if not better for less money depending what part of the leapfrog cycle we're in. The problem is in the chipsets. Intel's motherboard chipsets have a history of being extremely well supported in all OS's and very good quality. Every AMD-compatible chipset I've ever used has had problems. You name it: VIA, SiS, even AMD's own MP760 or whatever it was called. Invariably all the chipsets available for AMD processors have problems. Sometimes it's hardware bugs, sometimes it's instability. More often than not, it's just that driver support isn't really there in Linux or Windows, or IRQ tables are kinda hosed.
In the end, it's the chipsets that keep me buying Intel for production. If Intel ever released a chipset that supported an AMD processor, I'd deploy the two in a heartbeat, but fat chance of that happening soon.