Quite frankly, it's because the origin stories are always the best ones. From one perspective, it's because they include all the stages of the monomyth. From another, because it's far easier to identify, as a normal Joe, with Peter Parker than with Spider-Man. Origin stories are also explicitly about character development. In contrast, the usual superhero plot is episodic, because the next writer has to be able to take over -- and any change away from the status quo implies the possibility that the series might end. (Equally cynically, if I'm doing a movie without character development, why pay the licensing fees for a character?) The drama (in the technical sense) of character development helps the origin story appeal to a wider audience.
Considering movie-making from the business point of view should answer your other question. If what makes your movie pitch attractive is a particular twist on, say, the concept of the responsible use of power, why not tell a real-world story which will attract a wider audience (and be cheaper to boot)? Spider-Man wants anonymity to avoid exposing Mary Jane to danger, but when J. Jonah Jameson publishes proof that Spider-Man killed a bunch of thugs by knocking them out on a beach below the high-tide line, what does he do when people rightfully demand some accountability? (Maybe the cops ignored it when he called them in. Maybe the tide was unusually high that night because SHIELD was experimenting with a new weapon. Maybe he just screwed up by the numbers. Maybe he was too busy disarming a terrorist nuke to get them. Who knows?) US Navy SEAL #17 wants anonymity to avoid exposing his wife to danger, too, but the New York Times publishes proof that he killed two dozen Americans that one night in Afghanistan, what does he do? (Maybe the embassy ignored him when he told them to come fetch their employees. Maybe they were hostages, killed by a freak mudslide while he was scouting the exfiltration route. (Maybe the mudslide was caused by SHIELD testing a new weapon.;)) Maybe he just screwed up by the numbers. Maybe he was too busy disarming a terrorist nuke to get them. Who knows?) You may gain some sharper moral contrasts by using a super-hero (Spider-Man has never killed anyone, but that the SEAL's job), but that's about it.
Likewise, the Bond movies provide a thin veneer of plausibility over the general trope of 'unlikely hero saves the world' -- but he's a trained super-spy! He's not unlikely! Everybody knows how good the British intelligence services are! Magic or the equivalent super-high tech just concentrate power in a visually-pleasing and obvious way, and let you "play for high stakes" with a very small and simple cast of characters. It's less a flaw of the genre and more a question of why Hollywood would ever bother to film anything in expensive genres that don't require that expense. (Aside from spectacle, the answer is usually aversion to political risk; consider Avatar.)
This doesn't seem like it could be the right explanation. Suppose you measure 2^10 of your electrons on the x-axis. Some time later, I come along and measure the equivalent electrons in two groups of 2^9 each, half on the x- and half on the y- axis. One of those two groups won't be correlated, and one will. (If the probability isn't high enough, use more electrons.) That gives me one bit of classical information with an arbitrary degree of confidence.
This is why people suggest VM-based backwards compatibility is the direction for MS to go. It's not quite trivial to do (well, maybe it would be if they bought VMWare for "Fusion"), but it would at least give them a chance to get the rest of the system right.
So on a technical level, Microsoft could certainly use some TPC magic to allow the keyboard and mouse drivers to emit signed events, so that you could be certain that they came from a person. (Well, from the hardware, anyway.) Similar magic would allow trusted libraries to sign the conversion from mouse event to command. At that point, you could avoid prompting the user to confirm their action, because you know they did it. Now the question is: for things that don't have direct OS-level call mappings (e.g., aren't Explorer), what do you do?
I think the principle of least astonishment applies here; kind of a variant of what another poster suggests with SELinux-style capability-base computing. There are certain things you probably don't expect applications to do (very few should send e-mail, for example). Figuring that list out and reducing 'false positives' is an ugly, ugly task. MS might be able to do it because the control the whole application stack, but if they design it finely-grained enough to be really useful, all the other application developers will try to lazy out of it and give themselves too many privileges, and become exploitable.
Maybe I just haven't looked hard enough, but I'd like is a fast site that worked/more/ like a hard-copy paper. It seems like every news site I've ever looked makes it as hard as possible to browse their news. I/want/ an editor whose biases are clear selecting stories, and I/want/ to be able to read the first paragraph or two of a given story without having to wait for a page load. Print newspapers fit two to four stories with multiparagraph lead-ins on the top half of their front page; CNN, for example, manages 1 lead-in, three headlines with summaries, and a dozen hard-to-read headlines in a column to the right. The other two-thirds of the width of the screen is totally wasted; scrolling down reveals a dozen exceedingly arbitrary categories with two headlines each. The BBC's new design is even worse.
