As I understand it -- and I may be off base here -- when leaded gasoline was phased out, fuel refineries had to stop using cheap lead as a substitute for expensive octane.
"Pure" gasoline, before additives, is a hodgepodge of various no- and low-branch alkanes that averages around 6 to 7 carbon atoms per molecule. But linear heptane burns a bit too fast for the liking of a lot of engines, and hydrocarbons with even slight branching burn much faster than that. Enter "octane", which burns slower than "pure" gasoline and autoignites at a higher temperature, thus preventing premature detonation, or "knocking". ("Octane" is in quotes because, in the context of gasoline, "octane" is a function of the burn speed, not the number of carbon atoms. "Octane" doesn't strictly mean 8-carbon linear alkane, as it can also include heavier molecules with a small amount of branching.)
Raising the "octane rating" requires a more involved refining process, because the alkanes in the gasoline-diesel-kerosene spectrum all have very similar boiling points and are hard enough to distill cleanly when making "pure" gasoline. 8-carbon molecules are toeing the line, and heavier ones push pretty solidly into diesel territory. As an educated guess, refining high-octane gasoline probably cuts into diesel yields and reduces overall profits. It probably also requires a more expensive hydrocarbon cracking process, because branches are even less desirable in high-octane gasoline than they are in "pure" gasoline, and a lot of catalytic cracking processes are quite happy to spew out branched hydrocarbons.
Presumably, the reason "regular unleaded" gasoline is more expensive than "regular leaded" is the same reason that high-octane grades of gasoline are called "premium", and priced accordingly.
Actually, Google's pretty big on dogfooding stuff. Google's internal e-mail is most definitely the Google Apps flavor of GMail. Most of the features that Google adds to the public site have been stress-tested by deploying them internally, where Google's tens of thousands of employees can use them, abuse them, and file bugreports.
If I had to take a wild guess, you hit an anti-malware heuristic. For a while, a lot of malware e-mails consisted of an encrypted.ZIP file, with the decryption password in the body. Apparently this confused a significant number of people into believing that the.ZIP file's contents "must be" legitimate, so they ran arbitrary executable code. This, of course, also hid the payload from anti-virus software. As a result, a lot of anti-malware filters now suspect any message with an encrypted.ZIP attachment.
Besides, you should never, ever rely on the encryption features built in to any.ZIP program. The traditional PKZip/WinZip encryption algorithms are nearly useless and straightforward to brute force. If you don't like public key encryption, use GPG with a symmetric key.
They have wronged by handing out monopolies and they have wronged by subsidising them. Another wrong isn't going to fix the system. Just allow proper competition.
Oh, sure. All we need is for 10 companies to each separately dig their own trenches and lay their own fiber optic cable, at a duplicated cost of 10*$billions plus another order of magnitude in economic disruption. At that point, the Capitalism Fairy will magically turn an oligopoly of 10 large corporations into a healthy competitive market where no single player has the power to manipulate market prices.
And I have some prime beachfront real estate in Montana to sell you.
The reality is that phone lines, co-ax, and fiber optic cables share certain important traits with power lines, water mains, and sewage pipes. Namely: 1. The up-front installation costs are huge, which limits the number of players. (No entrepreneurs need apply, and therefore innovation is restricted.) 2. If each company needed to build separate infrastructure to each individual house, people would riot in the streets from all the closed roads, torn-up back yards, NIMBYist-angering overhead lines, and other offenses.
Situations like these, namely "utilities", are treated differently from the rest of the market. Why? Because they act differently from the rest of the market. They are natural monopolies: in the absence of government regulation, there tends to be only one provider for each of them. It is simply not possible for an unregulated market to have healthy competition in situations like these, and no amount of praying to Ayn Rand will make this not so.
The reality that the libertarian-utopia crowd doesn't want to acknowledge is that the best course of action with Internet access is probably to unprivatize the infrastructure, buying it back from the private local monopolists, then allowing free market competition to flourish among companies who provide Internet access on top of that public infrastructure. (In a sense, it ought to be public property already, since public tax dollars paid for huge chunks of it -- particularly the parts owned by the Bells. However, at this point it's a done deal, and for obvious reasons of fairness, contract law doesn't permit backsies.)
In a public-infrastructure/private-service scenario, bonus points are awarded if the Internet access companies running on top of that public infrastructure are permitted to bypass government inefficiencies and fix/upgrade the equipment themselves, but only if they are willing to pay for it themselves and with the understanding that their competitors will also reap the benefits. If the government does an inadequate job with maintenance, then in all likelihood this would quickly create a funding pool for infrastructure improvements that would, in some ways, be vaguely similar to the Free Software movement. (Do you think big software companies like Google hand out free code because they dislike money? No, they do it because it benefits the community, and thus benefits themselves. Even though open sourcing their code costs them in profit margins due to increased competition, the improvements (and testing! testing is expensive!) they get back from the community saves them even more money than they lose. It's the beginning of corporate ethics, the sorts of blindingly obvious insights that humans figured out thousands of years ago, but corporations are only just now learning. This is a good thing, and should be encouraged whenever possible.)
Disclaimer: as of recently, I now work for Google, but I don't speak for Google in any way, and I don't have anything to do with web search. I had noticed this and pondered it before I started my employment.
I'm pretty sure they didn't do this in the past, as I remembered looking at the source of the page and thinking "that is some lean html right there". Now, however.. they must be wasting gigabytes a day thanks to links like that. But I suppose they can afford it, and for them it's a nice metric to see which links got clicked, etc.
