Wish I could find the article; it said it so much better.
Ah hah! From an essay at the Cato Institute by Rasmus Fleischer (co-founder of The Piracy Bureau, parent organization of BitTorrent tracker The Pirate Bay):
Copyright law is mutating into something qualitatively different than what it has been in previous centuries...
Gray zones like these are omnipresent in 21st century copyright law. One reason for this development is the uncertain status of the very idea of "copying" today. Contrast today's world with the golden age of copyright, roughly speaking between 1800 and 1950. Back then, enforcement was easy. The act of reading a book was far removed from the act of printing one. Record presses and gramophones were safely distinct machines. Since then, things have changed.
The problem is that in most cases nobody knows how valuable the work is going to be. Therefore, it is hard to price the work until you see how many people will pay to read it.
I saw a brilliant argument last year, to the effect of: The copyright system (or even the entire IP system) is an imperfect approximation of "compensate creators for the value of their content". There's nothing special about "copying", per se; if nobody wanted to read your book, then nobody would want to copy it either.
But copying, until recently, was a useful proxy for "obtaining value". It's like a toll road. You're not paying for the value of "pass through this gated entrance"; you're paying for the entire value of the road. The entrance is just a convenient place to meter it.
Plain-paper Xerography changed the model; now individuals could copy works without much effort. And the digital world completely overturned it; works are copied in the course of using them, and there's no realistic way to prevent "unauthorized" copying. Which, really, is fine, because again: Copyright is only about copying because copying was a useful proxy.
What it really means: "We need a new way to compensate creators".
Wish I could find the article; it said it so much better.
With only a few gTLDs, you're right - there's no obvious difference between.com and.net. So everyone with a trademark wants to register their trademark in every gTLD, which only reinforces the lack of distinction between them. And so consumers don't really understand what a gTLD is; they think ".com" is part of the "noise" of the URL, like "http://www.". Which, again, becomes a self-fulfilling proposition.
I don't know if the right number of gTLDs is hundreds or thousands, but the right number is "more than most trademark holders are willing to pay for". In trademark law itself, there are categories of trademarks, and most don't rise to the level of what IIRC are called "famous marks" - trademarks that are protected no matter how they're used. (You could maybe sell a stuffed toy and call it the Caterpillar, but you couldn't call it the Google.)
We need to get gTLDs to that point, where most second-level domains (cisco.com) are unique only within the gTLD (.com), or a group of gTLDs (.com,.network,.wifi) and nobody cares about other gTLDs - because consumers wouldn't expect to see the trademark in that context (cisco.rap-music? No, you must be looking for sisqo.rap-music). It's self-reinforcing in both directions.
Think back to phone numbers, before 10-digit dialing. Companies tried to pick memorable phone numbers. Some companies that operated in multiple area codes reserved the same number in each area code - so you could always call 736-5000 and get a tow truck anywhere in New York, or whatever. Some brands were so strongly associated with their phone number that they were set up in EVERY area code - 936-3636 for dial-a-santa.
With ~200 area codes, and only a handful serving any one region, this was feasible. Once you went to a thousand area codes, with overlays and cell-phones confusing the mix, and ten-digit dialing, it became pointless. Now, the only "area-code-free" numbers are the x11 numbers.
I think it's more likely that they are trying to inflate their own numbers and don't have the power they pretend or are wording it in such a way that it seems they can do more than they can.
I was thinking the same thing... If Yelp's search results change frequently, and if reviews are regularly removed (for any reason at all), couldn't this be a combination of two classic scams, plus confirmation bias?
1. The Perfect Prediction scam: You send postcards to 10,000 people, predicting the winner of tomorrow night's game. On half the postcards, you write the home team's name; on the other half, the visiting team's name. You keep track of who got which postcard. The day before the next game, you send 5,000 postcards - to only the people who got the "accurate" postcard. Repeat a few times, and you'll have a few hundred people who think you have the inside scoop on fixed games, and who will pay you for the next "prediction".