Now, I was just thinking of using multiple columns and not having ads on the front page -- I mean, if I all want is headlines, I can get them a dozen other places and ways without going to your site -- but you could also do fun stuff like javascript articles in and out of the way. Now, I've never seen column-to-column wrapping done in a way that doesn't end up looking really silly. That being said, clicking on something [near what] I'm not going to read to pull the next article of that section into the same column is an interesting idea. I'll probably middle-click on an article I/do/ want to read, but you could do same thing on the single-article page, too. In terms of an emphasis on speed, I don't want shorter articles, I want to make looking for articles I'm interested in faster. (Yes, searching; but I need to know what to look for. That's part of the editor's job.)
At any rate, I feel like I spend more time navigating the site than reading at most news websites, specialized-interested ones excepted. (Ars Technica, for example, does a pretty good job of layout, although I'd like it to use more of the page horizontally and/to give a bit longer of a lead-in. But it can get away with a single chronologically-ordered column because of its narrow scope.) As I said earlier, I do/want/ something from the editors: the daily paper's website should decide what from the last day is important enough (given its well-known and undisguised biases) that I should know about it; I can get headlines from anywhere.
My primary objection to having a character's primary abilities be per-day limited is that it usually means that they don't do anything* every round. This is Not Fun, so it has to go. The implication is that there's good reason even for non-munchkins to "go to Motel 6" really frequently: the design of two of the four usual characters mean half the players stop having fun after some small number of encounters.
The extreme reaction to this was the Book of Nine Swords, where there were/no/ per-day limits. The problem with this is that it meant that the player did the same thing every fight. Part of that was flexibility; part of it was lack of resource management. As a general rule, I don't find "resource management" (which almost always means "speed" for adventuring parties) all that appropriate in every adventure of the high-heroic campaigns my group likes to run. In some cases, of course, it can be very dramatic and effective: can the party take the shortcut through the underground passage and warn somebody in time? (This is especially effective if the players are then expected to help the defense, as it gives them another factor to balance.) I think the key to doing per-day-limited abilities is going to be making the per-encounter and at-will abilities strong enough that the per-day abilities can be easily saved to compensate for bad luck. As the game is balanced now, you can't save your most powerful abilities for later, because you won't get there.
I think you think the 4E rogue will be only a DPS-machine because that's all they've mentioned about the class. But it's key; D&D is really about combat (mechanically). In the end, in our group, we've generally allowed the rogue to sneak attack whatever s/he wants (with the appropriate), because otherwise they just sit down and go home -- and that has exactly the same problem, if different causes -- than the wizard running out of spells.
FWIW regarding "everyone does damage", the books mention that they don't particularly like the name "blaster" for the wizard's role, because they recognize the importance of status effects and combat control -- the latter so much that they invented a new character class to make it happen! (The Defender role/really/ depends on the ability to control the battlefield, which is why 3E fighters who like the role and do well at it at low levels and in dungeons always end up disappointed.) I actually approve of "everyone does damage" to a certain extent; it's the most direct way of helping your buddies out in a fight. (Compare this to an ability like turning, which either wins the fight or does nothing. The alternate rule for doing damage is less flavorful, but better for teamwork and utility. A lich doesn't ignore the cleric because of its turning bonus; it's just much more likely to only take half damage. On the other hand, some of the pressure might fall of all-or-nothings if they're serious about balancing the game for multiple enemies; there's enough randomness (and/or ability difference) between the different enemy's saves that you can effectively have all-or-nothings with partial effects...)
- _Quinn
*: Yes, a high-level wizard has an effectively unlimited number of magic missiles. But they almost never matter in a CR-appropriate fight.
Actually, the final version(s) of this weapon system are intended to replace Tomahawks -- the article even mentions this towards the end. The production ammunition for the railgun will be GPS-guided, perhaps with some nifty high-tech scheme, probably with something lower-tech, like fins deploying as late as possible in the fight. It'll be substantially more expensive than conventional shells, but still an order of magnitude cheaper than cruise missiles -- and at Mach 8, will give less warning and be harder to intercept.
I think the Navy isn't actually all that interested in this as a ship-to-ship weapon; they don't any intention of allowing hostile surface ships within its maximum range of the carrier, whose planes can easily engage three times further away.
At least one of the new designs doesn't get 100% lift from its bag, and requires the airship to be moving to stay aloft. This eliminates most of the ground-handling problems: as long as the cargo weighs less than the missing lift, you can just turn the engines off and load and unload it as you would a plane or truck. Heck, it's a blimp, they have lots of space: go ahead and containerize the whole thing.
