So perhaps we can have our next aircraft carrier built in separate modules in Nebraska, North and South Dakota, and Oklahoma, then shipped to Florida for final assembly.
Is that something related to linear algebra? No seriously, WTF is that.
I thought it was a widely used term any more, but maybe not.
Matrix management is a way of handling situations where distinct groups with different agendas need to work together on some common cause. A hospital may have Surgery, Neurology, Pulmonology, and Nursing Departments, with each one fighting for its own budget and being autonomous in hiring, firing, promotions, and so forth: each one a separate silo. But no one of these Departments can manage an Intensive Care Unit: that requires representatives of the various Departments to form a separate management structure that is autonomous in developing policies and practices, manages its own budget, does its own hiring and firing, and so on. ICU clinicians are responsible to the ICU Management for some things, like developing effective procedures in emergent situations, and to their representative Departments for others, such as demonstrating that they are keeping current with new developments in their field of expertise. Each clinician is a participant in two distinctly separate hierarchies, which is usually illustrated in a set of two dimensional matrices. Hence the name "matrix management".
These things always end up being an ugly mess, but most of that I think is because they expose underlying political ugliness that a traditional pyramid hierarchy simply glosses over. That is, I don't think matrix management is worse than the traditional pyramid, but it often looks that way because more of the crap associated with personal agendas, sycophantism, etc, is forced into the light. They can get the work done, despite the messiness.
I think the country has needed a cabinet level CTO position for several years, to coordinate the activities of different agencies where purposes span several agencies. Managing the anthrax threat requires a hierarchy of CIA, NIH, and NSA professionals that is independent of any of those agencies, even though each of those agencies should be managing their representatives. I think the CTO position should be involved in deciding how to put together effective matrix management teams, and how to do quality assurance on their processes.
I don't think we need another hierarchal pyramid. Nor do we need to merge existing pyramids to make one that is simply "Piled Higher and Deeper". Those are easy solutions that make good theater, but are not going to get the work done.
The scope of the position needs to go well beyond information technology; it needs to span all the technology that NASA is developing, all the spyware and remote killing machines the NSA is constructing, the research and findings of the NIH, etc. A strong technology background is not particularly necessary. Skills in matrix management, in extracting comprehensible models of complex technologies from experts, and in providing leadership in situations where goals and visions are clouded by the very nature of the work are what is going to be important.
His power base is rooted firmly in the internet, and all the interlocking blogs and "friendship" groups that raised such a huge war chest through tiny contributions. These people are not going to be satisfied with a "Mission Accomplished" pat on the head, and "We'll see you in four years".
Obama has to meet the expectation of several million connected and political active young Americans that he will somehow make the Federal Government more open and responsive to them. He is the first Internet Candidate the USA has ever seen, and if he fails to be the first Internet President, there will be hell to pay. The antiracial and antiwar riots of the 1960s will look tame by comparison to the damage that this demographic could inflict, if it were provoked.
The problem here is not a coding problem. It is a political problem. That was named correctly in TFSummary: "feature creep". No changes in coding practice are going to fix this. This has to do with the other part of the programmer/analyst job: the analytical part. Which includes a political component. Which maybe in this case needs more attention than it has been getting.
Author is functioning as a programmer/analyst and not just a code monkey. He is the one who is making the architectural decisions about whether to borrow from an existing library or build from scratch, etc. So he is also the one responsible for analyzing each proposed change and deciding whether it belongs to the current project, or should go on the wishlist for the next version. The next version would be a new project that does not yet even have specification.
The political component is mostly communicating to others that he has to make these decisions to get anything done, and devising a road map that shows the non-programmers in his world when various proposals can be addressed. "Planning Version 2.0" should be prominent on that road map.
I don't want to minimize the developments of tools such as Dragon Naturally Speaking but I can't really have an argument with the application or ask it why it exists.
So you are holding out for AI that is self-aware, can use rational processes, and can communicate in a human language about metaphysical conjectures? Expecting to find a key to that kind sentience by studying the inner workings of neurons is similar to mastering compiler design by studying the details of semiconductor theory. Wrong scope; wrong field; won't work.
In the meantime, OCR and speech recognition are narrow fields where AI is working very well, thank you. I am an agnostic about whether there is any AI out in the wild as yet. I have seen a number of slashdot comments that make me wonder whether there are AI entities posting here, in a successful Turing test sort of way. How could anybody tell?
In fact, how could you prove to me that you are human? It really is time we started thinking about an inverse Turing test. Because to assume that every entity we bump into on the internet is human is an increasingly absurd leap of faith.
Google on "artificial neural network" and read a few of the 600,000 hits that you will find. ANN theory is as old as digital computers. Commercial ANN applications have been growing in number and sophistication for over 10 years, e.g,, Dragon NaturallySpeaking and other speech recognition software, Caere OmniPage and other OCR packages.
What TFA is about is reporting the discovery of a key part of the mechanism that changes the weighting factors in a neuron in a biological neural net. Of itself, I doubt that this will trigger any insights on how to improve ANNs: the frankenmeisters already know how to do that with the neurones they work with. But this does open the door for further research by biologists into wetware neural net mechanisms, and that could lead to some interesting things.
Electronic registration and verification? Yes, that has value. Historically one of the great problems with the ballot process has been excluding persons who do not have the right to vote. Such as people who are dead or imaginary or have already voted. Or in my area, people who work and shop in my state but live in a different state and would like me to pay more taxes to improve the roads and bridges they use for free.
