First, is anyone from GNOME talking to anyone from GNUstep? While approve of the Bonobo architecture ideas, it looks to me like the two are doing very similar things. (And WindowMaker rocks my world.) Second -- does anyone else think that an object-based OS doesn't really sound like UNIX?
-_Quinn
This first is the obvious problem -- understanding natural language seems to be an AI-complete problem. ('AI-complete' is a term borrowed from the normal classification of the difficulty of problems, and means that a problem is equivalent to and/or requires AI.) The other problem is more subtle -- define 'wedding photos' for me. People can and/want/ to describe the contents of something rather than where it resides. (I've seen people who prefer to search for something rather than type in a URL, even if they know it.) The problem of meta-data is very, very, big. How many people do you know who even enter so much meta-data as the title of a document when they save it? (I don't mean the name -- most WP's have an option to enter author/title/subject/keywords to help them find your document.) Even aside from formulating a meta-data standard, how do you/get/ that meta-data? Not all of it, and almost all of the useful bits, can't be -- except, perhaps, with AI.
"It is not about stopping progreess, it is about trying to encourage it."
your next sentence simple doesn't follow. The referenced article -- and most of the discussion recently -- about IP attacks the premise that IP still serves to encourage progress. Ignoring that for two sentences, effectively every inventor is somebody else's employee. The IP is not going to an individual/group -- it's going to the company (government institution, etc). If we want to encourage innovation by rewarding it, IP law ought to be changed so the individual/group gets more money. (And, incidentally, yes, I would invent. I'm a programmer (among other things) -- I do it all the time.)
The essence of patents is an attempt by the government to bribe those with secrets into exposing them by helping to exploit them commercially. This is distinct from the essence of copyright, by which it was decided that the creation of more works was beneficial to the people, and that the best way to promote more works was economic -- to allow the free market to handle it. With that in mind, the copyright system was set up to create such a market. The essence of the argument against copright is that the market (as consituted by, say, the MPAA) created by copyright law no longer -- or less than optimally -- benefits the people, and the same for patents, though the specifics, obviously, vary.
Hm. I've noticed stability issues with Netscape on RedHat 6.0, but not on any other platform. It may just be that the RH6.0 boxes don't have an up-to-date version of NS. I won't waste your time responding to some of your points, as they don't apply to me (I run NS under Slackware 7 w/o any stability issues at all, every window always maximized).
Point one and two are both good, though I've honestly never noticed image problems with NS in the course of my day-to-day browsing -- probably because of your fifth point. I'll take your word about JavaScript -- perhaps because NS's lack of functionality (odd, as they invented the language), I've not been very impressed by it. As for NS not handling straight HTML -- well, I don't do web design professionally, so I guess it doesn't matter that I've never seen it have any problems.:) Thanks for the reply -- I wanted to know what I was missing over here in Linux land until I get a build of Mozilla that runs fast enough not to hurt.
Actually, I had noticed this (the page reload), but not thought about it, because it doesn't affect me any -- with multiple desktops, the only apps I don't run full-screen are terminal emulators. Honestly, though, it's only in/very/ rare occasions (e.g. tables with ~3000 cells) that I've had to wait for the rendering any noticeable amount of time before the page finishes loading. Thanks.
Maybe this is a silly question, but what does IE 5.x do so much better than NS 4.7x? Why should I switch from NS 4.7 on Linux on IE 5.5 on Win9x? It's not more stable. It's certainly less secure. Am I just not getting around to enough sites on the web to notice Netscape's (supposedly) archaic renderer? For all the claims of being a faster browser, I have to admit to not caring at all -- if the page is complex enough that rendering time becomes significant, download time almost always overshadows it.
These are serious questions. Why should I change? What usability improvements does IE have? Why does NS 'suck so hard'?
Oddly enough, one of the 'big ideas' for improving air quality that was floated a while back and is (probably) wending its way toward production amount to installing a catalytic converter on the radiator/air intake, which has two benefits -- first, you clean up what's coming out the pipe in front of you, and second, burning cleaner air, your exhaust is cleaner. Some advocated using this technique to also remove non-car-produced pollutants from the air.
To become an engineer, you study the best way that it's been done before, and software engineering is no different. Since the best way it's been done before is UNIX, the ready availability of the blueprints for Linux makes it a natural choice for study. Mucking about with Linux is equivalent to moving I-beams around in the Twin Towers to see what happens -- except, naturally, quite a bit cheaper. The theory of software engineering -- computer science -- is more readily studied if the tools you use for the study are more readily available. The downtime associated with Windows means that it's not readily available...
All four of these must be fought in court, with the expectation of losing on #1. It'll be OK to lose #1, so long as Slashdot is forced by court injuction to perform the moderation. (That is, does not lose its common-carrier immunity.) The other three, likewise, must be fought, but expecting to take them to the Supreme Court for resolution.
Aside from moral reasons to fight -- I'm not familiar enough with Andover management to know if they're sufficient -- consider how many hits controversy generates. (Ever looked at the way ZDnet does business?) Furthermore, consider how many negative hits caving would generate. Thanks.
1: You bought MS's line, hook & sinker. The Macintosh (clone of a PARC system) made computers 'easy to use' -- the Mac just made them cheap enough to be feasible. Microsoft's Windows's jumped on the extremely cheap x86 hardware being developed to run VisiCalc and/or 1-2-3 faster and made a cheap knock-off of the Mac -- made 'easy to use' cheap enough to be popular. (I also take issue with 'easy to use,' but that's another argument for another time.)
2. The internet revolution started with DARPA, that guy from CERN whose name I can't remember, Mosaic, Netscape, and god help us, AOL. Microsoft never had anything to do with it. MS didn't provide any more home/office portability than you already had -- in fact, they provided less, because they refused to tell anyone how they saved their data.
3. I'll give you that. On the other hand, they made sure that every improvement worked to restrict interoperability and improve the power and reach of their monopoly -- e.g. IE.
1. It's illegal to use monopoly power to force other companies out of business. That means, for instance, if you've got a monopoly on gasoline, you can't use that to develop a monopoly on windshield-wiper fluid. Charging nothing for the browser was equivalent to that gas monopoly distrubiting a gallon of windshield-wiper fluid for nothing at every fill-up -- clearly illegal. That's why MS has to rely on the weak argument that they made the browser and OS component.
2. The government can force a monopoly to charge a 'fair' price for its product, but it tends to better serve the consumer by encouraging competition in that industry. I have no objection to whatever price MS puts on its software -- but when I can't buy a computer without that software, I do object. Furthermore, at almost every step up to Windows 95, MS had a competing prduct to price against, which they didn't and eventually squelched anyway. (DR-DOS, OS/2, etc)
The problem is less that MS bunded IE with Windows than that they used their monopoly power in desktop operating systems to their unfair advantage competing in a different market -- the market for browsers. 'Linux' is not a single company, and can't become a monopoly thereby; strictly speaking, 'Linux' doesn't require so much as glibc to be installed, much less a browser. For obvious reasons, achieving a desktop monopoly with GNU/Linux-based software will be very, very difficult, and not a problem we need to worry about...
Why, pray tell, would putting MS out of business destroy the economy? It would hammer MS's stock, sure, but MS itself doesn't employ enough people in a loose enough job market that the government has to look at it like Chrysler.