The big win is that we can more readily regulate emissions from stationary power plants (of higher efficiency that an IC engine), and weight is not a concern for the scrubbers and filters. Further along in the future, changing the problem from synthesizing gasoline to generating electricity more cleanly is an even bigger win.
The problem is that gesture recognition is a difficult problem in computer vision to begin with; using it to materially aid the translation process is at least as difficult a problem as spoken-language contextrual translation, and for the same reason. The real hope for cheap/fast/small translators is some sort of interactive device: the classic solution to hard AI problems is to make the human do the work. The device asks about if it recognized the sentence ("Did you write "The bathroom is in the volcano"? Please write "yes," write the sentence down again, or write a different one with the same meaning."), and then does a back-and-forth translation that's as literal as possible, and uses synonyms or definitions to check for figures of speech. ("OK. English isn't as cool as your language, so I'm going to have to say "The restroom is located inside in the lava-filled mountain," instead of what you wrote. Is that about right? (Please write "yes" or a correction.)") And so on.
You can expect the computer to do a good job looking up translations word-by-word, and recognizing simple grammatical constructs. If you can steer the users into forming simple literal sentences with basic vocabulary, the odds of getting a good translation should improve dramatically. It won't suffice for having a conversation, but live translators barely do.
I noticed the risk factor myself a long time ago; I couldn't figure out why my parents were so afraid to futz with things on the computer; I knew anything you changed you could change back. (Although some things would be harder than others.) I developed that initial confidence with floppy-based machines: in the very worst case, you take the disk out of the drive while the red light is off, before you do something totally stupid, and set it to read-only. Instant safety. No matter what happens, you can just restart the computer and get back to whatever you were doing. (OK, you saved your work first.)
I've read some more recent research about pretty much the same effect: male computer users, having higher 'self-efficacy' (they rank themselves are more likely to complete a given (computer) task before starting it), are more likely to 'twiddle' (experiment) with an application. Sometimes, this means productivity goes way down; other times, it goes way up (they experiment and find the tool to automate the task).
Reading PC written out again makes me wonder -- maybe it would help some people to/have/ those disks (-equivalent) back. A USB key with a switch on it that you plug into the computer when you want to do something. One way, "the computer box" can't write to your personal key, the other it can. Take all non-generic (Windows, Office) storage out of the box. Maybe even have different personal keys for different apps; plug in the iPod and get iTunes, plug in the camera, get iPhoto, plug in 'Word' key and get Word. You don't have to copy Word to the key; some sort of version-checking would suffice, so that if you stuck your personal key in another box somewhere else, it would complain as appropriate. You'd want like a four- or eight-port USB hub next to the monitor, and some sort of button on the hub or key 'hey, pay attention to me now.' You could buy blank keys at the store and have 'the box' turn them into Word keys or whatever, as appropriate. (And honestly, how many applications to people really tend to use? Brower, e-mail client, office software, a game program, and some random device-specific software.) Basically, make the box/act/ like the keys are like old-skool one-per-application floppies, even if the actual implementation is some two hard-drive only-one-live-at-a-time monstrosity so that you can always just turn it off and know that you didn't break anything.
I think that's what people mean when they talk about being like an appliance. There's no (nonobvious) penalty for turning it off and on again, and a low cost involved, and the appliance always recovers on it own.
I think Microsoft was trying to get at this with its Recovery Points, but they didn't sell it right and (AFAICT) it never worked right. I think the idea of some physically-seperable device could be important for people. (The 'smart card' / thin client systems that Sun (et alia) was selling a while ago have a whole nice list of features you should be able to bring along in a system like this. Location indepedence would be important for making it feel like the key has the application and your data.* ) It's intuitive.
-_Quinn
*: Technically, I suppose flash is cheap enough now that you could have a whole Linux distro on the key without wasting too much actual money, as opposed to space.
The economic benefit of a central bank is hardly arguable -- and as you point out, in the period the US was without a central bank, the European economies, which had central banks, were humming along quite nicely.
To be fair to (some of) the past governments responsible for the growth, the US also didn't maintain a standing army of any appreciable size at the beginning of the last century; didn't have a national bank (necessary for a modern economy; est. 1914) or, for that matter, federal insurance of deposit accounts (est. 1933 and ended runs on banks); didn't regulate the stock market (the SEC, est. 1934); didn't regulate telecommunication (universal access and interoperability) or automobile (safely); build or maintain modern roads (the automobile); act against fraudulent or unsafe foods or medicines (1906, FDA est. 1934); or have a federal law-enforcement agency (FBI est. 1908).