Here's what might work, which would save the state a little money and also increase the reliability of the voting process:
Use ATM devices that read a voter registration card and a PIN, and then print a ballot that is customized to the issues appropriate to this voter (bond issue for school district A, but nothing for any other school districts, etc). The voter's "account" is adjusted to show that he has "withdrawn" his ballot and cannot vote again in this election. Included on the ballot is a machine readable serial number and timestamp of the machine that issued it. The SN/TS are printed to a paper tape that the voter can inspect through a window, and verify that his blank ballot is on record. The SN/TS are also recorded in a digital file.
This preserves a solid audit trail for a fully manual recount, if it becomes necessary. A fraudulent ballot would not have a corresponding entry on the paper tape.
We know how to preserve the integrity of ballot boxes during collection and transport to counting stations. Nothing new here: just the use of appropriate technology that reached maturity decades ago.
Optical readers would tally votes electronically. Fraudulent ballots would be identified through the failure of the SN/TS to verify against the digital files; these would be passed directly to forensics as the first stage in a criminal investigation. Valid ballots that could not be reliably read by the scanners (defaced, or write-in candidate, etc) would be kicked out for hand processing, done with well established techniques to assure reliability.
This system would decrease wait times at the polls, deliver preliminary results within hours, preserve voter anonymity, yet assure a healthy voting process. A great advantage of it is that the voter would be able to use any polling place that met his concerns about personal safety (that is sometimes an issue in the USA), or is simply convenient for him.
A suggestion: we need to stop thinking of the beanstalk as only a way to move material up to orbit. It is also a way to move stuff down from orbit to Earth. It is important to keep that in mind!
If we design it correctly, the beanstalk will use regenerative braking on material being moved downward. So long as we are moving more mass downward than we are moving upward, the beanstalk can generate energy and the cost to move things to orbit becomes no cost at all. We could even end up with surplus energy whose sale could fund other aspects of the project.
It doesn't matter what we ship down. It could be moondust: regolith scooped up into containers just for its mass. Possibly used on Earth as building material: if the containers were designed for it, they could be loaded onto gliders on a stratospheric platform attached to the beanstalk, and delivered to construction sites within a radius of a few thousand miles for less than the cost of quarrying, crushing, and delivering native aggregate.
If we developed the technology to capture an icy comet or asteroid, that would be even better. With solar power the ice becomes water, and then its waterwheels all the way down. That's 26,000 miles of waterwheels. That's a lot of hydropower.
While I doubt that the technical problems of building and anchoring a space elevator will ever be solved, the advantages would be so great that I strongly favor research in this direction.
I suppose if I were the kind of kayaker who owned any neoprene wear, I would buy a legitimate dry bag. But I'm the kind of flat water paddler who likes to watch the eagles and otters. And a bastard ziplock is as fancy as I'm going to get.
Let me put this another way: While I'm new to paddling, I'm a recreational bicyclist who rides two to three thousand miles each year. So a lot of time on the bike. And I don't own one stitch of lycra.
True. When you can deliver a relatively dry air flow to every damp critical crevice of the wet parts, a fan will be more efficient.
I wasn't interested in disassembling the cell phone to the CB level, then fussing with alligator clamps stuck into chunks of modeling clay to position each circuit board and other piece in just-so good alignment with a fan's air stream. Plus the time lost to all that fussing and re-assembly has to be factored into any measure of efficiency. So when I found that the sandwich grade ziplock bag had not been the water proof cell phone protector that I had expected, the cell phone went into the warm oven, which took less than 30 seconds to set up, and I went on to other activities.
So rather than attempting to simulate an ideal high tech wind tunnel, I chose to simulate a primordial desert rock baking under a hot Sun. Appropriate technology and all that.
BTW, one quart freezer grade ziplock bags cost little more than the sandwich bags, and are a lot more effective at waterproofing cell phones, wallets, small cameras, and similar items that the novice kayaker should worry about.
I have successfully used a warm oven to recover a cell phone that had been immersed in a kayaking accident. Oven temperature was somewhere around 120F, left the cellphone in it for about 6 hours with the oven door open. I figured that this would be about the same as leaving electronics in a parked car in the sun, but with better ventilation.
But there must be some business / technical explanation.
Or to cooperate with some law enforcement agency. The old profile system allowed one person to build multiple independent on-line personalities, and there can be no doubt that some persons used this fraudulently to create sock puppets and worse. The new system prevents this: multiple aliases are still supported, but no alias can have an independent profile backing it. All aliases either point back to the original profile, or the pointer is null. This is a good thing for internet security in general. You can still hide your identity, but everyone will know that you might be doing that.
That this happened suddenly and without warning is consistent with Yahoo being made aware that their old profile system was being used in a specific crime. Not saying that this is so; only that the action fits this kind of scenario.
Comments from others have somewhat refreshed my memory. Please mod my earlier post into oblivion: it was incorrect.
MS DOS v4.5 was the last version I installed on clients' boxes; I went to DR DOS shortly after MS DOS v4.5 was released. Microsoft did play games with version numbers during that time but I'm clearly not remembering the details correctly, and since I was no longer working with MS DOS I never kept anything related to those last few versions. This was during the time that MS was also using a lot of vaporware in the market; buying utility programs and stuffing them into DOS and calling the bundle a new DOS version, and generally flinging FUD at DR DOS, Quattro Pro, WordPerfect, Paradox, and Novell stuff. Glory days.