My educated guess is that they're doing it to count hits, presumably so they can (a) collect stats about how far people go down the list of search results, and more importantly (b) collect stats about how accurate the search results were: if the first link was clicked quickly, it must've been a good result; if there are long delays, lots of links clicked, and the first 3 results were never clicked, it must've been an awful result.
Again, this is without any insider insights into what the web search team is actually doing, mind you, so take what I say with a grain of salt.
Regarding bloat: it's not as bad from a bandwidth perspective as you make it out, but the added latency was very annoying for me when they first started doing that. Adding another round trip to Google before I can see my result does have an annoyance impact. There are a few Greasemonkey scripts out there that purport to undo the Javascript and purify the result links; however, I never got any of them to work for me.
Hopefully Google's web search team is working on something more asynchronous/AJAX-y that can report back in the background.
MD5 has been cracked. That is to say, there are known methods of creating a file with a high probability of having the same MD5 as some original file.
This is not correct. There *is* an attack that permits a person to create two or more new files that share the same MD5 hash. (That the resulting files have the same MD5 hash is a certainty, not a mere "high probability".)
However, there *is not* an attack that permits a person to create one new file that has the same MD5 hash as an existing file. Files with chosen MD5 hashes are still not a reality (for now).
That said, it's time to smile and nod at MD5 while making a slow, non-threatening retreat. Sadly, SHA-1 is having similar problems, SHA-2 is based on many of the same SHA-1 and MD5 principles, and it's not clear that any of the alternative hash algorithms is built on more solid ground.
What really interests me is how viciously the
innoculation-is-good meme is defended.
Oh, well yeah, it's totally irrational to believe that innoculation can prevent disease. I mean, it only ERADICATED SMALLPOX.
Piffle, that. Every *day* I eradicate six impossible diseases before breakfast with carefully-formulated traditional medicine homeopathic organic vegan acupuncture shakes, down at the local Oxygen Bar. Beat that, innoculation!
Since you're obviously eager and willing to take a mercury injection...
I just ate a full maki roll of tuna sushi this week. That's more mercury than any one of the childhood shots I received back in the 80's (which still contained thiomersal back then).
You can't live without ingesting any of a hundred different poisons each day. That's why we have livers and kidneys. So long as the poisons don't overwhelm our bodies' abilities to eliminate and dispose of the poisons, we don't even notice.
I have little to no sympathy with Seidel. Thiomerisol, a mercury(!) compound, deals enormous damage
to a child's (and an adult's) brain. Basically it boils down to a needle full of lobotomy.
If she is defending Thiomerisol then either she hasn't done her homework or knowing the facts
she is on their payroll.
What the fuck hyperbole train did you just ride in on?
The amount of ethyl mercury in a dose of vaccine is tiny, and ethylmercury is eliminated so quickly (half-life 18 days or less) that it does not bioaccumulate. You're putting your kid in more danger by feeding them a tuna sandwich once a week than you are by giving them the standard childhood vaccinations, even if you go back in time to 1998 before the US started phasing out thiomersal. Unlike ethylmercury, methylmercury does build up in the body (half-life 44 days), and methylmercury is found in tuna and other large, long-lived ocean fish. (It's also found in large, long-lived land mammals like humans, and babies receive noteworthy amounts of mercury through breast milk.)
On top of that, thiomersal has been phased out of the routine childhood vaccines for years now. There was no resulting drop in autism rates; there was no resulting drop in mercury poisonings; there was no resulting increase in cognitive function, or test scores, or any measurable thing whatsoever. All the available evidence shows that removing thiomersal did absolutely nothing.
Hey, now! Mortal Kombat was actually fun if you didn't stare at it too hard, Resident Evil was tolerable for at least one viewing, and Street Fighter was... well, Street Fighter had MST3K comedy value. (Oh my God! They killed Raul Julia! You bastards!)
Having deliberately not watched a Boll movie from start to finish due to an utter lack of appeal for what I saw, I can't say for sure that they don't have any redeeming qualities, but they certainly seem at least one notch lower on the bad movie hierarchy than that.
I mean, sure, Paul Anderson has turned out some utter shite -- AvP comes to mind as one of his shiniest polished turds -- but he's never sunk to the Highlander-level depths of "Oh my God, now I have to invent a time machine just to retroactively kill myself and thereby never see this movie", or even to Vercingétorix levels of "Please remind me again why we're watching this".
Modern Intel motherboards (i810 forward) and AMD motherboards (768 forward) have a hardware RNG (Random Number Generator) that IIRC is based on diode noise. That's straight up quantum randomness, and most modern Linux distros automatically detect and use it if available.
Oh, come on. Header files -- at least, header files that implement APIs -- consist only of facts. A list of #defines and structs is a list of facts: it is not code, it is not a creative work, and it cannot be copyrighted. You cannot change the names of the struct members or the values of the #define constants while still following the API; the API restricts you to mechanically regurgitating the facts of how the API is defined.
SCO's big claim in the news media, back when they first announced their War On Linux, was that Linux was infringing on the Unix copyrights; their argument basically amounted to: "See, their header files are just like ours, once you strip out the comments". Which was utter BS, because the Linux headers implement POSIX and other standard APIs, and to implement POSIX the headers must contain exactly that code, word for word, because the POSIX API says so. If SCO had actually brought their BS to court, the judge would've laughed them out of the courtroom; and if Microsoft were to sue ReactOS over a silly header file, they'd face exactly the same reaction.
It's just as silly as saying that the NetBSD Editline library is violating the GPL because it implements the same API as the GNU Readline library. If you run both files through a C source code munger that eliminates comments and normalizes whitespace, they will be substantially identical — since they implement the same API, it's impossible for them to be broadly different, because there is no creativity (and therefore no copyright) in writing a header file that defines an API.