On the other hand, the US also didn't maintain a "war" on drugs, spy on its own people (as much, AFAIK), spy on everybody else (as much), etc.
From the point of view of revenue, however, the largest increase (roughly two-thirds of the current budget) is in the so-called "nondiscretionary" spending, almost all of it on Social Security, Medicare, and Medicaid, all programs enacted in the last century.
Some of this isn't as irrational as it sounds. Recommendations are a good way of deciding where to apply; there's a lot of choices, and generally not anything terribly obvious to distinguish them. Furthermore, if your school isn't a "name" school, as far as academics go, it's probably just one of many at about the right level of rigor for the applicant, so non-academic reasons become more important. If I'm going to spend four years someplace, I'd rather, all else roughly equal, spend it at the prettier place. There's a similar logic for "social scene," especially for a middle-tier university where you would expect a bunch of the applicants to mostly be looking for their paper, rather than to really have a strong intellectual life.
On the other hand, deciding to apply based on the school's sports teams' records is kind of disturbing. (Well, less so for applicants expecting to play college sports, but the vast majority really shouldn't...)
shouldn't they have defined a planet as, "In the Solar System, one of [Mercury,...,Pluto]; out of the solar system, a celestial body....", where the second clause is something along the lines of "spherical from its own gravity and not orbiting another planet." The subcategory of "double planet" could be defined by the barycenter being outside of either planet; the subcategory "dwarf planet" would, quite frankly, probably be arbitrary and something like "smaller than Mercury." You could then also have "minor planets" that are "smaller then Pluto."
As noted elsewhere in this thread, an individual Altix machine is ludricously expensive (even for an Itanium machine). However, the interconnect on an Altix extremely capable -- up to 512 processors* in a single system image, very low latency. For certain applications, this -- the NUMA model -- is very big win. Whether putting Itaniums into this very nice interconnect is cost-effective really depends on your specific application: how much cache does it need, its computation to communication ratio.
- _Quinn
*: I guess this means a 1024 cores, now that Montecito's out.
They're using unicast because of the hardware. I was just talking with Ron Minnich, from LANL, and in their tests, it was faster to do ad-hoc unicast trees than multicast in clusters with more than (IIRC) 128 nodes. I don't know if v9fs in particular uses ad-hoc unicast trees, but you don't need to have a global view of the machine at every node -- just at the control node, which reduces the cost to O(n).
Yes, bandwidth. Up until the PCIe bus, a pair of GigE ethernet cards saturated a PC's expansion bus. Until AMD built memory controllers into their chips, servers (read: non-x86 UNIX) crushed PCs in memory bandwidth. Until NCQ, SCSI drives crushed IDE drives in effective bandwidth.
So basically, yes, until very recently, there were very large and substantial bandwidth differences. They've gotten smaller. More important, however, are the "lights-out management" features. If you can't reinstall the OS from four floors away, it isn't a server.
The Apple IIGS in 1986 ran a full-color MacOS-equivalent (and superior, in some ways) called GS/OS very well, and it was essentially an underclocked Amiga. (The Apple IIGS also had very large ROMs; whether the Macintosh would've made the same impression it did had been released two years later I can't say.) Since it was totally compatible with Apple IIe (etc.) programs, it could have been the kind of "parlay" the article's author went on about, but it sank under Apple's neglect and unfathomable obsession with the the Macintosh.
Some (most that I know of, but that's a biased sample for obvious reasons) "classified" supercomputers aren't classified because of their hardware, but, for instance, because they run the nuclear bomb simulation codes, and must _not_ be accessible from outside its lab. I understand it's relatively common to run, e.g., LINPACK benchmarks during the set-up and initial tuning phase, before any of the classified applications are installed, and report the results. I believe this is the case for ASCI Q, which the last Top500 list put at ~13 teraflops (max, peak 20).
I think this is the same machine that was ranked #8 in the last Top500 list, but with more (the rest of its) processors installed.
It's not performing calculations. The specific example I'm familiar with is baseball players, for whom the algorithm for arriving where the ball will is to move to make it look like the ball is moving in a straight line (towards you). The relevant quote in this case is: "Asking if computers can think is like asking if submarines can swim."