That, and it might cause greenspan to lower rates for once:)
There are 4096-machine clusters, but the marginal performance gain per machine drops for many kinds of computation because of the 'segmented' archictecture. That is, the limited bandwidth inter-cluster relative to the bandwidth intra-cluster makes programming the whole cluster a problem of finding a 64-way very loosely-bound approach, each segment of which is 64-way loosely-bound (each segment of which is N-way tightly bound because of the possibility of SMP). Finding algorithms to split up problems this way is very difficult, and in some cases, impossible. For a given problem, there is a network width/speed for which the limiting factor is the processor speed (e.g. you're not losing performance to overhead); this is the case for many more problems at switched N bps than at shared N bps (in/out-side a cluster).
On the other hand, there are algorithmic techniques for masking (network) latency (e.g. time-skewing), so it's possible to make better use of 'loosely-coupled' (relative to the algorithm's interconnect requirements) compute elements (machines/clusters/etc) than you may think.
There is a large space for conventional books, and will be into the foreseeable future; posters on this article have a done a fine job of explaining why. But there are two things intensely disappointing about the distribution of King's story, but first I would just like to mention that even static text transcriptions are useful -- because search & analysis tools may be applied to them, like any other digital document.
The first disappointment is the lack of any of hypertext. Hypertext is a very basic form of a interactivity, and it shouldn't be too hard of a job authorial to add it. I can understand why the story itself wasn't presented as a hyperlink web -- a position with which, as a writer, I sympathize. However, I can easily imagine that hyperlinks to commentary, or sections of the story which are well-written but not relevant enough to include, and so on, could be very interesting; a scheme along the line of those web-annotation devices could very interesting -- especially if those annotations were moderated! (It shouldn't be too hard to adjust the slashdot code to allow contextual responses, right?)
Interactive Fiction is a different beast that normal fiction, and much more difficult. Not only are all the technical problems of writing a story present, and the efforts of imagination necessary to create a scene, but that imagination must extend to interactions that wouldn't occur in a linear plot. Reading interactive fiction is not the same as playing a text-based adventure game -- you must play the role of the character. (well, conventionally. Sufficiently advanced AI techniques might allow you to play a secondary character, cause plot events to happen, etc.) Not everyone role-plays well, and some don't enjoy it; for these people, interactive fiction doesn't do it. But playing a role can itself be entertainment; and playing that role in scenes with talented direction (the game-master or a well-written AI) is even better -- especially if the prose is well-written. (Actually, interactive fiction is close in spirit to pre-built adventures, where the computer is a rather unimaginative game master...) For folks who don't enjoy role-play, writing a script to run the game through on the 'normal path' (the author's cut, so to speak) would generate the normal, book-form novel.
More closely concerning Jon Katz's contentions: interactivity on the 'net is primarily driven by the desire for interaction with others. For effectively every other form of interaction, the internet is a 'mere enabler', a way to solve some technical problem. (How do I interactively control my stocks if I can't be in NY all the time?) The novel as a story-telling form is already strongly divorced from person-to-person to communication -- though there is still some demand for professional (live) storytellers, for much the same reasons that movies have yet to displace live acting, and CDs live bands. But the lack of interactivity in a novel is a concious 'design' choice now, not a mere accident of history. (e.g. you used WordPerfect and Ghostview instead of Inform and Frotz.) The demand for interactivity around the novel, then, can only be satisfied by interaction with other readers, one form of which I suggested above.
This has been an important first step toward 'net-enabling' the novel, but to suggest the net/requires/ interactive fiction is to suggest that the net is somehow something more than a communication medium, that it is required to mutate what is transmitted across it. (Being able to talk back over a phone does not change the nature of broadcast radio; if you had all the radio station's CD's, all you need is some way for the choices of the DJ (ha! try the top-40 list and advertisers -- but that's another story) to be communicated to your CD player, you'd be set. Similar thinking is behind the astounding popularity of play-list based mp3 players and very large collections of songs...) True, static broadcast is not utilising the abilities of the 'net as a communication medium, but it is premature to suggest that a medium of communication/ought/ to affect the form of that communication. (For technical reasons, it usually does -- but that doesn't it make it good or proper...) That a new medium for communication allows new forms of communications does not mean the best way to generate those forms is by modifying the old ones.
What if the problem isn't that the coders are the majority of users, but that the majority of users aren't coders? -- No, I'm serious, so bear with me here:)
The most frequent kind of question I get sounds stupid, but is actually smart and naive. For instance -- what's the difference between hard drive megabytes and RAM megabytes? -- the reason this is a `stupid' question is because of an artifact of the hardware used to design the system. If/When we get high-speed holographic memory, your RAM may very well be nonvoilatile. Much the same thing applies to UI's; frequently the answer to questions (especially my mom's) are "Well, you should be able to you, but you can't," because I don't feel like explaining some piece of MS idiocy or principle of computer architecture.
But back to my original question -- what if the problem is the majority of users aren't coders? Intuitively, piping is pretty obvious. What makes it difficult to use the arcanity of the commands, options, and symbols used. On the other hand, why can't I 'pick up' a directory listing and drop it into the in-pipe of a filter, which will then ask me what I want to get rid of? So when it's done, I should have new directory listing, so I open up my toolbox and drag the new directory listing over to my mailer, to which I give a file listing the addresses. Conversely, I might only want to send the last file listed off, so I drag the mailer onto that file's listing in the directory. There's nothing particularly impossible about coding the mechanisms to allow this -- and ongoing work in GNOME and KDE should make that work relatively straightforward.
That's a relatively simple example, so here's a more complicated one that more closely resembles conventional programming. Why shouldn't your GUI -- which is intended to increase your productivity -- start off with what you're going to be working on, that is, a document? (For the second time through this example, the last document you worked on.) So you've got a nice blank 8.5x11 piece of paper on-screen, with (probably) rulers and a blinking cursor in the upper left-hand corner. The directory tree is like 'disk navigator' as the K menu -- tear-off menus and the like, so you can drill-down to a specific tool you need, like spell-check, or a set of scripts for doing tables in XML, or the open/close attribute for underline and italics, etc. You create the UI to the particular document you're working on as you go along. (Includes things like collapse-to-button menu/toolbars, window-shaded toolboxes/menus, etc.) That interface is brought back when you go back to that document; you can say whether or not to maintain that interface across new documents. Weblinks are followed the same way, where you'd normally let the 'browser' UI set take over until you return to your original document. Full-screen games and the like can be handled with virtual desktops.
But that strongly resembles the applications advertised in the back of magazines, where you can 'write your own programs without ONE LINE OF CODE!' (etc). The key here is that you're not using a single application to do whatever it thinks is worthwhile to provide. The *power* of the command line stems from its uniform treatment of applications -- a black box with one pipe in and two pipes out. There's no reason that this exact treatment can't be handled graphically*. Now, it's true that you may well want to share information between applications in a more structured way, but that too, can be handled**.
I think most users want to be coders, to make their application do what they want, how they want, but don't have time to climb the learning curve into (conventional) programming, or learning every option and name from the command line.