[Historical aside: Readline was placed under the GPL (instead of the LGPL) because GNU wanted to strongarm... er, "incentivize"... people into releasing their code under the GPL. This was because, if a closed-source program wanted to load and use Readline, it had to follow the Readline API — of which GNU Readline was the only implementation, which (the FSF argued) meant the closed-source program was really a derivative work of GNU Readline, and thus subject to the GPL. But it also applied to BSD-licensed open source code, so the NetBSD folks wrote their BSD-licensed Editline knockoff. Now that Editline exists, a program can use the Readline API while being clearly outside the scope of the GPL.]
Now, if you have the Windows source code repository available to you, and you can verify that the ReactOS implementation is substantially similar to the Windows implementation, then your argument has legs. Until then, stop with the FUD.
Well, you can check for string length before applying one or another snippet.
GCC already does that; that's my entire point. You can't just look at one benchmark on copying long strings and say "Well, there's obviously no situation where the REP instructions are faster", which is how I read your previous post.
Instead, they've got compile-time logic that spots fixed sizes (e.g. memcpy(dst, src, sizeof(struct foo));) and inlines the appropriate code for the target CPU, and then they've got more flexible logic for when the length is variable that does a few quick size comparisons at runtime. (They already need to check for word alignment anyway, so it's not outrageous overhead.)
I saw a previous Slashdotter claim that the CLD instruction can take ~50 cycles on a modern processor (presumably because the pipeline has to clear before %EFLAGS can change). If that's true, then it's very understandable that the GCC folks would be chomping at the bit to strip those CLD's out wherever they can, because then the REP loops are reasonable in many more scenarios.
But what about performance for shorter strings? Which has the greater setup overhead, REP or vector? And on which CPUs?
These are the sorts of questions that GCC developers have to code for, and they've already decided that the REP instructions are better in some situations.
Now some geniuses have tagged it privacy - what does the state erasing a thumb drive it owns have to do with privacy?
RTFA. The reason the state is issuing these new fancy-schmancy thumb drives is that the new ones (claim to) have 256-bit AES encryption and (claim to) self-destruct after 10 consecutive wrong passwords. They're doing this whole switch because of privacy, because the thumb drives contain the private, personal case files of hundreds/thousands of citizens.
I like how you try to equate General Relativity with Reality. GR is a mathematical model that fits much of the data pretty well, but it is hardly an explanation of what is actually happening. People tend to forget little details like that.
I didn't forget that. But we've observed gravitational lensing. Gravitational lensing is an observational fact. No other theory before General Relativity had predicted gravitational lensing, and what's more, the amount of gravitational lensing we see is precisely consistent with General Relativity. That means, to the best of our ability to detect, General Relativity describes reality itself, and any new theory of gravity must describe reality at least equally well -- just as General Relativity would have been rejected outright if it were inconsistent with the facts known under Newtonian gravity.
I don't think it's possible to overestimate how difficult it is to pick some Joe Random's Theory of Gravity and have that theory predict that light itself, which has no mass, is somehow being attracted to things that do have mass. (Or, similarly bizarrely, that light has no mass yet does have momentum.) The String Theory family was carefully formulated such that it's guaranteed to reduce to General Relativity under the correct conditions. The TeVeS version of MOND might work. But, no matter how you slice it, just fiddling with the knobs of Newtonian gravity won't result in a theory compatible with Lorentz invariance and the Maxwell equations, unless you set out to do that from the start like Einstein did or the String Theory folks did, or perhaps unless you're very smart and very patient like the MOND folks are trying to do with TeVeS.
GR says that lensing of this sort is always caused by matter, and more matter causes stronger lensing (related by an equation). We don't see any matter there. Therefore, if there is matter, that matter neither emits nor appreciably absorbs light (i.e. "dark matter"), and GR says there must be matter where there is lensing. What it would take to unseat GR in the dark matter debate is a theory that simultaneously explains both (a) lensing by actual matter, and (b) lensing around empty patches of space that somehow lens light while containing no observed matter. It also needs to explain why nearly all galaxies spin at the same speed at the edges as they do near the center, while NGC 4736 spins slower at the edges than it does near the center.
In the past, people ran around saying that the scientists who proposed Quantum Mechanics were just "fooling themselves", running around believing that the model was reality. Well, the problem with that is that the model is DAMN good at predicting reality. The computer that you use to read Slashdot is proof enough of that -- semiconductor physics depends on a rather large amount of quantum mechanics, from the P-N junction downward. Intel and AMD exist, your computer exists, therefore QM is "true". Gravitational lensing exists, there is no observable matter there, therefore Dark Matter is "true". The observations stack up to the point that any new theory could only be a refinement of the old theory, like Einstein's was to Newton's.
Back when we were just monkeying around with redshifts and galactic rotation curves, MOND was a very reasonable model. But the Bullet Cluster, NGC 4736, and now this direct observation of dark matter filaments is entirely incompatible with MOND. MOND supporters readily acknowledge this -- MOND can't explain the Bullet Cluster without claiming that force can act in a different direction than acceleration (F=ma being a vector equation), which no MOND supporter is willing to countenance. (Last I heard, the MOND proponents were hoping that warm neutrinos plus MOND might be enough, though. Neutrinos are dark matter, but fairly mundane dark matter.) At this point, after the last 2 years of shocking observations, there IS no other theory left standing that fits th
Because Lisp wasn't so much invented as discovered. It was a mathematical formulation, with a toy syntax, meant for pure mathematical work, that was never meant to be implemented as a real programming language. Then a grad student had the bright idea to write eval in machine code... thus creating the first interpreted language. Ever.