Slashdot Mirror
Quite frankly, it's because the origin stories are always the best ones. From one perspective, it's because they include all the stages of the monomyth. From another, because it's far easier to identify, as a normal Joe, with Peter Parker than with Spider-Man. Origin stories are also explicitly about character development. In contrast, the usual superhero plot is episodic, because the next writer has to be able to take over -- and any change away from the status quo implies the possibility that the series might end. (Equally cynically, if I'm doing a movie without character development, why pay the licensing fees for a character?) The drama (in the technical sense) of character development helps the origin story appeal to a wider audience.
Considering movie-making from the business point of view should answer your other question. If what makes your movie pitch attractive is a particular twist on, say, the concept of the responsible use of power, why not tell a real-world story which will attract a wider audience (and be cheaper to boot)? Spider-Man wants anonymity to avoid exposing Mary Jane to danger, but when J. Jonah Jameson publishes proof that Spider-Man killed a bunch of thugs by knocking them out on a beach below the high-tide line, what does he do when people rightfully demand some accountability? (Maybe the cops ignored it when he called them in. Maybe the tide was unusually high that night because SHIELD was experimenting with a new weapon. Maybe he just screwed up by the numbers. Maybe he was too busy disarming a terrorist nuke to get them. Who knows?) US Navy SEAL #17 wants anonymity to avoid exposing his wife to danger, too, but the New York Times publishes proof that he killed two dozen Americans that one night in Afghanistan, what does he do? (Maybe the embassy ignored him when he told them to come fetch their employees. Maybe they were hostages, killed by a freak mudslide while he was scouting the exfiltration route. (Maybe the mudslide was caused by SHIELD testing a new weapon. ;)) Maybe he just screwed up by the numbers. Maybe he was too busy disarming a terrorist nuke to get them. Who knows?) You may gain some sharper moral contrasts by using a super-hero (Spider-Man has never killed anyone, but that the SEAL's job), but that's about it.
Likewise, the Bond movies provide a thin veneer of plausibility over the general trope of 'unlikely hero saves the world' -- but he's a trained super-spy! He's not unlikely! Everybody knows how good the British intelligence services are! Magic or the equivalent super-high tech just concentrate power in a visually-pleasing and obvious way, and let you "play for high stakes" with a very small and simple cast of characters. It's less a flaw of the genre and more a question of why Hollywood would ever bother to film anything in expensive genres that don't require that expense. (Aside from spectacle, the answer is usually aversion to political risk; consider Avatar.)
This doesn't seem like it could be the right explanation. Suppose you measure 2^10 of your electrons on the x-axis. Some time later, I come along and measure the equivalent electrons in two groups of 2^9 each, half on the x- and half on the y- axis. One of those two groups won't be correlated, and one will. (If the probability isn't high enough, use more electrons.) That gives me one bit of classical information with an arbitrary degree of confidence.
I can't resist: that's because reality has a well-known liberal bias. ;)
This is why people suggest VM-based backwards compatibility is the direction for MS to go. It's not quite trivial to do (well, maybe it would be if they bought VMWare for "Fusion"), but it would at least give them a chance to get the rest of the system right.
So on a technical level, Microsoft could certainly use some TPC magic to allow the keyboard and mouse drivers to emit signed events, so that you could be certain that they came from a person. (Well, from the hardware, anyway.) Similar magic would allow trusted libraries to sign the conversion from mouse event to command. At that point, you could avoid prompting the user to confirm their action, because you know they did it. Now the question is: for things that don't have direct OS-level call mappings (e.g., aren't Explorer), what do you do?
I think the principle of least astonishment applies here; kind of a variant of what another poster suggests with SELinux-style capability-base computing. There are certain things you probably don't expect applications to do (very few should send e-mail, for example). Figuring that list out and reducing 'false positives' is an ugly, ugly task. MS might be able to do it because the control the whole application stack, but if they design it finely-grained enough to be really useful, all the other application developers will try to lazy out of it and give themselves too many privileges, and become exploitable.
Maybe I just haven't looked hard enough, but I'd like is a fast site that worked /more/ like a hard-copy paper. It seems like every news site I've ever looked makes it as hard as possible to browse their news. I /want/ an editor whose biases are clear selecting stories, and I /want/ to be able to read the first paragraph or two of a given story without having to wait for a page load. Print newspapers fit two to four stories with multiparagraph lead-ins on the top half of their front page; CNN, for example, manages 1 lead-in, three headlines with summaries, and a dozen hard-to-read headlines in a column to the right. The other two-thirds of the width of the screen is totally wasted; scrolling down reveals a dozen exceedingly arbitrary categories with two headlines each. The BBC's new design is even worse.