2. The White Van Speakers scam: "Psst! Hey, wanna buy some speakers? I, uh, these were left over after we did a high-end home theatre job. Yeah, and the boss said throw them out, but I figured someone could use them, you know? I'll let 'em go for only $500." When you get them home, you discover they do work, but they're horrible. Who are you going to complain to? You bought what you thought were stolen speakers out of the back of a van.
It seems to me that even if Yelp isn't *actually* gaming their reviews, there might be a sales team that's discovered it's in their best interest to *claim* they are. The recipe:
1. Wait till the next "Google dance" on Yelp. 2. Check out the newly-sorted review page for your potential client. 3. If they now have good reviews at the top, claim you did it, just to show what you can do. 4. If they have bad reviews, claim you can fix them. 4a. And if they turned you down last month, hint vaguely that this is the result.
Naturally, when the next Yelp dance occurs, that client will have a newly randomized set of results - and they may not be better than the current ones. But who are they going to complain to? "Hey, I paid to have negative reviews removed, and they're still up there."
How irksome that judges, juries and lawyers should have to learn how technology works in order to do their jobs properly.
Aren't you then complaining how irksome it is that techies should have to learn how the law - and people - work?
What if we define "the law" as "a system of written laws, human actors, biases and tendencies which produces largely desirable but occasionally flawed results"?
That's how I read TFA. "Technically, you may have room to reasonably doubt one piece of technical evidence in isolation. But you probably won't reach reasonable doubt when all the evidence is considered collectively by a non-technical judge and jury. And even if you do, your life is still hell."
If you say "but they're technically WRONG!", I answer: Yes, but you should have RTFM.
The fact that confidence factors travel with the data is an important part of fuzzy logic.
Fascinating... thanks for the background. I guess the popular-science books focused more on the end result than the process - which itself ignores fuzzy logic:)
This sounds a lot like Flying Logic, my new favorite mind-mapping tool. It lets you draw a directed graph of your argument/evidence/etc, with confidence factors, edge weights, and operators. Then you can play with the confidence spinners and see how the (likely) outcome changes.
That was my first reaction too - and, by odd coincidence, that just popped into my head last night. "Hey, whatever happened to fuzzy logic?"
From what I remember, the practical applications all boiled down to one technique: lookup tables. Washing machines would eliminate the need to set temperature, spin speed, etc. separately. You'd tell it what fabric and what type of soil, and it would use fuzzy logic to determine the best laundry parameters. AFAICT, that's just a lookup table - with maybe some precursor to Bayesian stuff. There was even a slogan in Japan: "It's fuzzy!" (Yes, we know.)
But as TFA and some commenters pointed out: this could be great for streaming media, or anything else that doesn't need a precise answer. With things like OpenGL, what if you kicked off a few threads that explicitly didn't need precision? You'd have a fuzzy coprocessor, just like graphics and math and crypto coprocessors.
I wonder: are the failure modes predictable enough that you could wrap calculations with middleware, which would repeat them enough times to reach your desired level of accuracy (thus losing only precision, not accuracy)? If the failures are randomly distributed, and you can overclock things 7x, you could run everything 3x and still come out ahead.
Fascinating... so "teaching" doesn't work, but shaming does.
That's disappointing, though, since presumably more sites would be amenable to teaching than shaming.
I had proposed just such a phishing lesson for AOL customers back in the day, but of course marketing didn't like the idea of telling users "Don't trust us". And without any data, I couldn't particularly argue to the contrary.
Pretty sure they wouldn't have gone for the shaming either. Ah well.
Re:Ruby/Python in JVM/CLR not a Silver Bullet
on
Ruby 1.9.1 Released
·
· Score: 1
I'm no compiler/interpreter expert, but AIUI the JRuby folks (some of whom are now with Sun) are actively working with the JVM folks inside Sun to add dynamic-language optimizations to the JVM.