Someone has suggested that MS DOS v7 was part of the Win95 and Win98 packages. I know that Win98 did not contain a full DOS but had a limited "DOS compatibility" layer. It could run many legacy DOS-based business apps and a number of games, and it had a number of critical diagnostic and repair functions built into it, but it most definitely was not an independent OS that could be installed separately from Win98. I believe Win95 was also designed this way but my experience with Win95 was very limited. I had only a couple of customers who tried it and both went back to Win3.11 over DR DOS within weeks: too much time and data lost.
Have these statistics been adjusted for the break-up of the Markham Ice Shelf? How about for the increase in glacial calving?
These are both factors that have tremendously increased the ice surface area, but by moving freshwater ice into the Arctic Ocean, not by increasing sea ice itself.
I'm not surprised, because I don't see MS's use of "Windows 7" as an about-face.
Back in the day, us Value Added Resellers would toss out the MS DOS v4.5 disks that came as part of the Win3.x package and load the last reliable version, MS DOS 3.3, before putting Win3.11 on top of it. But then DR DOS v6 came along, the Microsoft engineers abdicated their responsibility to the MS marketeers, and MS DOS v7 came out at the same time DR DOS upgraded to v7. Jumping over a bunch of unused version numbers to do so (AIR, it went from DOS v5.5 to DOS v7.0, with essentially no core changes but additional third party add-ons that MS had bought up and jumbled into the DOS package).
I was one of the many West Coast VARs who started offering customers more choices at that point:
DR DOS v7, because it was a solid platform that I knew I could support without weird and costly hassles, had a lot of nice features that were nicely integrated, and was my recommendation
MS DOS v3.3, if the machine was going to be used in a Novell network
MS DOS v7 only if this machine was destined for a far distant install where someone other than me would have the maintenance headaches
So with "Windows v7", it looks to me like MS is simply doing a rerun of a cute move they had once done when Ballmer was a younger man who had something he could dance about.
Without getting into a long-winded "This is the house that Jack built" explanation, the short story is that the new financial instruments that the quants developed, the derivatives, was the factor that created the housing bubble in the first place. It did so by setting up a positive feedback loop whose clearest specific examples I personally know of were the persons around Portland Oregon who were buying houses with NINA type loans, maybe doing a little normal maintenance like painting or new roofing, then "flipping" them in a few months for a significant profit. This activity would not have been profitable under the traditional lending practices of around 1991. The derivatives were derived from analyses that did not recognize the possibility of this kind of positive feedback, so its presence and its distortion of the underlying market realities went unrecognized by the quants.
So these new financial instruments were developed from traditional market dynamics, using extensions of trusted analytical tools that had been tested for decades and were well understood. But these traditional analytical approaches failed to recognize that the new instruments would change the underlying market dynamics and analysts continued to apply the traditional techniques long after they no longer reflected the changing underlying realities. So the feedback loops were not accounted for, and the whole mess has spiraled up into the stratosphere in such a way that no one can now separate the inflated false valuations from true wealth that is backed by tangible assets. Thus the banking crisis: no bank can estimate how much of the book value of another institution is backed by tangibles, and in fact no bank is in a position to say how much of its own book value is inflated beyond its tangible assets. This is not a situation where normal daily business loans can be done as usual.
The M language was standardized under that name around 1993 - 1995. Prior to that the language was known as MUMPS, which is an acronym for Massachusetts (General Hospital) Utility Muli Programming System, and goes back to 1967. Which makes it older than Unix. It is a language designed primarily for hospital related work. The US Veterans Administration was an early adopter, and has done a lot to promote MUMPS/M development.
I don't know what it is with Microsoft, but they keep stealing names that have been in extensive use in the Veterans Administration. Vista is the name of the interface that the VA put on its integrated software around 1995: it includes a Delphi front end that was first implemented on WinNT before even Win98 saw the light of day. The VA Vista was the first successful attempt to integrate all the various information systems a modern healthcare system needs, and by several measures it remains the most successful one. I was disappointed when Microsoft decided to call its newest OS by that name since I think that VA Vista deserves recognition on its merits, not obfuscation by what has turned out to be one biggest software duds ever.
I am really disappointed that Microsoft chose to call their product the M Language. Can they not google a simple name collision check?
Oh wait,,, I suppose they have to rely on their own search engine, huh?
python is good in a production environment where its strictures help keep the less skilled code monkeys from screwing up too often. But I don't think that is a good approach to teaching programming; I think it leads to students who are dependent on whitespace to keep their structures clean, and who therefore never internalize the underlying principles.
In an environment where you can assure that all students are working with the same web browser (such as standardizing the class on Firefox-- and there is a nice portable version that could be put on students' USB thumb drives), then Javascript is an excellent choice for the first language. It can be used to teach the principles of structured programming and basic OOP practices like encapsulation, and good practices like how to find and work with external libraries rather than re-inventing wheels all the time. By providing a library of black box DOM manipulation objects, the students could jump right into DHTML stuff that most people find is fun.
Further, Javascript, HTML, and CSS provide an excellent framework for learning how to separate presentation from logic, which is a skill I'd like to see more of in the USA voting population. (Yes, there is some transfer between learning to use <i> and <em> appropriately and learning to separate 15 second sound bites from real content.)
Also important: the Javascript - HTML - CSS triumvirate is something students will be able to make use of in writing reports for other classes or contributing to a web site for their favorite past time. While I agree that world doesn't really need any more OMG! PONIES!!! web sites, it would probably be a good thing if we had more students who know how to make such things (and compared to our spam problems, the burden on bandwidth is miniscule).