Now, I was just thinking of using multiple columns and not having ads on the front page -- I mean, if I all want is headlines, I can get them a dozen other places and ways without going to your site -- but you could also do fun stuff like javascript articles in and out of the way. Now, I've never seen column-to-column wrapping done in a way that doesn't end up looking really silly. That being said, clicking on something [near what] I'm not going to read to pull the next article of that section into the same column is an interesting idea. I'll probably middle-click on an article I /do/ want to read, but you could do same thing on the single-article page, too. In terms of an emphasis on speed, I don't want shorter articles, I want to make looking for articles I'm interested in faster. (Yes, searching; but I need to know what to look for. That's part of the editor's job.)
At any rate, I feel like I spend more time navigating the site than reading at most news websites, specialized-interested ones excepted. (Ars Technica, for example, does a pretty good job of layout, although I'd like it to use more of the page horizontally and/to give a bit longer of a lead-in. But it can get away with a single chronologically-ordered column because of its narrow scope.) As I said earlier, I do /want/ something from the editors: the daily paper's website should decide what from the last day is important enough (given its well-known and undisguised biases) that I should know about it; I can get headlines from anywhere.
My primary objection to having a character's primary abilities be per-day limited is that it usually means that they don't do anything* every round. This is Not Fun, so it has to go. The implication is that there's good reason even for non-munchkins to "go to Motel 6" really frequently: the design of two of the four usual characters mean half the players stop having fun after some small number of encounters.
/no/ per-day limits. The problem with this is that it meant that the player did the same thing every fight. Part of that was flexibility; part of it was lack of resource management. As a general rule, I don't find "resource management" (which almost always means "speed" for adventuring parties) all that appropriate in every adventure of the high-heroic campaigns my group likes to run. In some cases, of course, it can be very dramatic and effective: can the party take the shortcut through the underground passage and warn somebody in time? (This is especially effective if the players are then expected to help the defense, as it gives them another factor to balance.) I think the key to doing per-day-limited abilities is going to be making the per-encounter and at-will abilities strong enough that the per-day abilities can be easily saved to compensate for bad luck. As the game is balanced now, you can't save your most powerful abilities for later, because you won't get there.
/really/ depends on the ability to control the battlefield, which is why 3E fighters who like the role and do well at it at low levels and in dungeons always end up disappointed.) I actually approve of "everyone does damage" to a certain extent; it's the most direct way of helping your buddies out in a fight. (Compare this to an ability like turning, which either wins the fight or does nothing. The alternate rule for doing damage is less flavorful, but better for teamwork and utility. A lich doesn't ignore the cleric because of its turning bonus; it's just much more likely to only take half damage. On the other hand, some of the pressure might fall of all-or-nothings if they're serious about balancing the game for multiple enemies; there's enough randomness (and/or ability difference) between the different enemy's saves that you can effectively have all-or-nothings with partial effects...)
The extreme reaction to this was the Book of Nine Swords, where there were
I think you think the 4E rogue will be only a DPS-machine because that's all they've mentioned about the class. But it's key; D&D is really about combat (mechanically). In the end, in our group, we've generally allowed the rogue to sneak attack whatever s/he wants (with the appropriate), because otherwise they just sit down and go home -- and that has exactly the same problem, if different causes -- than the wizard running out of spells.
FWIW regarding "everyone does damage", the books mention that they don't particularly like the name "blaster" for the wizard's role, because they recognize the importance of status effects and combat control -- the latter so much that they invented a new character class to make it happen! (The Defender role
- _Quinn
*: Yes, a high-level wizard has an effectively unlimited number of magic missiles. But they almost never matter in a CR-appropriate fight.
Actually, the final version(s) of this weapon system are intended to replace Tomahawks -- the article even mentions this towards the end. The production ammunition for the railgun will be GPS-guided, perhaps with some nifty high-tech scheme, probably with something lower-tech, like fins deploying as late as possible in the fight. It'll be substantially more expensive than conventional shells, but still an order of magnitude cheaper than cruise missiles -- and at Mach 8, will give less warning and be harder to intercept.
I think the Navy isn't actually all that interested in this as a ship-to-ship weapon; they don't any intention of allowing hostile surface ships within its maximum range of the carrier, whose planes can easily engage three times further away.
At least one of the new designs doesn't get 100% lift from its bag, and requires the airship to be moving to stay aloft. This eliminates most of the ground-handling problems: as long as the cargo weighs less than the missing lift, you can just turn the engines off and load and unload it as you would a plane or truck. Heck, it's a blimp, they have lots of space: go ahead and containerize the whole thing.