-_Quinn
* The program's name with three pipes coming out of it comes to mind; make sure the user can drag the pipes around as necessary and that outgoing and incoming pipes look clearly different. For program switches, have a listbox with selection; for arguments, combine a listbox (list of options requiring arguments) with a text-entry field (the argument.) All of this, plus context help (e.g. this switch does what?) can (usually) be gleaned automagically from the man pages. It might be worthwhile to crank out some XML to name the switches (e.g. -R(ecursive) is what would show up in the list for cp) so they aren't quite so cryptic initially.
** Add context menus which reflect on the application to determine what kinds of input you can hook up to it; splitter pipes that don't work unless all their inputs are OK are a good way to handle function parameters. See the reality toolkit for a demonstration of what I'm talking about. The KDE/GNOME kparts/bonobo models, so far as I can tell, are aiming at the reality toolkit in C++.
According to the judge, Microsoft's primary area of business is not its primae facie operating system, but the APIs which Microsoft has so far maintained exclusive control over. Linux and the BSDs, while they challenge Microsoft's power in the server arena (and are joining *NIX competitors from the Big Iron makers) are not in competition with Microsoft's APIs. Microsoft could, and IMHO should, cut loose from the Win95/NT kernels and port their APIs to a a system that works. (A BSD, for lisensing reasons, I'd imagine.) Spend a year without adding any features to those APIs (i.e. fix them so they at least approximate the reliability of the host system); release the new host & API package as Windows X.
This package would continue to sell as well as Windows 2000, because it's the only way to get Microsoft's APIs, whose invested presence in other software houses forms the single largest barrier to entry! Why hasn't MS Office been ported to Linux -- not because it can't be, and not because it wouldn't sell, but because it then doesn't additionally sell a copy of Windows! The cost of porting code to another OS forms a barrier to entry -- how many companies are writing games for Linux? Two? (Id and Loki.)
If you change 'rail road cars' to 'Windows APIs', 'oil' 'to operating systems', and 'Standard Oil' to 'Microsoft' in the sentence "Standard Oil regulating who got railroad tank cars, for instance, so small oil producers couldn't compete," I think you end up with a good understanding of the judge's position on the case...
Was anyone else amused that Mindcraft found that samba running in user space was a bad thing? (Look at the phase three report.) Hey... I wonder if this means someone should hack up a ksmbd?:)
I am curious, however, to see what the performance/reliability difference is when we're looking at a four-way linux cluster vs. the four-way SMP NT...
Sun makes good software because Sun doesn't make money on it. The useful (as opposed to necessary, like the OS) bits of Sun boxes' software aren't made by Sun anyway (bind, sendmail, perl, apache, etc...). While I'm sure Sun's engineers have deadline pressures to finish the next version of the O/S, it doesn't seem to be as feature-oriented as Windows new versions are, perhaps just because the new features we're seeing all are (or should be) done in user-space. Sun is competing on the merits of its hardware, not its software, which only has to be good enough to (a) display the power of their hardware, and (b) not look bad in comparison to a UNIX clone compiled for their hardware.
Similarly, SGI is moving to Linux because most of IRIX is uninteresting from the business model SGI has, which is Selling Really Good Graphics Workstations. They're doing some things, and have done some things (XFS, OpenGL) to make their use of Linux a symbiosis instead of parasitism. R. Kent Koeniger spoke at the Open Source/Open Science conference at Brookhave National Labs, explaining (paraphrase) that SGI wanted to make Linux suitable for use with Really Big Things, like multi-terrabyte filesystems, N-CPU multiprocessing, etc... which sounds like the system specs, minus the SGI-supplied OpenGL card and drivers, a massive graphics workstation...
Many people print out paper for a simple two-fold reason: they grew up making corrections that way. The kinesthetics (and ergonomics!) of pen-on-paper are 'better' than those of a strictly digital correction, and most monitors/still/ aren't large enough to fit a full page of text on the screen at once (at a readable size), which means it's quite frequently/faster/ to read something hardcopy, especially if you're going to be doing a whole lot of back-and-forth page flipping. (Not to mention you've had much more practice reading hardcopy than screencopy.) The physicality of hardcopy makes it easier to remember where a particular sentence is and go back to it.
The other reason, suprisingly enough, is that paper works better. (No, really!) Tools like PDAs don't, and shouldn't, try to implement the entire functionality of paper, just the small subset of it concerned with scheduling (and contact information, usually). It's not substantially easier to make massive modifications on a computer than on paper (though it's true that they tend to/look/ a lot better); paper doesn't have compatibility problems, or need power; on a character-per-gram basis for anything under a million or so words, paper is lighter than a computer. And of course, you can do free-hand sketching on paper much easier than you can the computer, unless you're sickly good with the mouse.
In terms of saving trees, look at the continual predictions of 'disposable computers' in the media. As usual, they've taken a perfectly legitimate trend (Moore's Law) and distorted it out of recognition, but it can make a certain kind of sense*. 'Disposable' processors aren't terribly useful without 'disposable' memory and displays; but what if you went ahead and ran off a hundred 8.5x11 plastic (touch sensitive) LCDs on which are burned (with the same process) a 386-class processor and (using tech from, IIRC, 3M) a plastic battery strip? (Assuming that some futuretech will handle non-volatile memory for us...) They're not/quite/ disposable, but they're close, and to displace paper, you just need instant copying. Stick a hundred-pack (same size as a 500-pack of 20-bond) into your digital copier, and el-blammo, you've got 100 perfect copies. If you need to put your sketch up on the overhead, put your paper down on it and hit the 'copy' button; instant transparency. Transfering information from page to page would work the same way. If communication is surface to surface (which make intuitive sense to me), then you can dispense with expensive wireless circuitry; and the nv-ram doesn't have to be very large -- find out what the average length of the average business memo is (I'd guess under ten pages) and make that the storage size. Flipping the page over changes which page is displayed.
Yeah, it's pretty much a standard vision of the future, but what's scary is how close it is...
* I'm already in medium-term disposable computing. I overclock my celeron, which burns out about four months later, at which point I buy a much faster one; the price per time-performance ratio is unbeatable.
The reason to ban internet taxes in the US is quite simple: they won't collect enough to make it worthwhile. People can move their servers to other countries and/or use (anonymous, encrptyed) digital cash solutions. See _Cryptonomicom_ for more information.
I love the computers and cars comparison, because this is where it comes in.:)
I don't CARE what the difference between glibc 1 and 2 is. I don't care what "CVS" is. I don't have to know the difference between kernel 2.2.6 and kernel 2.2.10 to run my desktop. (Faster and stabler than Windows, but that's another post). When I use my desktop, I never want to have to do more than turn the computer on and change my screen-saver. When I use my desktop, I want to doubleclick on something and start writing my reports. Yes, I *do* have to know about the mouse, both buttons, and my username and password, but I don't have to know anything technical about the O/S to use my desktop properly.
Am I missing something here, or is this already the case? RedHat 6.0 will boot from my CD and do everything except note that my mouse is an intellimouse on its own*; if they included AbiWord (for example) in their install, I'd never have to do anything else.
Nonetheless, I happen to know the differences between shocks and struts, and the grades of gasoline, and what the difference between 10W30 and 10W40 motor oil are because they make me a more effective user of my automobile. I also happen to be an inveterate tinkerer, so I also know how to rebuild an engine from scratch, if you'll send me the plans. Heck, I'm building my own car right now, from parts and plans other people supply; I'll be selling the plans into the public domain once I finish.