I don't know if this extends to "fixing the algorithm", but it seems like it could. It's early days.
The real issue is that all light bulbs really do need to have the rating of lumens.
Isn't there a problem due to the way that lumens are either (a) correctly measured, or (b) incorrectly converted from candela (I forget which)? Something about how incandescent bulbs (sans reflectors) give off light in 360 degrees, while LEDs are highly directional. So if you measure the amount of light hitting a point in the "right" place, the LED would appear to be just as bright as the bulb - but, of course, the total light output is far less. And cd or lumens (or both) are measured from a single point.
I remember seeing a rant about this five or six years ago, when Nichia first came out with brighter white LEDs. I sent a copy of that rant to a dealer claiming "equivalent to incandescent" LED fixtures; the dealer's response was something like "Well, but look - the numbers are identical!" which missed the point.
Has that now been solved with reflector design, and/or prohibitions against misleading conversions?
You don't need to certify anything. Companies with unscrupulous (or ascrupulous) business practices don't have isolated incidents; their profit motive inches them closer and closer to the entire underworld of sleaze. Their visible links to fraud are an excellent proxy for their hidden ones. All you have to do is make it searchable.
I'm having difficulty imagining a person who would watch an inauguration but not own a television.
It's Saturday night on Slashdot. Anyone reading this will be the type of person who owns a television AND would watch an inauguration, but who has difficulty imagining a person.
Does that mean they'll have to figure out ways to reduce the noise generated by arc-fault interruptors? I know that residential AC AFCI breakers have an obnoxious acoustic hum, but I don't know if that'd translate into powerline noise for DC versions..
I completely disagree. Maybe it's the backlight on my LCD, but I noticed a certain warmth on IPv4 that's distinctly lacking from IPv6. True, the soundstage is more distinct, but the highs are brittle. That said, I've found that setting the "unused" bits to 1 instead of 0 - in the analog domain, of course - really helps bring out the vocals.
It hasn't been this way forever, though. As AOL grew, peaks were our nightmare, and I'm pretty sure they happened during the week - I think 8-9pm Wednesdays (the West Coast was getting home from work, and the East Coast was still online).
More interesting was the big dip on Thursdays from 8pm-11pm. That'd be NBC's Must-See Thursdays: Friends, Seinfeld, ER, etc. I bet that doesn't happen anymore.
svn has one repository, and it's always central. In git, you have any number of repositories, and whether you call one "central" is an administrative question, not a technical one.
It solves the problem of "I don't want to check this into svn yet, but I want to save a checkpoint"; developers can have local repositories for coding, and you'll have one (or more) for integration and release. You can cherry-pick changes, merging any individual change at any time (or never).
And it's ridiculously fast. I never thought svn was that slow, but turns out it's dog-slow. If you remember being floored by the Amiga's blitter graphics: that sort of fast.
in the mid-90s by some of the cognitive researchers at the center of the whole anthropomorphic-character thing at Microsoft. I'm not having any luck finding it online, but the gist was: People react to computers the way they react to people.
It was filled with fascinating experiments. They'd have people work with one of two "expert system" programs - one of which subtly complimented the user's knowledge, and one of which didn't. Invariably, the friendly system would be rated as "more accurate".
Or: People watched a news program on one of two TV sets. One was unlabeled; the other was labeled "news" (and, IIRC, in the presence of other such "special purpose" TV sets labelled entertainment, sports, etc.) Those who watched on the special-purpose News TV were more likely to call the program "authoritative".
If you'd read that book, it was almost impossible NOT to come to the conclusion that things like Bob, Clippy, etc. would make average consumers far more comfortable - and productive - with computers. It was blindingly obvious from the studies, and, like so many wrong ideas, it seemed perfectly reasonable, and jibed with evolutionary psychology, etc. etc.
A few years ago, someone in the field told me that the truth eventually came out: The researchers had either fudged or misinterpreted their data. Oops.