The big win is that we can more readily regulate emissions from stationary power plants (of higher efficiency that an IC engine), and weight is not a concern for the scrubbers and filters. Further along in the future, changing the problem from synthesizing gasoline to generating electricity more cleanly is an even bigger win.
Turns out that chicks aren't free.
Recall the phrase "regular army"? Back in the day "regulated" meant "trained." That is, in modern English, the Amendment would read:
"A well-trained militia being necessary to the security of a free State, the right of the People to keep and bear arms shall not be infringed."
Which doesn't answer the question, but gives you a better idea of what was really meant by the justification half of the Amendment.
The problem is that gesture recognition is a difficult problem in computer vision to begin with; using it to materially aid the translation process is at least as difficult a problem as spoken-language contextrual translation, and for the same reason. The real hope for cheap/fast/small translators is some sort of interactive device: the classic solution to hard AI problems is to make the human do the work. The device asks about if it recognized the sentence ("Did you write "The bathroom is in the volcano"? Please write "yes," write the sentence down again, or write a different one with the same meaning."), and then does a back-and-forth translation that's as literal as possible, and uses synonyms or definitions to check for figures of speech. ("OK. English isn't as cool as your language, so I'm going to have to say "The restroom is located inside in the lava-filled mountain," instead of what you wrote. Is that about right? (Please write "yes" or a correction.)") And so on.
You can expect the computer to do a good job looking up translations word-by-word, and recognizing simple grammatical constructs. If you can steer the users into forming simple literal sentences with basic vocabulary, the odds of getting a good translation should improve dramatically. It won't suffice for having a conversation, but live translators barely do.
I noticed the risk factor myself a long time ago; I couldn't figure out why my parents were so afraid to futz with things on the computer; I knew anything you changed you could change back. (Although some things would be harder than others.) I developed that initial confidence with floppy-based machines: in the very worst case, you take the disk out of the drive while the red light is off, before you do something totally stupid, and set it to read-only. Instant safety. No matter what happens, you can just restart the computer and get back to whatever you were doing. (OK, you saved your work first.)
/have/ those disks (-equivalent) back. A USB key with a switch on it that you plug into the computer when you want to do something. One way, "the computer box" can't write to your personal key, the other it can. Take all non-generic (Windows, Office) storage out of the box. Maybe even have different personal keys for different apps; plug in the iPod and get iTunes, plug in the camera, get iPhoto, plug in 'Word' key and get Word. You don't have to copy Word to the key; some sort of version-checking would suffice, so that if you stuck your personal key in another box somewhere else, it would complain as appropriate. You'd want like a four- or eight-port USB hub next to the monitor, and some sort of button on the hub or key 'hey, pay attention to me now.' You could buy blank keys at the store and have 'the box' turn them into Word keys or whatever, as appropriate. (And honestly, how many applications to people really tend to use? Brower, e-mail client, office software, a game program, and some random device-specific software.) Basically, make the box /act/ like the keys are like old-skool one-per-application floppies, even if the actual implementation is some two hard-drive only-one-live-at-a-time monstrosity so that you can always just turn it off and know that you didn't break anything.
I've read some more recent research about pretty much the same effect: male computer users, having higher 'self-efficacy' (they rank themselves are more likely to complete a given (computer) task before starting it), are more likely to 'twiddle' (experiment) with an application. Sometimes, this means productivity goes way down; other times, it goes way up (they experiment and find the tool to automate the task).
Reading PC written out again makes me wonder -- maybe it would help some people to
I think that's what people mean when they talk about being like an appliance. There's no (nonobvious) penalty for turning it off and on again, and a low cost involved, and the appliance always recovers on it own.
I think Microsoft was trying to get at this with its Recovery Points, but they didn't sell it right and (AFAICT) it never worked right. I think the idea of some physically-seperable device could be important for people. (The 'smart card' / thin client systems that Sun (et alia) was selling a while ago have a whole nice list of features you should be able to bring along in a system like this. Location indepedence would be important for making it feel like the key has the application and your data.* ) It's intuitive.
-_Quinn
*: Technically, I suppose flash is cheap enough now that you could have a whole Linux distro on the key without wasting too much actual money, as opposed to space.
The economic benefit of a central bank is hardly arguable -- and as you point out, in the period the US was without a central bank, the European economies, which had central banks, were humming along quite nicely.
To be fair to (some of) the past governments responsible for the growth, the US also didn't maintain a standing army of any appreciable size at the beginning of the last century; didn't have a national bank (necessary for a modern economy; est. 1914) or, for that matter, federal insurance of deposit accounts (est. 1933 and ended runs on banks); didn't regulate the stock market (the SEC, est. 1934); didn't regulate telecommunication (universal access and interoperability) or automobile (safely); build or maintain modern roads (the automobile); act against fraudulent or unsafe foods or medicines (1906, FDA est. 1934); or have a federal law-enforcement agency (FBI est. 1908).