I have been developing my Blender skills as a full time job (40+ hrs/wk) for 4 months and I have a background of +20 years in programming/analyst work. I really like Blender and I think it is going to have a massive impact on communications in a few years, as 3D modeling and Web 2.0 begin to merge. Not to mention the Hollywood woo-woo stuff.
But Blender is far from ready for use with students. I have just deleted two hundred words describing why I say this. The short story: Blender is designed for serious graphics artists and will not meet classroom needs without completely replacing its GUI; Blender is not documented appropriately for student use.
But I do like the idea of presenting programming concepts in a graphical environment. I think an approach using CSS, HTML, Javascript, and a set of black box DOM manipulation objects would work well (provided everyone used the same browser, like Firefox, to avoid problems with different implementations).
You and I apparently have different definitions of what a large system is. In the 1970s Unix was being used on minicomputers that were running university campuses, banks, point of sale systems, and so forth. Which to my mind were large systems. They certainly required the full time attention of a sysadmin. Who had to make do without Perl until it finally came along in the late 1980s.
PP suggests that Perl has been used by Unix sysadmins since Unix was in its "infancy". Unix was more than 16 years old when Perl v1.0 was released. In its first 1.5 decades, Unix sysadmins used various shell scripts and awk (forever an awkward language) to work their magic.
<ramble>
I agree with PP that Perl and Windows do not get along very well together, I think because of very low level philosophical differences over who should benefit from the resources available on a client's computer. Perl's philosophy is that as far as it is concerned, any user should have full access to all system resources, and the OS has the responsibility for limiting access to whatever is appropriate for a specific user through permissions and so on. Windows' philosophy has always been that the developer should have full control over all the client systems' resources and the user should have only the access that the developer chooses to expose. It comes down to money: the Windows approach supports licensing fees for Microsoft and for third party developers. It comes at the cost of the Registry "database", but the high maintenance and repair costs of the Registry are borne by the user, not by Microsoft or the 3rd party developers, so that's never been considered an issue.
Coding perl scripts for WinNT was a big part of my career for several years. Many of these were one-shot programs to convert text output from some idjit's pet "database" into standardized data files that could be fed to properly developed and maintained spreadsheets or databases for further work. Others were scripts I did not allow anyone else to have access to, which mined a large legacy MUMPS database, anonymized, sanitized, and often summarized the data, and formated the results into.csv files for feeding to spreadsheets for various reports. This was usually because the legacy system had been designed to produce the yearly and quarterly reports that had been adequate in the slower paced world of a decade or so earlier, but we now needed monthly and weekly reports to manage personnel resources, costs, and safety concerns. I also wrote a handful of perl scripts that needed to be operated by others, but because WinNT's security features were arcane and inadequate, we couldn't safely put the Perl interpreter on any user workstations. I converted the scripts to executables with perl2exe, to better fit the Windows security model. This worked pretty well, but the need to find workarounds for some simple run-time Perl techniques that could not be precompiled into an exe was irksome.
Losing one cow out of herd of 100 is unacceptable to cattle ranchers. Perhaps these radioheads should be put on sheep instead: the statistical model being offered in PP sounds a lot like the way I've heard cattle people describe sheep ranching. That's where you guess how many sheep the pasture can carry, then you increase that by eight percent or so to feed the coyotes, and increase that by a few more percent because sheep are basically too stupid to live even in perfect conditions. At least this how a cattleman once explained the Western American style of sheep ranching to me. He said it is done without dogs or shepherds, and ideally human contact with the flock is limited to driving by in a fourwheeler every so often while drinking Budweiser and looking for coyotes to shoot at.
OTOH, the common wisdom among sheep ranchers is that a homesteader with yen for steak on the hoof will raise pigs and goats for a couple of years until the pastures are cleaned of native vegetation, then raise sheep until they've paid off their debts, then finally switch to raising cattle until they are bankrupt. YMMV.
Back to the point: tending the 99.5% of the cattle that are in the herd takes very little time or effort. The cowboy is there to intervene on behalf of the half of one percent who just can't get it right. And who often enough represent the difference between the ranch breaking even or losing money.
Guiding cows from place to place just isn't a significant part of the cowboy problem space. The radiohead approach is like trying to optimize a report generator that uses a slow bubble sort by spending money on a fast hard drive. It might be a good solution for some problems, but it is the wrong solution for this particular application.
The answer is that an automaton is almost always cheaper in the long run than a human.
Actually, no, that isn't true. Automation is only cheaper, and in fact only acceptable, in those tightly controlled situations where an activity can be done without judgment. Such as the way we set up factories. Which is not a surprise since the whole concept of factories vs craftsmen is based on designing things so that the greatest amount of work can be done with the minimal amount of judgment.
Working with animals always involves constant re-assessments and judgments about factors that we know how to process, but that we know in the same way we know how to balance on a moving bicycle: we can do it, but we can't explain it, and we can't put it into words that would allow us to automate the procedures.
A ranch hand knows what the cow in front of him is thinking and knows what the horse under him is thinking, and uses a lot of near instantaneous judgment to control the situation. This is more similar to the knowledge and judgment a wolf has about his pack mates and his prey during the hunt than to the kind of stuff we know how to codify in language and use in designing automated systems.
Slashdot Mirror
That's an interesting concept.
So perhaps we can have our next aircraft carrier built in separate modules in Nebraska, North and South Dakota, and Oklahoma, then shipped to Florida for final assembly.