I always see that from the writer's viewpoint, as if he's saying "Look, I know this isn't news, and I'm just getting around to writing about it a few years later, but I really do have something interesting to say about it! So I will acknowledge its apparent staleness with a jokey aside before I get to the point."
Good thing writing isn't some sort of Rorschach test where we can each imbue it with our own insecurities, eh?
Slashdot Mirror
Ah hah! From an essay at the Cato Institute by Rasmus Fleischer (co-founder of The Piracy Bureau, parent organization of BitTorrent tracker The Pirate Bay):
The Future Of Copyright
I saw a brilliant argument last year, to the effect of: The copyright system (or even the entire IP system) is an imperfect approximation of "compensate creators for the value of their content". There's nothing special about "copying", per se; if nobody wanted to read your book, then nobody would want to copy it either.
But copying, until recently, was a useful proxy for "obtaining value". It's like a toll road. You're not paying for the value of "pass through this gated entrance"; you're paying for the entire value of the road. The entrance is just a convenient place to meter it.
Plain-paper Xerography changed the model; now individuals could copy works without much effort. And the digital world completely overturned it; works are copied in the course of using them, and there's no realistic way to prevent "unauthorized" copying. Which, really, is fine, because again: Copyright is only about copying because copying was a useful proxy.
What it really means: "We need a new way to compensate creators".
Wish I could find the article; it said it so much better.
Why is that any more likely to work than training people not to be nasty?
With only a few gTLDs, you're right - there's no obvious difference between .com and .net. So everyone with a trademark wants to register their trademark in every gTLD, which only reinforces the lack of distinction between them. And so consumers don't really understand what a gTLD is; they think ".com" is part of the "noise" of the URL, like "http://www.". Which, again, becomes a self-fulfilling proposition.
I don't know if the right number of gTLDs is hundreds or thousands, but the right number is "more than most trademark holders are willing to pay for". In trademark law itself, there are categories of trademarks, and most don't rise to the level of what IIRC are called "famous marks" - trademarks that are protected no matter how they're used. (You could maybe sell a stuffed toy and call it the Caterpillar, but you couldn't call it the Google.)
We need to get gTLDs to that point, where most second-level domains (cisco.com) are unique only within the gTLD (.com), or a group of gTLDs (.com, .network, .wifi) and nobody cares about other gTLDs - because consumers wouldn't expect to see the trademark in that context (cisco.rap-music? No, you must be looking for sisqo.rap-music). It's self-reinforcing in both directions.
Think back to phone numbers, before 10-digit dialing. Companies tried to pick memorable phone numbers. Some companies that operated in multiple area codes reserved the same number in each area code - so you could always call 736-5000 and get a tow truck anywhere in New York, or whatever. Some brands were so strongly associated with their phone number that they were set up in EVERY area code - 936-3636 for dial-a-santa.
With ~200 area codes, and only a handful serving any one region, this was feasible. Once you went to a thousand area codes, with overlays and cell-phones confusing the mix, and ten-digit dialing, it became pointless. Now, the only "area-code-free" numbers are the x11 numbers.
gTLDs are like area codes.
I was thinking the same thing... If Yelp's search results change frequently, and if reviews are regularly removed (for any reason at all), couldn't this be a combination of two classic scams, plus confirmation bias?
1. The Perfect Prediction scam: You send postcards to 10,000 people, predicting the winner of tomorrow night's game. On half the postcards, you write the home team's name; on the other half, the visiting team's name. You keep track of who got which postcard. The day before the next game, you send 5,000 postcards - to only the people who got the "accurate" postcard. Repeat a few times, and you'll have a few hundred people who think you have the inside scoop on fixed games, and who will pay you for the next "prediction".
2. The White Van Speakers scam: "Psst! Hey, wanna buy some speakers? I, uh, these were left over after we did a high-end home theatre job. Yeah, and the boss said throw them out, but I figured someone could use them, you know? I'll let 'em go for only $500." When you get them home, you discover they do work, but they're horrible. Who are you going to complain to? You bought what you thought were stolen speakers out of the back of a van.