Those same mathematical principles apply to any programming language, and thus each new language is just slightly more Lisp-like than the one it replaces. Java is C++ with garbage collection (invented in Lisp). Perl added closures (invented in Lisp) to many peoples' everyday vocabularies -- every time you type "map {... }" or "sort {... }" in Perl, that's Lisp peeking through. Perl 6 is going the extra step of adding Lisp-style macros and a user-modifiable grammar, thus making it (in the mathematical sense) a dialect of Lisp, albeit one with a funny syntax.
Go read Paul Graham's What Made Lisp Different, then claim that it's just a coincidence that every new programming language borrows more and more ideas from something invented in the 1950s. Then do yourself a favor and buy Higher-Order Perl.
If our fundamental laws are a bit off, then this bending/distorting of the light would be explained by that...
That's the thing: no, it wouldn't!
Have you even read the Wikipedia article on MOND? MOND is the only serious proposal to modify our laws of gravity. It proposes a change in F=ma, such that tiny accelerations result in a surprisingly large force. This would not affect the amount of mass, and thus would not cause a change in gravitational lensing. Any observation of gravitational lensing in a place where there's no visible matter is ipso facto evidence in favor of dark matter and evidence in disfavor of MOND.
Beyond MOND, no other modification of gravity has yet been proposed that's consistent with General Relativity. (Even MOND might not be, depending on how well TeVeS works. AFAIK, the math is still under scrutiny.) And without compatibility with General Relativity, gravitational lensing is impossible, so it contradicts reality.
At this point, thanks to the Bullet Cluster and other observations, even MOND supporters acknowledge that dark matter exists at the scale of galactic clusters and up. The only question remaining is, at the scale of a single galaxy, whether the galactic rotation curve is due to MOND or due to a halo of dark matter around the galaxy. And that itself is a point of fresh debate ever since the recent observations of NGC 4736, a galaxy that follows the classic (non-MOND) Newtonian predictions. Dark matter theory can easily account for this by suggesting that this one galaxy was somehow separated from the dark matter halo that once surrounded it; MOND (or any other modification to gravity) must somehow explain why this one galaxy does not conform to the laws of physics.
At this point, the existence of dark matter is as good as proven.
I call bullshit. It's not scientists putting this crap out there. It's bad science reporting. If it's not a scary disease or a global disaster, then it gets about 3 words (not in a row) of actual scientific content, surrounded by fuzzy blather and bad analogies. (If it is a scary disease or a global disaster, then it gets 6 scientifically meaningful words, 4 of them wrong.)
I'm reminded of an anecdote (attribution lost, possibly apocryphal) of a politician being briefed on the progress of the SDI/Star Wars project. The scientist is describing the current technology, saying that after years of effort they can fire a laser that delivers 10^n W (for some n that I forget) to a target. However, for the laser to be useful, they need a laser than can deliver 10^2n W. The politician, not understanding exponentials, remarks: "My God, we're halfway there!"
My gut feeling is that, from strictly a hardware perspective, we're already capable of building a human-level AI. The problem is that, from a software perspective, we've focused too much on approaches that will never work.
As far as I'm concerned, the #1 problem is the Big Damn Database approach, which is basically a cargo cult in disguise. Though expert systems are useful in their niches, "1. Expert system 2. ??? 3. AI!" is not a workable roadmap to the future. I'm certain that it's far easier to start with an ignorant AI and teach it a pile of facts than it is to start with a pile of facts and teach it to develop a personality.
The #2 problem is the Down To The Synapse approach. This, unlike BDD, could quite possibly create "A"I if given enough hardware. But I think that, while DTTS will lead to a better understanding of medicine, it won't advance the AI field. It won't lead to an improved understanding of how human cognition works — it certainly won't teach us anything we didn't already know from Phineas Gage and company.
Even if we go to all the trouble of developing a supercomputer capable of DTTS emulation of a human brain — so what? If we ask this emulated AI to compute 2+2, millions of simulated synapses will fire, trillions of transistors will flip states, phenomenal amounts of electricity will pour into the supercomputer, just for the AI to give the very same answer that a simple circuit consisting of a few dozen transistors could've answered in a tiny fraction of the time, using the amount of electricity stored on your fingertip when you rub your shoes on the carpet during winter. And that's not even a Strong AI question. That's not to say that working DTTS won't be profound in some sense, but we know we can build it better, yet we won't have the faintest idea of where to go next.
That brings me to my core idea — goals first, emotions close behind. Anyone who's pondered the "is/ought" problem in philosophy already knows the truth of this, even if they don't know they know the truth of it. The people building cockroach robots were on the right track all along; they're just thinking too small. MIT's Kismet, for instance, gives an idea of where AI needs to head.
That said, I think building a full-on robot like Kismet is premature. A robot requires an enormous number of systems to process sensory data, and those processing systems are largely peripheral to the core idea of AI. If we had an AI already, we could put the AI in the robot, try a few things, and ask the AI what works best. So, ideally, I think we need to look at a pure software approach to AI before we go off building robot bodies for them to inhabit.
And how to do that? I think Electric Funstuff's Sim-hilarities captures the essence of that. If we give AIs a virtual world to live in — say, an MMO — then that removes a lot of the need for divining meaning from sensory input, allowing a sharper focus on the "intelligence" aspect of AI. Start with that, grow from there, and I can definitely see human-level AI by 2029.