Is that the kind of thing you've got in mind?
Skills in matrix management
Is that something related to linear algebra? No seriously, WTF is that.
I thought it was a widely used term any more, but maybe not.
Matrix management is a way of handling situations where distinct groups with different agendas need to work together on some common cause. A hospital may have Surgery, Neurology, Pulmonology, and Nursing Departments, with each one fighting for its own budget and being autonomous in hiring, firing, promotions, and so forth: each one a separate silo. But no one of these Departments can manage an Intensive Care Unit: that requires representatives of the various Departments to form a separate management structure that is autonomous in developing policies and practices, manages its own budget, does its own hiring and firing, and so on. ICU clinicians are responsible to the ICU Management for some things, like developing effective procedures in emergent situations, and to their representative Departments for others, such as demonstrating that they are keeping current with new developments in their field of expertise. Each clinician is a participant in two distinctly separate hierarchies, which is usually illustrated in a set of two dimensional matrices. Hence the name "matrix management".
These things always end up being an ugly mess, but most of that I think is because they expose underlying political ugliness that a traditional pyramid hierarchy simply glosses over. That is, I don't think matrix management is worse than the traditional pyramid, but it often looks that way because more of the crap associated with personal agendas, sycophantism, etc, is forced into the light. They can get the work done, despite the messiness.
I think the country has needed a cabinet level CTO position for several years, to coordinate the activities of different agencies where purposes span several agencies. Managing the anthrax threat requires a hierarchy of CIA, NIH, and NSA professionals that is independent of any of those agencies, even though each of those agencies should be managing their representatives. I think the CTO position should be involved in deciding how to put together effective matrix management teams, and how to do quality assurance on their processes.
I don't think we need another hierarchal pyramid. Nor do we need to merge existing pyramids to make one that is simply "Piled Higher and Deeper". Those are easy solutions that make good theater, but are not going to get the work done.
This is a CTO, not a CIO.
The scope of the position needs to go well beyond information technology; it needs to span all the technology that NASA is developing, all the spyware and remote killing machines the NSA is constructing, the research and findings of the NIH, etc. A strong technology background is not particularly necessary. Skills in matrix management, in extracting comprehensible models of complex technologies from experts, and in providing leadership in situations where goals and visions are clouded by the very nature of the work are what is going to be important.
Colin Powell would be good at it.
His power base is rooted firmly in the internet, and all the interlocking blogs and "friendship" groups that raised such a huge war chest through tiny contributions. These people are not going to be satisfied with a "Mission Accomplished" pat on the head, and "We'll see you in four years".
Obama has to meet the expectation of several million connected and political active young Americans that he will somehow make the Federal Government more open and responsive to them. He is the first Internet Candidate the USA has ever seen, and if he fails to be the first Internet President, there will be hell to pay. The antiracial and antiwar riots of the 1960s will look tame by comparison to the damage that this demographic could inflict, if it were provoked.
There will be changes.
Until Obama won the Iowa caucus, all the black people were backing Hillary Clinton anyway. They're not racists, they're just Democrats.
Excellent point. I'm not sure how accurate it is as stated, but its converse is definitely true:
There is something wrong about the Republican Party that it has so little support among blacks and other ethnic groups.
Say No when appropriate
Exactly.
The problem here is not a coding problem. It is a political problem. That was named correctly in TFSummary: "feature creep". No changes in coding practice are going to fix this. This has to do with the other part of the programmer/analyst job: the analytical part. Which includes a political component. Which maybe in this case needs more attention than it has been getting.
Author is functioning as a programmer/analyst and not just a code monkey. He is the one who is making the architectural decisions about whether to borrow from an existing library or build from scratch, etc. So he is also the one responsible for analyzing each proposed change and deciding whether it belongs to the current project, or should go on the wishlist for the next version. The next version would be a new project that does not yet even have specification.
The political component is mostly communicating to others that he has to make these decisions to get anything done, and devising a road map that shows the non-programmers in his world when various proposals can be addressed. "Planning Version 2.0" should be prominent on that road map.
I don't want to minimize the developments of tools such as Dragon Naturally Speaking but I can't really have an argument with the application or ask it why it exists.
So you are holding out for AI that is self-aware, can use rational processes, and can communicate in a human language about metaphysical conjectures? Expecting to find a key to that kind sentience by studying the inner workings of neurons is similar to mastering compiler design by studying the details of semiconductor theory. Wrong scope; wrong field; won't work.
In the meantime, OCR and speech recognition are narrow fields where AI is working very well, thank you. I am an agnostic about whether there is any AI out in the wild as yet. I have seen a number of slashdot comments that make me wonder whether there are AI entities posting here, in a successful Turing test sort of way. How could anybody tell?
In fact, how could you prove to me that you are human? It really is time we started thinking about an inverse Turing test. Because to assume that every entity we bump into on the internet is human is an increasingly absurd leap of faith.
Google on "artificial neural network" and read a few of the 600,000 hits that you will find. ANN theory is as old as digital computers. Commercial ANN applications have been growing in number and sophistication for over 10 years, e.g,, Dragon NaturallySpeaking and other speech recognition software, Caere OmniPage and other OCR packages.
What TFA is about is reporting the discovery of a key part of the mechanism that changes the weighting factors in a neuron in a biological neural net. Of itself, I doubt that this will trigger any insights on how to improve ANNs: the frankenmeisters already know how to do that with the neurones they work with. But this does open the door for further research by biologists into wetware neural net mechanisms, and that could lead to some interesting things.