As it stands, most people are ignoring the vast majority of things they can do with 'computers' (in most cases, with their favored productivity app(s)) because they are either unwilling or unable to learn about the rest. I am an elitist; I don't necessarily believe that the Linux Command-Line Way is the Right Way, but there is a learning curve, especially for Windows**. (I went from an Apple IIGS to a Win3.1 PC box, and I was lost as all get-out until I sat down and read the manual cover-to-cover. The Apple IIGS was a triumph -- an totally general information appliance. Only if you stuck the GS/OS or Finder disks in specifically did you even realize there/was/ an operating system.)
However, alot of the general resisistance to dumbing-down comes from (geek) culture rather than technical issues***. It goes against what's valued (especially in the gift culture that's developing these applications and operating systems about which you're pontificating) to dumb things down, though not to make things better. In many cases, the two are diametrically opposed. (Also, the less-choice-is-simpler camp attacks the liberterian in every geek.) Our gift culture values displays of competence, pretty much. Social standing is generally taken in terms of competence (up to wizardry in areas); it's no suprise that Joe WinBloze user is universally reviled. No one objects to mpeg123 front-ends, but using a front-end in this case actually gains you a significant amount of functionality; and the command-line is still there if you ever (cronjob alarm clock, anyone?) need to script it. So there's a certain prestige attatched to the mpeg123 front-end builder.
I personally used fvwm95 until I got my hands on KDE, so I don't find it particularly disturbing. It also means that I personally know that it's possible to move from the wrapper to the real thing (so to speak -- not to be dissing fvwm95). On the other hand, while WinLinux is technically impressive (slurping h/w settings out of the windows registry and translating that into Linux drivers & setup is decidedly non-trivial), not many/.'ers liked the idea. Part of that reason is that a UMSDOS-based linux distro is not the Right Thing, especially if you're looking to allow the user to gradually move up to greater competence.
Elegant (in the classic sense) designs are always easy to use, and that's what we should be aiming towards. The opposition stems from the dumbing-down process not generating an elegant design. If you/really/ want someone to someone to use your software right away, you can walk them through it with intelligent defaults. The key is that the walk-through is NOT an elegant design, and that it must not be the only way to use the software.
Intuition is wrong enough frequently that it's not always a good design...
-_Quinn
* Admittedly, the defaults aren't particularly brilliant, but that's life.
** The major advance of Windows 3.1 was the GUI for applications, not for the system. This is where Win9x went horribly wrong. It's a given that the computer won't always be right, so treating cases where it isn't as exceptions is a Bad Idea and tends to break things. (That is to say, I/like/ autoprobing and plug and play, but don't expect them to work.) I know it's admitting to being horribly inadequate, but I don't know TeX, and find that WordPerfect 8 and/or AbiWord satisfy my writing urges completely (except for WP8's broken postscript generator). I know I'm not a Real Coder because I like using jcc over emacs or pico. (Though I do keep an xterm handy for doing the makes, because my Makefiles would give makemake the shaking fits.)
*** IE, we're not building a homogeneous system here; interoperability is going to be a bitch. IE, mounting is the Right Thing to do; is automounting? (For the "Why should I have to 'mount' CD-ROMs when they 'just work' in Windows?" questions.) It is the Right Thing to distribute applications as source; should we adjust the LSB so that there's a user-writable subset of directories for scripted installs? Should we ship our distro with '.rpm' as a MIME-type for netscape so we can 'just download' a binary-form application?
*Sigh.* Another article that fails to grasp the vitally important distinction between Artificial Intelligence and Machine Intelligence. Taking the burger-bot example, I order the "burger with everything, no pickles."
The robot has to (a) recognize what I said (voice recognition; a solved problem, given the enough hardware); (b) recognize what I actually want, the ketchup, onions, mustard, burger, buns, cheese, etc (expert system); and (c) put the burger together (vision system feeding an expert system that generates motion commands).
Everything but (c) has been done, and done well, albeit, incredibly expensively. This bot will look artificially look intelligent because you'll only ever see it in the Burger King.
Conventionally, this would be an artificial intelligence; so far, so good. However, AI conventionally also includes what are better called machine intelligences, machines on thinking and acting on a level at or beyond that of a 'normal human.' In other words, an MI (a machine capable of holding a conversation with you in a manner indistinguishable from a(nother) human) working at a burger joint would get just as bored and surly as the teenagers who work there now would -- which would defeat the point, no?
So why is this distinction so important? AI's might be able to learn, but they will be cheaper than MI's, definitely no citizen material, and mass-producable; each burger-bot is identical to the next, and comes with a certain store of 'knowledge' and 'skills' (databases and expert systems) that can be copied from machine to machine. (We're doing that now, so we know it works.)
What makes everyone think the same will hold for MIs? Why would an MI understand the hardware, firmware, and software it's constructed out of any more so than I understand my bioware? Suppose we overclock an MI's hardware by a factor of two... the MI is still only thinking twice as fast as normal human, not thinking anything more significant. Supposing that we manage to teach an MI how to design circuitry, what makes you think that it'll be any less prone to error than a human? It's still thinking its way through the design process, same as Joe IEEE Member does; granting a human perfect memory doesn't make him/her immune to error. An MI would be able to run, say, a chip validation program without thinking its way through it, but it would be using it the same way you and I would. An MI would not be anymore aware of what its processors are doing than you and I know what our neurons are doing; you think and your hands type, and there's a total concious disconnect between the two actions. Now, granting that conciousness in a finite-state machine (which I, for one, am not), there's still no gaurantee that you'll be able to save the state in such a fashion that it can be duplicated at your leisure on other hardware later.
I think the best you can hope for you with MIs is a very powerful subconcious, with the attendant powerful intuition. Data bases (Cyc, for example) would have either be searched conciously (like looking up every word I'm trying to type!) and hence uselessly, or subconciously, like I'm 'accessing my database' on English to crank out these paragraphs. A given MI might read a textbook on relativity and understand it, but there's no gaurantee, again, that it will be able to directly implant that knowledge in another MI; it would have become part of its particular neural net (non-transferrable, in general) or re-ordered its database (possibly transferrable, but the task of synchronizing the databases would be enourmous.) So an MI might be able to say, I'm guessing the problem is with the transmission, but unlike an expert system, it wouldn't be able to provide you with an exact trace of the logic behind that conclusion, unless it took the time to think it through 'conciously'.
Now, because of advances in h/w, an MI might be able to come up with that intuition in a tenth of the time a human might be able to, and 'learn' ten-times faster, but that only cuts the time it take to teach it 'from scratch' down to two or three years... and each additional job four or five months.
I just don't think an intelligent system can have full knowledge of itself; just on the storage level, a full image of itself would always take twice as much storage as it was currently using; the knowledge of its first-order self would not be exhaustive of its second-order self, and so on...