Slashdot Mirror
As I understand it -- and I may be off base here -- when leaded gasoline was phased out, fuel refineries had to stop using cheap lead as a substitute for expensive octane.
"Pure" gasoline, before additives, is a hodgepodge of various no- and low-branch alkanes that averages around 6 to 7 carbon atoms per molecule. But linear heptane burns a bit too fast for the liking of a lot of engines, and hydrocarbons with even slight branching burn much faster than that. Enter "octane", which burns slower than "pure" gasoline and autoignites at a higher temperature, thus preventing premature detonation, or "knocking". ("Octane" is in quotes because, in the context of gasoline, "octane" is a function of the burn speed, not the number of carbon atoms. "Octane" doesn't strictly mean 8-carbon linear alkane, as it can also include heavier molecules with a small amount of branching.)
Raising the "octane rating" requires a more involved refining process, because the alkanes in the gasoline-diesel-kerosene spectrum all have very similar boiling points and are hard enough to distill cleanly when making "pure" gasoline. 8-carbon molecules are toeing the line, and heavier ones push pretty solidly into diesel territory. As an educated guess, refining high-octane gasoline probably cuts into diesel yields and reduces overall profits. It probably also requires a more expensive hydrocarbon cracking process, because branches are even less desirable in high-octane gasoline than they are in "pure" gasoline, and a lot of catalytic cracking processes are quite happy to spew out branched hydrocarbons.
Presumably, the reason "regular unleaded" gasoline is more expensive than "regular leaded" is the same reason that high-octane grades of gasoline are called "premium", and priced accordingly.
Hang on. This all sounds suspiciously familiar. Is he by any chance related to Fred Phelps?
Actually, Google's pretty big on dogfooding stuff. Google's internal e-mail is most definitely the Google Apps flavor of GMail. Most of the features that Google adds to the public site have been stress-tested by deploying them internally, where Google's tens of thousands of employees can use them, abuse them, and file bugreports.
If I had to take a wild guess, you hit an anti-malware heuristic. For a while, a lot of malware e-mails consisted of an encrypted .ZIP file, with the decryption password in the body. Apparently this confused a significant number of people into believing that the .ZIP file's contents "must be" legitimate, so they ran arbitrary executable code. This, of course, also hid the payload from anti-virus software. As a result, a lot of anti-malware filters now suspect any message with an encrypted .ZIP attachment.
Besides, you should never, ever rely on the encryption features built in to any .ZIP program. The traditional PKZip/WinZip encryption algorithms are nearly useless and straightforward to brute force. If you don't like public key encryption, use GPG with a symmetric key.
Oh, sure. All we need is for 10 companies to each separately dig their own trenches and lay their own fiber optic cable, at a duplicated cost of 10*$billions plus another order of magnitude in economic disruption. At that point, the Capitalism Fairy will magically turn an oligopoly of 10 large corporations into a healthy competitive market where no single player has the power to manipulate market prices.
And I have some prime beachfront real estate in Montana to sell you.
The reality is that phone lines, co-ax, and fiber optic cables share certain important traits with power lines, water mains, and sewage pipes. Namely: 1. The up-front installation costs are huge, which limits the number of players. (No entrepreneurs need apply, and therefore innovation is restricted.) 2. If each company needed to build separate infrastructure to each individual house, people would riot in the streets from all the closed roads, torn-up back yards, NIMBYist-angering overhead lines, and other offenses.
Situations like these, namely "utilities", are treated differently from the rest of the market. Why? Because they act differently from the rest of the market. They are natural monopolies: in the absence of government regulation, there tends to be only one provider for each of them. It is simply not possible for an unregulated market to have healthy competition in situations like these, and no amount of praying to Ayn Rand will make this not so.
The reality that the libertarian-utopia crowd doesn't want to acknowledge is that the best course of action with Internet access is probably to unprivatize the infrastructure, buying it back from the private local monopolists, then allowing free market competition to flourish among companies who provide Internet access on top of that public infrastructure. (In a sense, it ought to be public property already, since public tax dollars paid for huge chunks of it -- particularly the parts owned by the Bells. However, at this point it's a done deal, and for obvious reasons of fairness, contract law doesn't permit backsies.)
In a public-infrastructure/private-service scenario, bonus points are awarded if the Internet access companies running on top of that public infrastructure are permitted to bypass government inefficiencies and fix/upgrade the equipment themselves, but only if they are willing to pay for it themselves and with the understanding that their competitors will also reap the benefits. If the government does an inadequate job with maintenance, then in all likelihood this would quickly create a funding pool for infrastructure improvements that would, in some ways, be vaguely similar to the Free Software movement. (Do you think big software companies like Google hand out free code because they dislike money? No, they do it because it benefits the community, and thus benefits themselves. Even though open sourcing their code costs them in profit margins due to increased competition, the improvements (and testing! testing is expensive!) they get back from the community saves them even more money than they lose. It's the beginning of corporate ethics, the sorts of blindingly obvious insights that humans figured out thousands of years ago, but corporations are only just now learning. This is a good thing, and should be encouraged whenever possible.)
Disclaimer: as of recently, I now work for Google, but I don't speak for Google in any way, and I don't have anything to do with web search. I had noticed this and pondered it before I started my employment.
My educated guess is that they're doing it to count hits, presumably so they can (a) collect stats about how far people go down the list of search results, and more importantly (b) collect stats about how accurate the search results were: if the first link was clicked quickly, it must've been a good result; if there are long delays, lots of links clicked, and the first 3 results were never clicked, it must've been an awful result.
Again, this is without any insider insights into what the web search team is actually doing, mind you, so take what I say with a grain of salt.