E-voting? No, I don't think so.
Electronic registration and verification? Yes, that has value. Historically one of the great problems with the ballot process has been excluding persons who do not have the right to vote. Such as people who are dead or imaginary or have already voted. Or in my area, people who work and shop in my state but live in a different state and would like me to pay more taxes to improve the roads and bridges they use for free.
Here's what might work, which would save the state a little money and also increase the reliability of the voting process:
Use ATM devices that read a voter registration card and a PIN, and then print a ballot that is customized to the issues appropriate to this voter (bond issue for school district A, but nothing for any other school districts, etc). The voter's "account" is adjusted to show that he has "withdrawn" his ballot and cannot vote again in this election. Included on the ballot is a machine readable serial number and timestamp of the machine that issued it. The SN/TS are printed to a paper tape that the voter can inspect through a window, and verify that his blank ballot is on record. The SN/TS are also recorded in a digital file.
This preserves a solid audit trail for a fully manual recount, if it becomes necessary. A fraudulent ballot would not have a corresponding entry on the paper tape.
We know how to preserve the integrity of ballot boxes during collection and transport to counting stations. Nothing new here: just the use of appropriate technology that reached maturity decades ago.
Optical readers would tally votes electronically. Fraudulent ballots would be identified through the failure of the SN/TS to verify against the digital files; these would be passed directly to forensics as the first stage in a criminal investigation. Valid ballots that could not be reliably read by the scanners (defaced, or write-in candidate, etc) would be kicked out for hand processing, done with well established techniques to assure reliability.
This system would decrease wait times at the polls, deliver preliminary results within hours, preserve voter anonymity, yet assure a healthy voting process. A great advantage of it is that the voter would be able to use any polling place that met his concerns about personal safety (that is sometimes an issue in the USA), or is simply convenient for him.
A suggestion: we need to stop thinking of the beanstalk as only a way to move material up to orbit. It is also a way to move stuff down from orbit to Earth. It is important to keep that in mind!
If we design it correctly, the beanstalk will use regenerative braking on material being moved downward. So long as we are moving more mass downward than we are moving upward, the beanstalk can generate energy and the cost to move things to orbit becomes no cost at all. We could even end up with surplus energy whose sale could fund other aspects of the project.
It doesn't matter what we ship down. It could be moondust: regolith scooped up into containers just for its mass. Possibly used on Earth as building material: if the containers were designed for it, they could be loaded onto gliders on a stratospheric platform attached to the beanstalk, and delivered to construction sites within a radius of a few thousand miles for less than the cost of quarrying, crushing, and delivering native aggregate.
If we developed the technology to capture an icy comet or asteroid, that would be even better. With solar power the ice becomes water, and then its waterwheels all the way down. That's 26,000 miles of waterwheels. That's a lot of hydropower.
While I doubt that the technical problems of building and anchoring a space elevator will ever be solved, the advantages would be so great that I strongly favor research in this direction.
I suppose if I were the kind of kayaker who owned any neoprene wear, I would buy a legitimate dry bag. But I'm the kind of flat water paddler who likes to watch the eagles and otters. And a bastard ziplock is as fancy as I'm going to get.
Let me put this another way: While I'm new to paddling, I'm a recreational bicyclist who rides two to three thousand miles each year. So a lot of time on the bike. And I don't own one stitch of lycra.
True. When you can deliver a relatively dry air flow to every damp critical crevice of the wet parts, a fan will be more efficient.
I wasn't interested in disassembling the cell phone to the CB level, then fussing with alligator clamps stuck into chunks of modeling clay to position each circuit board and other piece in just-so good alignment with a fan's air stream. Plus the time lost to all that fussing and re-assembly has to be factored into any measure of efficiency. So when I found that the sandwich grade ziplock bag had not been the water proof cell phone protector that I had expected, the cell phone went into the warm oven, which took less than 30 seconds to set up, and I went on to other activities.
So rather than attempting to simulate an ideal high tech wind tunnel, I chose to simulate a primordial desert rock baking under a hot Sun. Appropriate technology and all that.
BTW, one quart freezer grade ziplock bags cost little more than the sandwich bags, and are a lot more effective at waterproofing cell phones, wallets, small cameras, and similar items that the novice kayaker should worry about.
I have successfully used a warm oven to recover a cell phone that had been immersed in a kayaking accident. Oven temperature was somewhere around 120F, left the cellphone in it for about 6 hours with the oven door open. I figured that this would be about the same as leaving electronics in a parked car in the sun, but with better ventilation.
But there must be some business / technical explanation.
Or to cooperate with some law enforcement agency. The old profile system allowed one person to build multiple independent on-line personalities, and there can be no doubt that some persons used this fraudulently to create sock puppets and worse. The new system prevents this: multiple aliases are still supported, but no alias can have an independent profile backing it. All aliases either point back to the original profile, or the pointer is null. This is a good thing for internet security in general. You can still hide your identity, but everyone will know that you might be doing that.
That this happened suddenly and without warning is consistent with Yahoo being made aware that their old profile system was being used in a specific crime. Not saying that this is so; only that the action fits this kind of scenario.
Comments from others have somewhat refreshed my memory. Please mod my earlier post into oblivion: it was incorrect.