Slashdot Mirror
First, is anyone from GNOME talking to anyone from GNUstep? While approve of the Bonobo architecture ideas, it looks to me like the two are doing very similar things. (And WindowMaker rocks my world.) Second -- does anyone else think that an object-based OS doesn't really sound like UNIX? -_Quinn
This first is the obvious problem -- understanding natural language seems to be an AI-complete problem. ('AI-complete' is a term borrowed from the normal classification of the difficulty of problems, and means that a problem is equivalent to and/or requires AI.) The other problem is more subtle -- define 'wedding photos' for me. People can and /want/ to describe the contents of something rather than where it resides. (I've seen people who prefer to search for something rather than type in a URL, even if they know it.) The problem of meta-data is very, very, big. How many people do you know who even enter so much meta-data as the title of a document when they save it? (I don't mean the name -- most WP's have an option to enter author/title/subject/keywords to help them find your document.) Even aside from formulating a meta-data standard, how do you /get/ that meta-data? Not all of it, and almost all of the useful bits, can't be -- except, perhaps, with AI.
-_Quinn
While you're right to say,
"It is not about stopping progreess, it is about trying to encourage it."
your next sentence simple doesn't follow. The referenced article -- and most of the discussion recently -- about IP attacks the premise that IP still serves to encourage progress. Ignoring that for two sentences, effectively every inventor is somebody else's employee. The IP is not going to an individual/group -- it's going to the company (government institution, etc). If we want to encourage innovation by rewarding it, IP law ought to be changed so the individual/group gets more money. (And, incidentally, yes, I would invent. I'm a programmer (among other things) -- I do it all the time.)
The essence of patents is an attempt by the government to bribe those with secrets into exposing them by helping to exploit them commercially. This is distinct from the essence of copyright, by which it was decided that the creation of more works was beneficial to the people, and that the best way to promote more works was economic -- to allow the free market to handle it. With that in mind, the copyright system was set up to create such a market. The essence of the argument against copright is that the market (as consituted by, say, the MPAA) created by copyright law no longer -- or less than optimally -- benefits the people, and the same for patents, though the specifics, obviously, vary.
-_Quinn
Hm. I've noticed stability issues with Netscape on RedHat 6.0, but not on any other platform. It may just be that the RH6.0 boxes don't have an up-to-date version of NS. I won't waste your time responding to some of your points, as they don't apply to me (I run NS under Slackware 7 w/o any stability issues at all, every window always maximized).
:) Thanks for the reply -- I wanted to know what I was missing over here in Linux land until I get a build of Mozilla that runs fast enough not to hurt.
Point one and two are both good, though I've honestly never noticed image problems with NS in the course of my day-to-day browsing -- probably because of your fifth point. I'll take your word about JavaScript -- perhaps because NS's lack of functionality (odd, as they invented the language), I've not been very impressed by it. As for NS not handling straight HTML -- well, I don't do web design professionally, so I guess it doesn't matter that I've never seen it have any problems.
-_Quinn
Actually, I had noticed this (the page reload), but not thought about it, because it doesn't affect me any -- with multiple desktops, the only apps I don't run full-screen are terminal emulators. Honestly, though, it's only in /very/ rare occasions (e.g. tables with ~3000 cells) that I've had to wait for the rendering any noticeable amount of time before the page finishes loading. Thanks.
-_Quinn
Maybe this is a silly question, but what does IE 5.x do so much better than NS 4.7x? Why should I switch from NS 4.7 on Linux on IE 5.5 on Win9x? It's not more stable. It's certainly less secure. Am I just not getting around to enough sites on the web to notice Netscape's (supposedly) archaic renderer? For all the claims of being a faster browser, I have to admit to not caring at all -- if the page is complex enough that rendering time becomes significant, download time almost always overshadows it.
These are serious questions. Why should I change? What usability improvements does IE have? Why does NS 'suck so hard'?
-_Quinn
Oddly enough, one of the 'big ideas' for improving air quality that was floated a while back and is (probably) wending its way toward production amount to installing a catalytic converter on the radiator/air intake, which has two benefits -- first, you clean up what's coming out the pipe in front of you, and second, burning cleaner air, your exhaust is cleaner. Some advocated using this technique to also remove non-car-produced pollutants from the air.
-_Quinn
Did anyone else initially think that "GPF comics" meant Gender Protection Fault?
-_Quinn
To become an engineer, you study the best way that it's been done before, and software engineering is no different. Since the best way it's been done before is UNIX, the ready availability of the blueprints for Linux makes it a natural choice for study. Mucking about with Linux is equivalent to moving I-beams around in the Twin Towers to see what happens -- except, naturally, quite a bit cheaper. The theory of software engineering -- computer science -- is more readily studied if the tools you use for the study are more readily available. The downtime associated with Windows means that it's not readily available...
-_Quinn
Natalie Gritman covered in hot ports.
All four of these must be fought in court, with the expectation of losing on #1. It'll be OK to lose #1, so long as Slashdot is forced by court injuction to perform the moderation. (That is, does not lose its common-carrier immunity.) The other three, likewise, must be fought, but expecting to take them to the Supreme Court for resolution.
Aside from moral reasons to fight -- I'm not familiar enough with Andover management to know if they're sufficient -- consider how many hits controversy generates. (Ever looked at the way ZDnet does business?) Furthermore, consider how many negative hits caving would generate. Thanks.
-_Quinn
Try WindowMaker for GNUstep if you liked NeXT :)
1: You bought MS's line, hook & sinker. The Macintosh (clone of a PARC system) made computers 'easy to use' -- the Mac just made them cheap enough to be feasible. Microsoft's Windows's jumped on the extremely cheap x86 hardware being developed to run VisiCalc and/or 1-2-3 faster and made a cheap knock-off of the Mac -- made 'easy to use' cheap enough to be popular. (I also take issue with 'easy to use,' but that's another argument for another time.)
2. The internet revolution started with DARPA, that guy from CERN whose name I can't remember, Mosaic, Netscape, and god help us, AOL. Microsoft never had anything to do with it. MS didn't provide any more home/office portability than you already had -- in fact, they provided less, because they refused to tell anyone how they saved their data.
3. I'll give you that. On the other hand, they made sure that every improvement worked to restrict interoperability and improve the power and reach of their monopoly -- e.g. IE.
1. It's illegal to use monopoly power to force other companies out of business. That means, for instance, if you've got a monopoly on gasoline, you can't use that to develop a monopoly on windshield-wiper fluid. Charging nothing for the browser was equivalent to that gas monopoly distrubiting a gallon of windshield-wiper fluid for nothing at every fill-up -- clearly illegal. That's why MS has to rely on the weak argument that they made the browser and OS component.
2. The government can force a monopoly to charge a 'fair' price for its product, but it tends to better serve the consumer by encouraging competition in that industry. I have no objection to whatever price MS puts on its software -- but when I can't buy a computer without that software, I do object. Furthermore, at almost every step up to Windows 95, MS had a competing prduct to price against, which they didn't and eventually squelched anyway. (DR-DOS, OS/2, etc)
-_Quinn
The problem is less that MS bunded IE with Windows than that they used their monopoly power in desktop operating systems to their unfair advantage competing in a different market -- the market for browsers. 'Linux' is not a single company, and can't become a monopoly thereby; strictly speaking, 'Linux' doesn't require so much as glibc to be installed, much less a browser. For obvious reasons, achieving a desktop monopoly with GNU/Linux-based software will be very, very difficult, and not a problem we need to worry about...