Regarding bloat: it's not as bad from a bandwidth perspective as you make it out, but the added latency was very annoying for me when they first started doing that. Adding another round trip to Google before I can see my result does have an annoyance impact. There are a few Greasemonkey scripts out there that purport to undo the Javascript and purify the result links; however, I never got any of them to work for me.
Hopefully Google's web search team is working on something more asynchronous/AJAX-y that can report back in the background.
This is not correct. There *is* an attack that permits a person to create two or more new files that share the same MD5 hash. (That the resulting files have the same MD5 hash is a certainty, not a mere "high probability".)
However, there *is not* an attack that permits a person to create one new file that has the same MD5 hash as an existing file. Files with chosen MD5 hashes are still not a reality (for now).
That said, it's time to smile and nod at MD5 while making a slow, non-threatening retreat. Sadly, SHA-1 is having similar problems, SHA-2 is based on many of the same SHA-1 and MD5 principles, and it's not clear that any of the alternative hash algorithms is built on more solid ground.
Oh, well yeah, it's totally irrational to believe that innoculation can prevent disease. I mean, it only ERADICATED SMALLPOX.
Piffle, that. Every *day* I eradicate six impossible diseases before breakfast with carefully-formulated traditional medicine homeopathic organic vegan acupuncture shakes, down at the local Oxygen Bar. Beat that, innoculation!
I just ate a full maki roll of tuna sushi this week. That's more mercury than any one of the childhood shots I received back in the 80's (which still contained thiomersal back then).
You can't live without ingesting any of a hundred different poisons each day. That's why we have livers and kidneys. So long as the poisons don't overwhelm our bodies' abilities to eliminate and dispose of the poisons, we don't even notice.
What the fuck hyperbole train did you just ride in on?
The amount of ethyl mercury in a dose of vaccine is tiny, and ethylmercury is eliminated so quickly (half-life 18 days or less) that it does not bioaccumulate. You're putting your kid in more danger by feeding them a tuna sandwich once a week than you are by giving them the standard childhood vaccinations, even if you go back in time to 1998 before the US started phasing out thiomersal. Unlike ethylmercury, methylmercury does build up in the body (half-life 44 days), and methylmercury is found in tuna and other large, long-lived ocean fish. (It's also found in large, long-lived land mammals like humans, and babies receive noteworthy amounts of mercury through breast milk.)
The reality is that toxicity depends on dose. Oxygen is a deadly poison at a high enough concentrations: divers at 600m generally use breathing gas that's 98% He and 2% O2, because 21% O2 would kill them more-or-less instantly. Iron, an essential nutrient, is acutely toxic at a dose that's not much larger than a healthy amount: iron overdose is a leading cause of poisoning deaths in young children, and it used to be even worse thanks to the iron in Flintstone's chewables. (I myself had my stomach pumped when I was 4.) Methyl salicylate, better known as Ben-Gay and closely related to aspirin, killed a cross-country runner last year because she didn't know that it's poisonous in large doses.
On top of that, thiomersal has been phased out of the routine childhood vaccines for years now. There was no resulting drop in autism rates; there was no resulting drop in mercury poisonings; there was no resulting increase in cognitive function, or test scores, or any measurable thing whatsoever. All the available evidence shows that removing thiomersal did absolutely nothing.
On top of that, thanks in large part to the autism-vaccine controversy, mumps is making a comeback, and pertussis is now endemic in the area around Boulder, CO, thanks explicitly to unvaccinated children and a failure to reach herd immunity (which for pertussis is 92-94% vaccination).
I mean, hell, at least autism won't kill you.
But a ball of water the size of a basketball would rapidly boil off into space, so it's not in hydrostatic equilibrium, is it?
Hey, now! Mortal Kombat was actually fun if you didn't stare at it too hard, Resident Evil was tolerable for at least one viewing, and Street Fighter was... well, Street Fighter had MST3K comedy value. (Oh my God! They killed Raul Julia! You bastards!)
Having deliberately not watched a Boll movie from start to finish due to an utter lack of appeal for what I saw, I can't say for sure that they don't have any redeeming qualities, but they certainly seem at least one notch lower on the bad movie hierarchy than that.
I mean, sure, Paul Anderson has turned out some utter shite -- AvP comes to mind as one of his shiniest polished turds -- but he's never sunk to the Highlander-level depths of "Oh my God, now I have to invent a time machine just to retroactively kill myself and thereby never see this movie", or even to Vercingétorix levels of "Please remind me again why we're watching this".
Modern Intel motherboards (i810 forward) and AMD motherboards (768 forward) have a hardware RNG (Random Number Generator) that IIRC is based on diode noise. That's straight up quantum randomness, and most modern Linux distros automatically detect and use it if available.
I strongly recommend that you go read "What Colour are your bits?", which discusses this exact scenario.
Oh, come on. Header files -- at least, header files that implement APIs -- consist only of facts. A list of #defines and structs is a list of facts: it is not code, it is not a creative work, and it cannot be copyrighted. You cannot change the names of the struct members or the values of the #define constants while still following the API; the API restricts you to mechanically regurgitating the facts of how the API is defined.
SCO's big claim in the news media, back when they first announced their War On Linux, was that Linux was infringing on the Unix copyrights; their argument basically amounted to: "See, their header files are just like ours, once you strip out the comments". Which was utter BS, because the Linux headers implement POSIX and other standard APIs, and to implement POSIX the headers must contain exactly that code, word for word, because the POSIX API says so. If SCO had actually brought their BS to court, the judge would've laughed them out of the courtroom; and if Microsoft were to sue ReactOS over a silly header file, they'd face exactly the same reaction.