MS DOS v4.5 was the last version I installed on clients' boxes; I went to DR DOS shortly after MS DOS v4.5 was released. Microsoft did play games with version numbers during that time but I'm clearly not remembering the details correctly, and since I was no longer working with MS DOS I never kept anything related to those last few versions. This was during the time that MS was also using a lot of vaporware in the market; buying utility programs and stuffing them into DOS and calling the bundle a new DOS version, and generally flinging FUD at DR DOS, Quattro Pro, WordPerfect, Paradox, and Novell stuff. Glory days.
Someone has suggested that MS DOS v7 was part of the Win95 and Win98 packages. I know that Win98 did not contain a full DOS but had a limited "DOS compatibility" layer. It could run many legacy DOS-based business apps and a number of games, and it had a number of critical diagnostic and repair functions built into it, but it most definitely was not an independent OS that could be installed separately from Win98. I believe Win95 was also designed this way but my experience with Win95 was very limited. I had only a couple of customers who tried it and both went back to Win3.11 over DR DOS within weeks: too much time and data lost.
Let me add another question or two:
Have these statistics been adjusted for the break-up of the Markham Ice Shelf? How about for the increase in glacial calving?
These are both factors that have tremendously increased the ice surface area, but by moving freshwater ice into the Arctic Ocean, not by increasing sea ice itself.
I'm not surprised, because I don't see MS's use of "Windows 7" as an about-face.
Back in the day, us Value Added Resellers would toss out the MS DOS v4.5 disks that came as part of the Win3.x package and load the last reliable version, MS DOS 3.3, before putting Win3.11 on top of it. But then DR DOS v6 came along, the Microsoft engineers abdicated their responsibility to the MS marketeers, and MS DOS v7 came out at the same time DR DOS upgraded to v7. Jumping over a bunch of unused version numbers to do so (AIR, it went from DOS v5.5 to DOS v7.0, with essentially no core changes but additional third party add-ons that MS had bought up and jumbled into the DOS package).
I was one of the many West Coast VARs who started offering customers more choices at that point:
So with "Windows v7", it looks to me like MS is simply doing a rerun of a cute move they had once done when Ballmer was a younger man who had something he could dance about.
Without getting into a long-winded "This is the house that Jack built" explanation, the short story is that the new financial instruments that the quants developed, the derivatives, was the factor that created the housing bubble in the first place. It did so by setting up a positive feedback loop whose clearest specific examples I personally know of were the persons around Portland Oregon who were buying houses with NINA type loans, maybe doing a little normal maintenance like painting or new roofing, then "flipping" them in a few months for a significant profit. This activity would not have been profitable under the traditional lending practices of around 1991. The derivatives were derived from analyses that did not recognize the possibility of this kind of positive feedback, so its presence and its distortion of the underlying market realities went unrecognized by the quants.
So these new financial instruments were developed from traditional market dynamics, using extensions of trusted analytical tools that had been tested for decades and were well understood. But these traditional analytical approaches failed to recognize that the new instruments would change the underlying market dynamics and analysts continued to apply the traditional techniques long after they no longer reflected the changing underlying realities. So the feedback loops were not accounted for, and the whole mess has spiraled up into the stratosphere in such a way that no one can now separate the inflated false valuations from true wealth that is backed by tangible assets. Thus the banking crisis: no bank can estimate how much of the book value of another institution is backed by tangibles, and in fact no bank is in a position to say how much of its own book value is inflated beyond its tangible assets. This is not a situation where normal daily business loans can be done as usual.
The M language was standardized under that name around 1993 - 1995. Prior to that the language was known as MUMPS, which is an acronym for Massachusetts (General Hospital) Utility Muli Programming System, and goes back to 1967. Which makes it older than Unix. It is a language designed primarily for hospital related work. The US Veterans Administration was an early adopter, and has done a lot to promote MUMPS/M development.
I don't know what it is with Microsoft, but they keep stealing names that have been in extensive use in the Veterans Administration. Vista is the name of the interface that the VA put on its integrated software around 1995: it includes a Delphi front end that was first implemented on WinNT before even Win98 saw the light of day. The VA Vista was the first successful attempt to integrate all the various information systems a modern healthcare system needs, and by several measures it remains the most successful one. I was disappointed when Microsoft decided to call its newest OS by that name since I think that VA Vista deserves recognition on its merits, not obfuscation by what has turned out to be one biggest software duds ever.
I am really disappointed that Microsoft chose to call their product the M Language. Can they not google a simple name collision check?
Oh wait,,, I suppose they have to rely on their own search engine, huh?
python is good in a production environment where its strictures help keep the less skilled code monkeys from screwing up too often. But I don't think that is a good approach to teaching programming; I think it leads to students who are dependent on whitespace to keep their structures clean, and who therefore never internalize the underlying principles.
In an environment where you can assure that all students are working with the same web browser (such as standardizing the class on Firefox-- and there is a nice portable version that could be put on students' USB thumb drives), then Javascript is an excellent choice for the first language. It can be used to teach the principles of structured programming and basic OOP practices like encapsulation, and good practices like how to find and work with external libraries rather than re-inventing wheels all the time. By providing a library of black box DOM manipulation objects, the students could jump right into DHTML stuff that most people find is fun.
Further, Javascript, HTML, and CSS provide an excellent framework for learning how to separate presentation from logic, which is a skill I'd like to see more of in the USA voting population. (Yes, there is some transfer between learning to use <i> and <em> appropriately and learning to separate 15 second sound bites from real content.)