-_Quinn
Why, pray tell, would putting MS out of business destroy the economy? It would hammer MS's stock, sure, but MS itself doesn't employ enough people in a loose enough job market that the government has to look at it like Chrysler.
:)
That, and it might cause greenspan to lower rates for once
-_Quinn
There are 4096-machine clusters, but the marginal performance gain per machine drops for many kinds of computation because of the 'segmented' archictecture. That is, the limited bandwidth inter-cluster relative to the bandwidth intra-cluster makes programming the whole cluster a problem of finding a 64-way very loosely-bound approach, each segment of which is 64-way loosely-bound (each segment of which is N-way tightly bound because of the possibility of SMP). Finding algorithms to split up problems this way is very difficult, and in some cases, impossible. For a given problem, there is a network width/speed for which the limiting factor is the processor speed (e.g. you're not losing performance to overhead); this is the case for many more problems at switched N bps than at shared N bps (in/out-side a cluster).
On the other hand, there are algorithmic techniques for masking (network) latency (e.g. time-skewing), so it's possible to make better use of 'loosely-coupled' (relative to the algorithm's interconnect requirements) compute elements (machines/clusters/etc) than you may think.
-_Quinn
There is a large space for conventional books, and will be into the foreseeable future; posters on this article have a done a fine job of explaining why. But there are two things intensely disappointing about the distribution of King's story, but first I would just like to mention that even static text transcriptions are useful -- because search & analysis tools may be applied to them, like any other digital document.
/requires/ interactive fiction is to suggest that the net is somehow something more than a communication medium, that it is required to mutate what is transmitted across it. (Being able to talk back over a phone does not change the nature of broadcast radio; if you had all the radio station's CD's, all you need is some way for the choices of the DJ (ha! try the top-40 list and advertisers -- but that's another story) to be communicated to your CD player, you'd be set. Similar thinking is behind the astounding popularity of play-list based mp3 players and very large collections of songs...) True, static broadcast is not utilising the abilities of the 'net as a communication medium, but it is premature to suggest that a medium of communication /ought/ to affect the form of that communication. (For technical reasons, it usually does -- but that doesn't it make it good or proper...) That a new medium for communication allows new forms of communications does not mean the best way to generate those forms is by modifying the old ones.
The first disappointment is the lack of any of hypertext. Hypertext is a very basic form of a interactivity, and it shouldn't be too hard of a job authorial to add it. I can understand why the story itself wasn't presented as a hyperlink web -- a position with which, as a writer, I sympathize. However, I can easily imagine that hyperlinks to commentary, or sections of the story which are well-written but not relevant enough to include, and so on, could be very interesting; a scheme along the line of those web-annotation devices could very interesting -- especially if those annotations were moderated! (It shouldn't be too hard to adjust the slashdot code to allow contextual responses, right?)
Interactive Fiction is a different beast that normal fiction, and much more difficult. Not only are all the technical problems of writing a story present, and the efforts of imagination necessary to create a scene, but that imagination must extend to interactions that wouldn't occur in a linear plot. Reading interactive fiction is not the same as playing a text-based adventure game -- you must play the role of the character. (well, conventionally. Sufficiently advanced AI techniques might allow you to play a secondary character, cause plot events to happen, etc.) Not everyone role-plays well, and some don't enjoy it; for these people, interactive fiction doesn't do it. But playing a role can itself be entertainment; and playing that role in scenes with talented direction (the game-master or a well-written AI) is even better -- especially if the prose is well-written. (Actually, interactive fiction is close in spirit to pre-built adventures, where the computer is a rather unimaginative game master...) For folks who don't enjoy role-play, writing a script to run the game through on the 'normal path' (the author's cut, so to speak) would generate the normal, book-form novel.
More closely concerning Jon Katz's contentions: interactivity on the 'net is primarily driven by the desire for interaction with others. For effectively every other form of interaction, the internet is a 'mere enabler', a way to solve some technical problem. (How do I interactively control my stocks if I can't be in NY all the time?) The novel as a story-telling form is already strongly divorced from person-to-person to communication -- though there is still some demand for professional (live) storytellers, for much the same reasons that movies have yet to displace live acting, and CDs live bands. But the lack of interactivity in a novel is a concious 'design' choice now, not a mere accident of history. (e.g. you used WordPerfect and Ghostview instead of Inform and Frotz.) The demand for interactivity around the novel, then, can only be satisfied by interaction with other readers, one form of which I suggested above.
This has been an important first step toward 'net-enabling' the novel, but to suggest the net
-_Quinn
What if the problem isn't that the coders are the majority of users, but that the majority of users aren't coders? -- No, I'm serious, so bear with me here :)
The most frequent kind of question I get sounds stupid, but is actually smart and naive. For instance -- what's the difference between hard drive megabytes and RAM megabytes? -- the reason this is a `stupid' question is because of an artifact of the hardware used to design the system. If/When we get high-speed holographic memory, your RAM may very well be nonvoilatile. Much the same thing applies to UI's; frequently the answer to questions (especially my mom's) are "Well, you should be able to you, but you can't," because I don't feel like explaining some piece of MS idiocy or principle of computer architecture.
But back to my original question -- what if the problem is the majority of users aren't coders? Intuitively, piping is pretty obvious. What makes it difficult to use the arcanity of the commands, options, and symbols used. On the other hand, why can't I 'pick up' a directory listing and drop it into the in-pipe of a filter, which will then ask me what I want to get rid of? So when it's done, I should have new directory listing, so I open up my toolbox and drag the new directory listing over to my mailer, to which I give a file listing the addresses. Conversely, I might only want to send the last file listed off, so I drag the mailer onto that file's listing in the directory. There's nothing particularly impossible about coding the mechanisms to allow this -- and ongoing work in GNOME and KDE should make that work relatively straightforward.
That's a relatively simple example, so here's a more complicated one that more closely resembles conventional programming. Why shouldn't your GUI -- which is intended to increase your productivity -- start off with what you're going to be working on, that is, a document? (For the second time through this example, the last document you worked on.) So you've got a nice blank 8.5x11 piece of paper on-screen, with (probably) rulers and a blinking cursor in the upper left-hand corner. The directory tree is like 'disk navigator' as the K menu -- tear-off menus and the like, so you can drill-down to a specific tool you need, like spell-check, or a set of scripts for doing tables in XML, or the open/close attribute for underline and italics, etc. You create the UI to the particular document you're working on as you go along. (Includes things like collapse-to-button menu/toolbars, window-shaded toolboxes/menus, etc.) That interface is brought back when you go back to that document; you can say whether or not to maintain that interface across new documents. Weblinks are followed the same way, where you'd normally let the 'browser' UI set take over until you return to your original document. Full-screen games and the like can be handled with virtual desktops.
But that strongly resembles the applications advertised in the back of magazines, where you can 'write your own programs without ONE LINE OF CODE!' (etc). The key here is that you're not using a single application to do whatever it thinks is worthwhile to provide. The *power* of the command line stems from its uniform treatment of applications -- a black box with one pipe in and two pipes out. There's no reason that this exact treatment can't be handled graphically*. Now, it's true that you may well want to share information between applications in a more structured way, but that too, can be handled**.
I think most users want to be coders, to make their application do what they want, how they want, but don't have time to climb the learning curve into (conventional) programming, or learning every option and name from the command line.