It seems to me that even if Yelp isn't *actually* gaming their reviews, there might be a sales team that's discovered it's in their best interest to *claim* they are. The recipe:
1. Wait till the next "Google dance" on Yelp.
2. Check out the newly-sorted review page for your potential client.
3. If they now have good reviews at the top, claim you did it, just to show what you can do.
4. If they have bad reviews, claim you can fix them.
4a. And if they turned you down last month, hint vaguely that this is the result.
Naturally, when the next Yelp dance occurs, that client will have a newly randomized set of results - and they may not be better than the current ones. But who are they going to complain to? "Hey, I paid to have negative reviews removed, and they're still up there."
"So... do you have kids?"
Aren't you then complaining how irksome it is that techies should have to learn how the law - and people - work?
What if we define "the law" as "a system of written laws, human actors, biases and tendencies which produces largely desirable but occasionally flawed results"?
That's how I read TFA. "Technically, you may have room to reasonably doubt one piece of technical evidence in isolation. But you probably won't reach reasonable doubt when all the evidence is considered collectively by a non-technical judge and jury. And even if you do, your life is still hell."
If you say "but they're technically WRONG!", I answer: Yes, but you should have RTFM.
Hand's on the other foot now, ain't it?
Fascinating... thanks for the background. I guess the popular-science books focused more on the end result than the process - which itself ignores fuzzy logic :)
This sounds a lot like Flying Logic, my new favorite mind-mapping tool. It lets you draw a directed graph of your argument/evidence/etc, with confidence factors, edge weights, and operators. Then you can play with the confidence spinners and see how the (likely) outcome changes.
That was my first reaction too - and, by odd coincidence, that just popped into my head last night. "Hey, whatever happened to fuzzy logic?"
From what I remember, the practical applications all boiled down to one technique: lookup tables. Washing machines would eliminate the need to set temperature, spin speed, etc. separately. You'd tell it what fabric and what type of soil, and it would use fuzzy logic to determine the best laundry parameters. AFAICT, that's just a lookup table - with maybe some precursor to Bayesian stuff. There was even a slogan in Japan: "It's fuzzy!" (Yes, we know.)
But as TFA and some commenters pointed out: this could be great for streaming media, or anything else that doesn't need a precise answer. With things like OpenGL, what if you kicked off a few threads that explicitly didn't need precision? You'd have a fuzzy coprocessor, just like graphics and math and crypto coprocessors.
I wonder: are the failure modes predictable enough that you could wrap calculations with middleware, which would repeat them enough times to reach your desired level of accuracy (thus losing only precision, not accuracy)? If the failures are randomly distributed, and you can overclock things 7x, you could run everything 3x and still come out ahead.
Don't be ridiculous. They use sugar pills.
(Or Folger's crystals!)
Fascinating... so "teaching" doesn't work, but shaming does.
That's disappointing, though, since presumably more sites would be amenable to teaching than shaming.
I had proposed just such a phishing lesson for AOL customers back in the day, but of course marketing didn't like the idea of telling users "Don't trust us". And without any data, I couldn't particularly argue to the contrary.
Pretty sure they wouldn't have gone for the shaming either. Ah well.
I'm no compiler/interpreter expert, but AIUI the JRuby folks (some of whom are now with Sun) are actively working with the JVM folks inside Sun to add dynamic-language optimizations to the JVM.
I don't know if this extends to "fixing the algorithm", but it seems like it could. It's early days.
I just want to acknowledge and validate your Arch Deluxe reference.
Isn't there a problem due to the way that lumens are either (a) correctly measured, or (b) incorrectly converted from candela (I forget which)? Something about how incandescent bulbs (sans reflectors) give off light in 360 degrees, while LEDs are highly directional. So if you measure the amount of light hitting a point in the "right" place, the LED would appear to be just as bright as the bulb - but, of course, the total light output is far less. And cd or lumens (or both) are measured from a single point.