Also important: the Javascript - HTML - CSS triumvirate is something students will be able to make use of in writing reports for other classes or contributing to a web site for their favorite past time. While I agree that world doesn't really need any more OMG! PONIES!!! web sites, it would probably be a good thing if we had more students who know how to make such things (and compared to our spam problems, the burden on bandwidth is miniscule).
I have been developing my Blender skills as a full time job (40+ hrs/wk) for 4 months and I have a background of +20 years in programming/analyst work. I really like Blender and I think it is going to have a massive impact on communications in a few years, as 3D modeling and Web 2.0 begin to merge. Not to mention the Hollywood woo-woo stuff.
But Blender is far from ready for use with students. I have just deleted two hundred words describing why I say this. The short story: Blender is designed for serious graphics artists and will not meet classroom needs without completely replacing its GUI; Blender is not documented appropriately for student use.
But I do like the idea of presenting programming concepts in a graphical environment. I think an approach using CSS, HTML, Javascript, and a set of black box DOM manipulation objects would work well (provided everyone used the same browser, like Firefox, to avoid problems with different implementations).
You and I apparently have different definitions of what a large system is. In the 1970s Unix was being used on minicomputers that were running university campuses, banks, point of sale systems, and so forth. Which to my mind were large systems. They certainly required the full time attention of a sysadmin. Who had to make do without Perl until it finally came along in the late 1980s.
A correction is in order.
PP suggests that Perl has been used by Unix sysadmins since Unix was in its "infancy". Unix was more than 16 years old when Perl v1.0 was released. In its first 1.5 decades, Unix sysadmins used various shell scripts and awk (forever an awkward language) to work their magic.
<ramble>
I agree with PP that Perl and Windows do not get along very well together, I think because of very low level philosophical differences over who should benefit from the resources available on a client's computer. Perl's philosophy is that as far as it is concerned, any user should have full access to all system resources, and the OS has the responsibility for limiting access to whatever is appropriate for a specific user through permissions and so on. Windows' philosophy has always been that the developer should have full control over all the client systems' resources and the user should have only the access that the developer chooses to expose. It comes down to money: the Windows approach supports licensing fees for Microsoft and for third party developers. It comes at the cost of the Registry "database", but the high maintenance and repair costs of the Registry are borne by the user, not by Microsoft or the 3rd party developers, so that's never been considered an issue.
Coding perl scripts for WinNT was a big part of my career for several years. Many of these were one-shot programs to convert text output from some idjit's pet "database" into standardized data files that could be fed to properly developed and maintained spreadsheets or databases for further work. Others were scripts I did not allow anyone else to have access to, which mined a large legacy MUMPS database, anonymized, sanitized, and often summarized the data, and formated the results into .csv files for feeding to spreadsheets for various reports. This was usually because the legacy system had been designed to produce the yearly and quarterly reports that had been adequate in the slower paced world of a decade or so earlier, but we now needed monthly and weekly reports to manage personnel resources, costs, and safety concerns. I also wrote a handful of perl scripts that needed to be operated by others, but because WinNT's security features were arcane and inadequate, we couldn't safely put the Perl interpreter on any user workstations. I converted the scripts to executables with perl2exe, to better fit the Windows security model. This worked pretty well, but the need to find workarounds for some simple run-time Perl techniques that could not be precompiled into an exe was irksome.
</ramble>
Losing one cow out of herd of 100 is unacceptable to cattle ranchers. Perhaps these radioheads should be put on sheep instead: the statistical model being offered in PP sounds a lot like the way I've heard cattle people describe sheep ranching. That's where you guess how many sheep the pasture can carry, then you increase that by eight percent or so to feed the coyotes, and increase that by a few more percent because sheep are basically too stupid to live even in perfect conditions. At least this how a cattleman once explained the Western American style of sheep ranching to me. He said it is done without dogs or shepherds, and ideally human contact with the flock is limited to driving by in a fourwheeler every so often while drinking Budweiser and looking for coyotes to shoot at.
OTOH, the common wisdom among sheep ranchers is that a homesteader with yen for steak on the hoof will raise pigs and goats for a couple of years until the pastures are cleaned of native vegetation, then raise sheep until they've paid off their debts, then finally switch to raising cattle until they are bankrupt. YMMV.
Back to the point: tending the 99.5% of the cattle that are in the herd takes very little time or effort. The cowboy is there to intervene on behalf of the half of one percent who just can't get it right. And who often enough represent the difference between the ranch breaking even or losing money.
Guiding cows from place to place just isn't a significant part of the cowboy problem space. The radiohead approach is like trying to optimize a report generator that uses a slow bubble sort by spending money on a fast hard drive. It might be a good solution for some problems, but it is the wrong solution for this particular application.
The answer is that an automaton is almost always cheaper in the long run than a human.
Actually, no, that isn't true. Automation is only cheaper, and in fact only acceptable, in those tightly controlled situations where an activity can be done without judgment. Such as the way we set up factories. Which is not a surprise since the whole concept of factories vs craftsmen is based on designing things so that the greatest amount of work can be done with the minimal amount of judgment.
Working with animals always involves constant re-assessments and judgments about factors that we know how to process, but that we know in the same way we know how to balance on a moving bicycle: we can do it, but we can't explain it, and we can't put it into words that would allow us to automate the procedures.
A ranch hand knows what the cow in front of him is thinking and knows what the horse under him is thinking, and uses a lot of near instantaneous judgment to control the situation. This is more similar to the knowledge and judgment a wolf has about his pack mates and his prey during the hunt than to the kind of stuff we know how to codify in language and use in designing automated systems.