-_Quinn
* The program's name with three pipes coming out of it comes to mind; make sure the user can drag the pipes around as necessary and that outgoing and incoming pipes look clearly different. For program switches, have a listbox with selection; for arguments, combine a listbox (list of options requiring arguments) with a text-entry field (the argument.) All of this, plus context help (e.g. this switch does what?) can (usually) be gleaned automagically from the man pages. It might be worthwhile to crank out some XML to name the switches (e.g. -R(ecursive) is what would show up in the list for cp) so they aren't quite so cryptic initially.
** Add context menus which reflect on the application to determine what kinds of input you can hook up to it; splitter pipes that don't work unless all their inputs are OK are a good way to handle function parameters. See the reality toolkit for a demonstration of what I'm talking about. The KDE/GNOME kparts/bonobo models, so far as I can tell, are aiming at the reality toolkit in C++.
According to the judge, Microsoft's primary area of business is not its primae facie operating system, but the APIs which Microsoft has so far maintained exclusive control over. Linux and the BSDs, while they challenge Microsoft's power in the server arena (and are joining *NIX competitors from the Big Iron makers) are not in competition with Microsoft's APIs. Microsoft could, and IMHO should, cut loose from the Win95/NT kernels and port their APIs to a a system that works. (A BSD, for lisensing reasons, I'd imagine.) Spend a year without adding any features to those APIs (i.e. fix them so they at least approximate the reliability of the host system); release the new host & API package as Windows X.
This package would continue to sell as well as Windows 2000, because it's the only way to get Microsoft's APIs, whose invested presence in other software houses forms the single largest barrier to entry! Why hasn't MS Office been ported to Linux -- not because it can't be, and not because it wouldn't sell, but because it then doesn't additionally sell a copy of Windows! The cost of porting code to another OS forms a barrier to entry -- how many companies are writing games for Linux? Two? (Id and Loki.)
If you change 'rail road cars' to 'Windows APIs', 'oil' 'to operating systems', and 'Standard Oil' to 'Microsoft' in the sentence "Standard Oil regulating who got railroad tank cars, for instance, so small oil producers couldn't compete," I think you end up with a good understanding of the judge's position on the case...
-_Quinn
Was anyone else amused that Mindcraft found that samba running in user space was a bad thing? (Look at the phase three report.) Hey... I wonder if this means someone should hack up a ksmbd? :)
I am curious, however, to see what the performance/reliability difference is when we're looking at a four-way linux cluster vs. the four-way SMP NT...
-_Quinn
Sun makes good software because Sun doesn't make money on it. The useful (as opposed to necessary, like the OS) bits of Sun boxes' software aren't made by Sun anyway (bind, sendmail, perl, apache, etc...). While I'm sure Sun's engineers have deadline pressures to finish the next version of the O/S, it doesn't seem to be as feature-oriented as Windows new versions are, perhaps just because the new features we're seeing all are (or should be) done in user-space. Sun is competing on the merits of its hardware, not its software, which only has to be good enough to (a) display the power of their hardware, and (b) not look bad in comparison to a UNIX clone compiled for their hardware.
Similarly, SGI is moving to Linux because most of IRIX is uninteresting from the business model SGI has, which is Selling Really Good Graphics Workstations. They're doing some things, and have done some things (XFS, OpenGL) to make their use of Linux a symbiosis instead of parasitism. R. Kent Koeniger spoke at the Open Source/Open Science conference at Brookhave National Labs, explaining (paraphrase) that SGI wanted to make Linux suitable for use with Really Big Things, like multi-terrabyte filesystems, N-CPU multiprocessing, etc... which sounds like the system specs, minus the SGI-supplied OpenGL card and drivers, a massive graphics workstation...
-_Quinn
-_Quinn
Many people print out paper for a simple two-fold reason: they grew up making corrections that way. The kinesthetics (and ergonomics!) of pen-on-paper are 'better' than those of a strictly digital correction, and most monitors /still/ aren't large enough to fit a full page of text on the screen at once (at a readable size), which means it's quite frequently /faster/ to read something hardcopy, especially if you're going to be doing a whole lot of back-and-forth page flipping. (Not to mention you've had much more practice reading hardcopy than screencopy.) The physicality of hardcopy makes it easier to remember where a particular sentence is and go back to it.
/look/ a lot better); paper doesn't have compatibility problems, or need power; on a character-per-gram basis for anything under a million or so words, paper is lighter than a computer. And of course, you can do free-hand sketching on paper much easier than you can the computer, unless you're sickly good with the mouse.
/quite/ disposable, but they're close, and to displace paper, you just need instant copying. Stick a hundred-pack (same size as a 500-pack of 20-bond) into your digital copier, and el-blammo, you've got 100 perfect copies. If you need to put your sketch up on the overhead, put your paper down on it and hit the 'copy' button; instant transparency. Transfering information from page to page would work the same way. If communication is surface to surface (which make intuitive sense to me), then you can dispense with expensive wireless circuitry; and the nv-ram doesn't have to be very large -- find out what the average length of the average business memo is (I'd guess under ten pages) and make that the storage size. Flipping the page over changes which page is displayed.
The other reason, suprisingly enough, is that paper works better. (No, really!) Tools like PDAs don't, and shouldn't, try to implement the entire functionality of paper, just the small subset of it concerned with scheduling (and contact information, usually). It's not substantially easier to make massive modifications on a computer than on paper (though it's true that they tend to
In terms of saving trees, look at the continual predictions of 'disposable computers' in the media. As usual, they've taken a perfectly legitimate trend (Moore's Law) and distorted it out of recognition, but it can make a certain kind of sense*. 'Disposable' processors aren't terribly useful without 'disposable' memory and displays; but what if you went ahead and ran off a hundred 8.5x11 plastic (touch sensitive) LCDs on which are burned (with the same process) a 386-class processor and (using tech from, IIRC, 3M) a plastic battery strip? (Assuming that some futuretech will handle non-volatile memory for us...) They're not
Yeah, it's pretty much a standard vision of the future, but what's scary is how close it is...
* I'm already in medium-term disposable computing. I overclock my celeron, which burns out about four months later, at which point I buy a much faster one; the price per time-performance ratio is unbeatable.
-_Quinn
The reason to ban internet taxes in the US is quite simple: they won't collect enough to make it worthwhile. People can move their servers to other countries and/or use (anonymous, encrptyed) digital cash solutions. See _Cryptonomicom_ for more information.
-_Quinn
I love the computers and cars comparison, because this is where it comes in. :)
/was/ an operating system.)
/.'ers liked the idea. Part of that reason is that a UMSDOS-based linux distro is not the Right Thing, especially if you're looking to allow the user to gradually move up to greater competence.
/really/ want someone to someone to use your software right away, you can walk them through it with intelligent defaults. The key is that the walk-through is NOT an elegant design, and that it must not be the only way to use the software.
/like/ autoprobing and plug and play, but don't expect them to work.)
I don't CARE what the difference between glibc 1 and 2 is. I don't care what "CVS" is. I don't have to know the difference between kernel 2.2.6 and kernel 2.2.10 to run my desktop. (Faster and stabler than Windows, but that's another post). When I use my desktop, I never want to have to do more than turn the computer on and change my screen-saver. When I use my desktop, I want to doubleclick on something and start writing my reports. Yes, I *do* have to know about the mouse, both buttons, and my username and password, but I don't have to know anything technical about the O/S to use my desktop properly.