I remember seeing a rant about this five or six years ago, when Nichia first came out with brighter white LEDs. I sent a copy of that rant to a dealer claiming "equivalent to incandescent" LED fixtures; the dealer's response was something like "Well, but look - the numbers are identical!" which missed the point.
Has that now been solved with reflector design, and/or prohibitions against misleading conversions?
I'd noticed that all my DNS queries for non-existent domains were actually returning NXDOMAIN, instead of an advertisement...
You don't need to certify anything. Companies with unscrupulous (or ascrupulous) business practices don't have isolated incidents; their profit motive inches them closer and closer to the entire underworld of sleaze. Their visible links to fraud are an excellent proxy for their hidden ones. All you have to do is make it searchable.
Working on that.
It's Saturday night on Slashdot. Anyone reading this will be the type of person who owns a television AND would watch an inauguration, but who has difficulty imagining a person.
Does that mean they'll have to figure out ways to reduce the noise generated by arc-fault interruptors? I know that residential AC AFCI breakers have an obnoxious acoustic hum, but I don't know if that'd translate into powerline noise for DC versions..
I completely disagree. Maybe it's the backlight on my LCD, but I noticed a certain warmth on IPv4 that's distinctly lacking from IPv6. True, the soundstage is more distinct, but the highs are brittle. That said, I've found that setting the "unused" bits to 1 instead of 0 - in the analog domain, of course - really helps bring out the vocals.
It hasn't been this way forever, though. As AOL grew, peaks were our nightmare, and I'm pretty sure they happened during the week - I think 8-9pm Wednesdays (the West Coast was getting home from work, and the East Coast was still online).
More interesting was the big dip on Thursdays from 8pm-11pm. That'd be NBC's Must-See Thursdays: Friends, Seinfeld, ER, etc. I bet that doesn't happen anymore.
svn has one repository, and it's always central. In git, you have any number of repositories, and whether you call one "central" is an administrative question, not a technical one.
It solves the problem of "I don't want to check this into svn yet, but I want to save a checkpoint"; developers can have local repositories for coding, and you'll have one (or more) for integration and release. You can cherry-pick changes, merging any individual change at any time (or never).
And it's ridiculously fast. I never thought svn was that slow, but turns out it's dog-slow. If you remember being floored by the Amiga's blitter graphics: that sort of fast.
Yeah, but it's the white guys that'd get pulled off the plane for calling "shotgun".
in the mid-90s by some of the cognitive researchers at the center of the whole anthropomorphic-character thing at Microsoft. I'm not having any luck finding it online, but the gist was: People react to computers the way they react to people.
It was filled with fascinating experiments. They'd have people work with one of two "expert system" programs - one of which subtly complimented the user's knowledge, and one of which didn't. Invariably, the friendly system would be rated as "more accurate".
Or: People watched a news program on one of two TV sets. One was unlabeled; the other was labeled "news" (and, IIRC, in the presence of other such "special purpose" TV sets labelled entertainment, sports, etc.) Those who watched on the special-purpose News TV were more likely to call the program "authoritative".
If you'd read that book, it was almost impossible NOT to come to the conclusion that things like Bob, Clippy, etc. would make average consumers far more comfortable - and productive - with computers. It was blindingly obvious from the studies, and, like so many wrong ideas, it seemed perfectly reasonable, and jibed with evolutionary psychology, etc. etc.
A few years ago, someone in the field told me that the truth eventually came out: The researchers had either fudged or misinterpreted their data. Oops.
Yes, and it's already been invented a few times:
I always see that from the writer's viewpoint, as if he's saying "Look, I know this isn't news, and I'm just getting around to writing about it a few years later, but I really do have something interesting to say about it! So I will acknowledge its apparent staleness with a jokey aside before I get to the point."
Good thing writing isn't some sort of Rorschach test where we can each imbue it with our own insecurities, eh?