It's just as silly as saying that the NetBSD Editline library is violating the GPL because it implements the same API as the GNU Readline library. If you run both files through a C source code munger that eliminates comments and normalizes whitespace, they will be substantially identical — since they implement the same API, it's impossible for them to be broadly different, because there is no creativity (and therefore no copyright) in writing a header file that defines an API.
[Historical aside: Readline was placed under the GPL (instead of the LGPL) because GNU wanted to strongarm... er, "incentivize"... people into releasing their code under the GPL. This was because, if a closed-source program wanted to load and use Readline, it had to follow the Readline API — of which GNU Readline was the only implementation, which (the FSF argued) meant the closed-source program was really a derivative work of GNU Readline, and thus subject to the GPL. But it also applied to BSD-licensed open source code, so the NetBSD folks wrote their BSD-licensed Editline knockoff. Now that Editline exists, a program can use the Readline API while being clearly outside the scope of the GPL.]
Now, if you have the Windows source code repository available to you, and you can verify that the ReactOS implementation is substantially similar to the Windows implementation, then your argument has legs. Until then, stop with the FUD.
GCC already does that; that's my entire point. You can't just look at one benchmark on copying long strings and say "Well, there's obviously no situation where the REP instructions are faster", which is how I read your previous post.
Instead, they've got compile-time logic that spots fixed sizes (e.g. memcpy(dst, src, sizeof(struct foo));) and inlines the appropriate code for the target CPU, and then they've got more flexible logic for when the length is variable that does a few quick size comparisons at runtime. (They already need to check for word alignment anyway, so it's not outrageous overhead.)
I saw a previous Slashdotter claim that the CLD instruction can take ~50 cycles on a modern processor (presumably because the pipeline has to clear before %EFLAGS can change). If that's true, then it's very understandable that the GCC folks would be chomping at the bit to strip those CLD's out wherever they can, because then the REP loops are reasonable in many more scenarios.
But what about performance for shorter strings? Which has the greater setup overhead, REP or vector? And on which CPUs?
These are the sorts of questions that GCC developers have to code for, and they've already decided that the REP instructions are better in some situations.
RTFA. The reason the state is issuing these new fancy-schmancy thumb drives is that the new ones (claim to) have 256-bit AES encryption and (claim to) self-destruct after 10 consecutive wrong passwords. They're doing this whole switch because of privacy, because the thumb drives contain the private, personal case files of hundreds/thousands of citizens.
I didn't forget that. But we've observed gravitational lensing. Gravitational lensing is an observational fact. No other theory before General Relativity had predicted gravitational lensing, and what's more, the amount of gravitational lensing we see is precisely consistent with General Relativity. That means, to the best of our ability to detect, General Relativity describes reality itself, and any new theory of gravity must describe reality at least equally well -- just as General Relativity would have been rejected outright if it were inconsistent with the facts known under Newtonian gravity.
I don't think it's possible to overestimate how difficult it is to pick some Joe Random's Theory of Gravity and have that theory predict that light itself, which has no mass, is somehow being attracted to things that do have mass. (Or, similarly bizarrely, that light has no mass yet does have momentum.) The String Theory family was carefully formulated such that it's guaranteed to reduce to General Relativity under the correct conditions. The TeVeS version of MOND might work. But, no matter how you slice it, just fiddling with the knobs of Newtonian gravity won't result in a theory compatible with Lorentz invariance and the Maxwell equations, unless you set out to do that from the start like Einstein did or the String Theory folks did, or perhaps unless you're very smart and very patient like the MOND folks are trying to do with TeVeS.
GR says that lensing of this sort is always caused by matter, and more matter causes stronger lensing (related by an equation). We don't see any matter there. Therefore, if there is matter, that matter neither emits nor appreciably absorbs light (i.e. "dark matter"), and GR says there must be matter where there is lensing. What it would take to unseat GR in the dark matter debate is a theory that simultaneously explains both (a) lensing by actual matter, and (b) lensing around empty patches of space that somehow lens light while containing no observed matter. It also needs to explain why nearly all galaxies spin at the same speed at the edges as they do near the center, while NGC 4736 spins slower at the edges than it does near the center.
In the past, people ran around saying that the scientists who proposed Quantum Mechanics were just "fooling themselves", running around believing that the model was reality. Well, the problem with that is that the model is DAMN good at predicting reality. The computer that you use to read Slashdot is proof enough of that -- semiconductor physics depends on a rather large amount of quantum mechanics, from the P-N junction downward. Intel and AMD exist, your computer exists, therefore QM is "true". Gravitational lensing exists, there is no observable matter there, therefore Dark Matter is "true". The observations stack up to the point that any new theory could only be a refinement of the old theory, like Einstein's was to Newton's.
Back when we were just monkeying around with redshifts and galactic rotation curves, MOND was a very reasonable model. But the Bullet Cluster, NGC 4736, and now this direct observation of dark matter filaments is entirely incompatible with MOND. MOND supporters readily acknowledge this -- MOND can't explain the Bullet Cluster without claiming that force can act in a different direction than acceleration (F=ma being a vector equation), which no MOND supporter is willing to countenance. (Last I heard, the MOND proponents were hoping that warm neutrinos plus MOND might be enough, though. Neutrinos are dark matter, but fairly mundane dark matter.) At this point, after the last 2 years of shocking observations, there IS no other theory left standing that fits th
Because Lisp wasn't so much invented as discovered. It was a mathematical formulation, with a toy syntax, meant for pure mathematical work, that was never meant to be implemented as a real programming language. Then a grad student had the bright idea to write eval in machine code... thus creating the first interpreted language. Ever.