On the other hand, the US also didn't maintain a "war" on drugs, spy on its own people (as much, AFAIK), spy on everybody else (as much), etc.
From the point of view of revenue, however, the largest increase (roughly two-thirds of the current budget) is in the so-called "nondiscretionary" spending, almost all of it on Social Security, Medicare, and Medicaid, all programs enacted in the last century.
Some of this isn't as irrational as it sounds. Recommendations are a good way of deciding where to apply; there's a lot of choices, and generally not anything terribly obvious to distinguish them. Furthermore, if your school isn't a "name" school, as far as academics go, it's probably just one of many at about the right level of rigor for the applicant, so non-academic reasons become more important. If I'm going to spend four years someplace, I'd rather, all else roughly equal, spend it at the prettier place. There's a similar logic for "social scene," especially for a middle-tier university where you would expect a bunch of the applicants to mostly be looking for their paper, rather than to really have a strong intellectual life.
On the other hand, deciding to apply based on the school's sports teams' records is kind of disturbing. (Well, less so for applicants expecting to play college sports, but the vast majority really shouldn't...)
shouldn't they have defined a planet as, "In the Solar System, one of [Mercury,...,Pluto]; out of the solar system, a celestial body ....", where the second clause is something along the lines of "spherical from its own gravity and not orbiting another planet." The subcategory of "double planet" could be defined by the barycenter being outside of either planet; the subcategory "dwarf planet" would, quite frankly, probably be arbitrary and something like "smaller than Mercury." You could then also have "minor planets" that are "smaller then Pluto."
As noted elsewhere in this thread, an individual Altix machine is ludricously expensive (even for an Itanium machine). However, the interconnect on an Altix extremely capable -- up to 512 processors* in a single system image, very low latency. For certain applications, this -- the NUMA model -- is very big win. Whether putting Itaniums into this very nice interconnect is cost-effective really depends on your specific application: how much cache does it need, its computation to communication ratio.
- _Quinn
*: I guess this means a 1024 cores, now that Montecito's out.
They're using unicast because of the hardware. I was just talking with Ron Minnich, from LANL, and in their tests, it was faster to do ad-hoc unicast trees than multicast in clusters with more than (IIRC) 128 nodes. I don't know if v9fs in particular uses ad-hoc unicast trees, but you don't need to have a global view of the machine at every node -- just at the control node, which reduces the cost to O(n).
Yes, bandwidth. Up until the PCIe bus, a pair of GigE ethernet cards saturated a PC's expansion bus. Until AMD built memory controllers into their chips, servers (read: non-x86 UNIX) crushed PCs in memory bandwidth. Until NCQ, SCSI drives crushed IDE drives in effective bandwidth.
So basically, yes, until very recently, there were very large and substantial bandwidth differences. They've gotten smaller. More important, however, are the "lights-out management" features. If you can't reinstall the OS from four floors away, it isn't a server.
The Apple IIGS in 1986 ran a full-color MacOS-equivalent (and superior, in some ways) called GS/OS very well, and it was essentially an underclocked Amiga. (The Apple IIGS also had very large ROMs; whether the Macintosh would've made the same impression it did had been released two years later I can't say.) Since it was totally compatible with Apple IIe (etc.) programs, it could have been the kind of "parlay" the article's author went on about, but it sank under Apple's neglect and unfathomable obsession with the the Macintosh.
- _Quinn
Some (most that I know of, but that's a biased sample for obvious reasons) "classified" supercomputers aren't classified because of their hardware, but, for instance, because they run the nuclear bomb simulation codes, and must _not_ be accessible from outside its lab. I understand it's relatively common to run, e.g., LINPACK benchmarks during the set-up and initial tuning phase, before any of the classified applications are installed, and report the results. I believe this is the case for ASCI Q, which the last Top500 list put at ~13 teraflops (max, peak 20).
I think this is the same machine that was ranked #8 in the last Top500 list, but with more (the rest of its) processors installed.
- _Quinn
I thought everybody called them slidewalks.
- _Quinn
It's not performing calculations. The specific example I'm familiar with is baseball players, for whom the algorithm for arriving where the ball will is to move to make it look like the ball is moving in a straight line (towards you). The relevant quote in this case is: "Asking if computers can think is like asking if submarines can swim."
- _Quinn