Am I missing something here, or is this already the case? RedHat 6.0 will boot from my CD and do everything except note that my mouse is an intellimouse on its own*; if they included AbiWord (for example) in their install, I'd never have to do anything else.
Nonetheless, I happen to know the differences between shocks and struts, and the grades of gasoline, and what the difference between 10W30 and 10W40 motor oil are because they make me a more effective user of my automobile. I also happen to be an inveterate tinkerer, so I also know how to rebuild an engine from scratch, if you'll send me the plans. Heck, I'm building my own car right now, from parts and plans other people supply; I'll be selling the plans into the public domain once I finish.
As it stands, most people are ignoring the vast majority of things they can do with 'computers' (in most cases, with their favored productivity app(s)) because they are either unwilling or unable to learn about the rest. I am an elitist; I don't necessarily believe that the Linux Command-Line Way is the Right Way, but there is a learning curve, especially for Windows**. (I went from an Apple IIGS to a Win3.1 PC box, and I was lost as all get-out until I sat down and read the manual cover-to-cover. The Apple IIGS was a triumph -- an totally general information appliance. Only if you stuck the GS/OS or Finder disks in specifically did you even realize there
However, alot of the general resisistance to dumbing-down comes from (geek) culture rather than technical issues***. It goes against what's valued (especially in the gift culture that's developing these applications and operating systems about which you're pontificating) to dumb things down, though not to make things better. In many cases, the two are diametrically opposed. (Also, the less-choice-is-simpler camp attacks the liberterian in every geek.) Our gift culture values displays of competence, pretty much. Social standing is generally taken in terms of competence (up to wizardry in areas); it's no suprise that Joe WinBloze user is universally reviled. No one objects to mpeg123 front-ends, but using a front-end in this case actually gains you a significant amount of functionality; and the command-line is still there if you ever (cronjob alarm clock, anyone?) need to script it. So there's a certain prestige attatched to the mpeg123 front-end builder.
I personally used fvwm95 until I got my hands on KDE, so I don't find it particularly disturbing. It also means that I personally know that it's possible to move from the wrapper to the real thing (so to speak -- not to be dissing fvwm95). On the other hand, while WinLinux is technically impressive (slurping h/w settings out of the windows registry and translating that into Linux drivers & setup is decidedly non-trivial), not many
Elegant (in the classic sense) designs are always easy to use, and that's what we should be aiming towards. The opposition stems from the dumbing-down process not generating an elegant design. If you
Intuition is wrong enough frequently that it's not always a good design...
-_Quinn
* Admittedly, the defaults aren't particularly brilliant, but that's life.
** The major advance of Windows 3.1 was the GUI for applications, not for the system. This is where Win9x went horribly wrong. It's a given that the computer won't always be right, so treating cases where it isn't as exceptions is a Bad Idea and tends to break things. (That is to say, I
I know it's admitting to being horribly inadequate, but I don't know TeX, and find that WordPerfect 8 and/or AbiWord satisfy my writing urges completely (except for WP8's broken postscript generator). I know I'm not a Real Coder because I like using jcc over emacs or pico. (Though I do keep an xterm handy for doing the makes, because my Makefiles would give makemake the shaking fits.)
*** IE, we're not building a homogeneous system here; interoperability is going to be a bitch. IE, mounting is the Right Thing to do; is automounting? (For the "Why should I have to 'mount' CD-ROMs when they 'just work' in Windows?" questions.) It is the Right Thing to distribute applications as source; should we adjust the LSB so that there's a user-writable subset of directories for scripted installs? Should we ship our distro with '.rpm' as a MIME-type for netscape so we can 'just download' a binary-form application?
*Sigh.* Another article that fails to grasp the vitally important distinction between Artificial Intelligence and Machine Intelligence. Taking the burger-bot example, I order the "burger with everything, no pickles."
The robot has to (a) recognize what I said (voice recognition; a solved problem, given the enough hardware); (b) recognize what I actually want, the ketchup, onions, mustard, burger, buns, cheese, etc (expert system); and (c) put the burger together (vision system feeding an expert system that generates motion commands).
Everything but (c) has been done, and done well, albeit, incredibly expensively. This bot will look artificially look intelligent because you'll only ever see it in the Burger King.
Conventionally, this would be an artificial intelligence; so far, so good. However, AI conventionally also includes what are better called machine intelligences, machines on thinking and acting on a level at or beyond that of a 'normal human.' In other words, an MI (a machine capable of holding a conversation with you in a manner indistinguishable from a(nother) human) working at a burger joint would get just as bored and surly as the teenagers who work there now would -- which would defeat the point, no?
So why is this distinction so important? AI's might be able to learn, but they will be cheaper than MI's, definitely no citizen material, and mass-producable; each burger-bot is identical to the next, and comes with a certain store of 'knowledge' and 'skills' (databases and expert systems) that can be copied from machine to machine. (We're doing that now, so we know it works.)
What makes everyone think the same will hold for MIs? Why would an MI understand the hardware, firmware, and software it's constructed out of any more so than I understand my bioware? Suppose we overclock an MI's hardware by a factor of two... the MI is still only thinking twice as fast as normal human, not thinking anything more significant. Supposing that we manage to teach an MI how to design circuitry, what makes you think that it'll be any less prone to error than a human? It's still thinking its way through the design process, same as Joe IEEE Member does; granting a human perfect memory doesn't make him/her immune to error. An MI would be able to run, say, a chip validation program without thinking its way through it, but it would be using it the same way you and I would. An MI would not be anymore aware of what its processors are doing than you and I know what our neurons are doing; you think and your hands type, and there's a total concious disconnect between the two actions. Now, granting that conciousness in a finite-state machine (which I, for one, am not), there's still no gaurantee that you'll be able to save the state in such a fashion that it can be duplicated at your leisure on other hardware later.
I think the best you can hope for you with MIs is a very powerful subconcious, with the attendant powerful intuition. Data bases (Cyc, for example) would have either be searched conciously (like looking up every word I'm trying to type!) and hence uselessly, or subconciously, like I'm 'accessing my database' on English to crank out these paragraphs. A given MI might read a textbook on relativity and understand it, but there's no gaurantee, again, that it will be able to directly implant that knowledge in another MI; it would have become part of its particular neural net (non-transferrable, in general) or re-ordered its database (possibly transferrable, but the task of synchronizing the databases would be enourmous.) So an MI might be able to say, I'm guessing the problem is with the transmission, but unlike an expert system, it wouldn't be able to provide you with an exact trace of the logic behind that conclusion, unless it took the time to think it through 'conciously'.
Now, because of advances in h/w, an MI might be able to come up with that intuition in a tenth of the time a human might be able to, and 'learn' ten-times faster, but that only cuts the time it take to teach it 'from scratch' down to two or three years... and each additional job four or five months.
I just don't think an intelligent system can have full knowledge of itself; just on the storage level, a full image of itself would always take twice as much storage as it was currently using; the knowledge of its first-order self would not be exhaustive of its second-order self, and so on...
Futurists expect too much from the future.
-_Quinn