Those same mathematical principles apply to any programming language, and thus each new language is just slightly more Lisp-like than the one it replaces. Java is C++ with garbage collection (invented in Lisp). Perl added closures (invented in Lisp) to many peoples' everyday vocabularies -- every time you type "map { ... }" or "sort { ... }" in Perl, that's Lisp peeking through. Perl 6 is going the extra step of adding Lisp-style macros and a user-modifiable grammar, thus making it (in the mathematical sense) a dialect of Lisp, albeit one with a funny syntax.
Go read Paul Graham's What Made Lisp Different, then claim that it's just a coincidence that every new programming language borrows more and more ideas from something invented in the 1950s. Then do yourself a favor and buy Higher-Order Perl.
That's the thing: no, it wouldn't!
Have you even read the Wikipedia article on MOND? MOND is the only serious proposal to modify our laws of gravity. It proposes a change in F=ma, such that tiny accelerations result in a surprisingly large force. This would not affect the amount of mass, and thus would not cause a change in gravitational lensing. Any observation of gravitational lensing in a place where there's no visible matter is ipso facto evidence in favor of dark matter and evidence in disfavor of MOND.
Beyond MOND, no other modification of gravity has yet been proposed that's consistent with General Relativity. (Even MOND might not be, depending on how well TeVeS works. AFAIK, the math is still under scrutiny.) And without compatibility with General Relativity, gravitational lensing is impossible, so it contradicts reality.
At this point, thanks to the Bullet Cluster and other observations, even MOND supporters acknowledge that dark matter exists at the scale of galactic clusters and up. The only question remaining is, at the scale of a single galaxy, whether the galactic rotation curve is due to MOND or due to a halo of dark matter around the galaxy. And that itself is a point of fresh debate ever since the recent observations of NGC 4736, a galaxy that follows the classic (non-MOND) Newtonian predictions. Dark matter theory can easily account for this by suggesting that this one galaxy was somehow separated from the dark matter halo that once surrounded it; MOND (or any other modification to gravity) must somehow explain why this one galaxy does not conform to the laws of physics.
At this point, the existence of dark matter is as good as proven.
What's up with dark energy, OTOH, is still anyone's guess...
I call bullshit. It's not scientists putting this crap out there. It's bad science reporting. If it's not a scary disease or a global disaster, then it gets about 3 words (not in a row) of actual scientific content, surrounded by fuzzy blather and bad analogies. (If it is a scary disease or a global disaster, then it gets 6 scientifically meaningful words, 4 of them wrong.)
I'm reminded of an anecdote (attribution lost, possibly apocryphal) of a politician being briefed on the progress of the SDI/Star Wars project. The scientist is describing the current technology, saying that after years of effort they can fire a laser that delivers 10^n W (for some n that I forget) to a target. However, for the laser to be useful, they need a laser than can deliver 10^2n W. The politician, not understanding exponentials, remarks: "My God, we're halfway there!"
Warning: rambling post ahead.
My gut feeling is that, from strictly a hardware perspective, we're already capable of building a human-level AI. The problem is that, from a software perspective, we've focused too much on approaches that will never work.
As far as I'm concerned, the #1 problem is the Big Damn Database approach, which is basically a cargo cult in disguise. Though expert systems are useful in their niches, "1. Expert system 2. ??? 3. AI!" is not a workable roadmap to the future. I'm certain that it's far easier to start with an ignorant AI and teach it a pile of facts than it is to start with a pile of facts and teach it to develop a personality.
The #2 problem is the Down To The Synapse approach. This, unlike BDD, could quite possibly create "A"I if given enough hardware. But I think that, while DTTS will lead to a better understanding of medicine, it won't advance the AI field. It won't lead to an improved understanding of how human cognition works — it certainly won't teach us anything we didn't already know from Phineas Gage and company.
Even if we go to all the trouble of developing a supercomputer capable of DTTS emulation of a human brain — so what? If we ask this emulated AI to compute 2+2, millions of simulated synapses will fire, trillions of transistors will flip states, phenomenal amounts of electricity will pour into the supercomputer, just for the AI to give the very same answer that a simple circuit consisting of a few dozen transistors could've answered in a tiny fraction of the time, using the amount of electricity stored on your fingertip when you rub your shoes on the carpet during winter. And that's not even a Strong AI question. That's not to say that working DTTS won't be profound in some sense, but we know we can build it better, yet we won't have the faintest idea of where to go next.
That brings me to my core idea — goals first, emotions close behind. Anyone who's pondered the "is/ought" problem in philosophy already knows the truth of this, even if they don't know they know the truth of it. The people building cockroach robots were on the right track all along; they're just thinking too small. MIT's Kismet, for instance, gives an idea of where AI needs to head.
That said, I think building a full-on robot like Kismet is premature. A robot requires an enormous number of systems to process sensory data, and those processing systems are largely peripheral to the core idea of AI. If we had an AI already, we could put the AI in the robot, try a few things, and ask the AI what works best. So, ideally, I think we need to look at a pure software approach to AI before we go off building robot bodies for them to inhabit.
And how to do that? I think Electric Funstuff's Sim-hilarities captures the essence of that. If we give AIs a virtual world to live in — say, an MMO — then that removes a lot of the need for divining meaning from sensory input, allowing a sharper focus on the "intelligence" aspect of AI. Start with that, grow from there, and I can definitely see human-level AI by 2029.
But that would put Cheney in charge.