It ain't over till it's over. The universe will still be inhabitable for a long while, and perhaps we can reverse entropy or escape the universe some millions of years in the future. Who knows?:)
Suppose blacks are more likely to live in the city, and whites are more likely to live in suburbia (just for the sake of the argument). Suppose furthermore that blacks are stopped more frequently, so they're more often arrested even if the baseline crime rate is the same (again, for the sake of the argument). Then even if you hide race and gender info from the AI, it might just say "people living in the city are more likely to offend than people living in suburbia". All this because being black is correlated with living in the city, and being black is also correlated with offending according to the data, due to preexisting selection effects.
How about you let the users opt in, and if you need a more representative sample, you actively get a focus group. It costs more, but it's not like MS is broke or anything.
There's this country in Asia that has considerable experience with a government-controlled internet. Perhaps the Conservatives could arrange a trip there to learn?
I was asking how it could tell different lynxes apart, or different w3ms apart, the way it could tell e.g. Firefox on Linux and Firefox on Windows apart. It doesn't seem possible unless the headers or the response times differ.
If headers are all that make different text browsers look different, perhaps the developers could talk to each other to make their browser more like one another, to thwart just this kind of privacy invasion.
Try to write any sort of backtracking solver (e.g. integer or constraint programming, Sudoku, alpha-beta pruning, MCTS game AI) without recursion. Or even just a password brute-forcer. Listing every password of length n is a pain without recursion; you either get a loop n deep or you end up implementing your own stack.
What these problems have in common is that they grow so fast that the computer doesn't really have time to reach stack overflow. Consider chess, for instance. No chess AI can go 100 turns deep, but pretty much every stack can handle 100 entries.
That's interesting. One problem I've happened upon when playing with adversarial learning like that (though in a genetic algorithm context) is that the programs forget what has happened before; e.g. in a rock-paper-scissors setting, the first system learns rock, then the second learns paper, then the first learns scissors, then the second learns rock and you're back where you started. Presumably they have some way of avoiding this with GANs.
If it's possible to do by a robot arm, it should be possible to do by faking the input from the stylus system. All you'd need is something like a finite element model of the physical system involving the robot and stylus (in the very worst case).
I suspect there's a selection effect: the kind of people who get filthy rich are the kind who are extremely motivated by money. So they wouldn't stop working. If they see money as score and the highest score wins, then they keep working even after they're rich. On the other hand, the kind of people who would quit after $10M may not have the ruthless drive to get there no matter what it takes.
1. Make a program that asks the CPU if it's malware.
2. Have the program do malwary stuff if the CPU says it's not malware, and do benign stuff otherwise.
3. Profit! (Or laugh.)
There are two ways of constructing a software design: One way is to make it so simple that there are obviously no deficiencies, and the other way is to make it so complicated that there are no obvious deficiencies.
Since DREs are usually general-purpose computers running Windows or something, no points for guessing which category they're part of. And the only way you can have a DRE is to make it as simple as possible, which means no Turing-complete hardware. Just a row of buttons, a matrix display, and counters. Or you know, paper and pencil.
Kurzweil (on the one hand) and all those people thinking the terminator scenario will happen (on the other) feel like theists who try to quash their theism by force, but it just pops up elsewhere in another shape.
Kurzweil's "we're all going to be immortal and the singularity will bring plenty to all" is: technology will let us make God and we will all go to techno-Heaven.
The terminator/golem scenario with the out-of-control superintelligences turning the whole world into computer material is: technology will let us make God and we will all go to techno-Hell. The more you get into the really bizarre end of the theology, the more obvious it is.
I can think of something more useful. See if the site computer has a modem or phone connection or something, and then bridge over onto the Kwangmyong intranet. Port scan and download everything!... although at dialup speeds, it would take a while.
Statistically speaking, the variance decreases with the sample size. That is, if you have a legislature of 500, the chance of getting a majority of say, people who'd like to turn the country into 1984 is pretty close to 0%. But if you select the president by lottery, you may be very unlucky and get a Big Brother. In other words, yes, it is too risky.
I agree with you that "a king by another name" isn't really a good way to do things. A parliamentary sortition system would solve this without any problem, because the parliament would work as an electoral college. But you're not going to change the US into a parliamentary form of government. Hence an explicit electoral college instead of an implicit one.
Maybe it is a little silly to compare the chances of getting sortition with the chances of becoming more parliamentary. But I think the latter is a lot less likely than the former, for some reason... at least until the former has run for a while. Perhaps the system would decide to reorganize in that manner after a while.
Slashdot Mirror
It ain't over till it's over. The universe will still be inhabitable for a long while, and perhaps we can reverse entropy or escape the universe some millions of years in the future. Who knows? :)
Welcome to panexperientialism.
Not so easy in practice.
Suppose blacks are more likely to live in the city, and whites are more likely to live in suburbia (just for the sake of the argument). Suppose furthermore that blacks are stopped more frequently, so they're more often arrested even if the baseline crime rate is the same (again, for the sake of the argument). Then even if you hide race and gender info from the AI, it might just say "people living in the city are more likely to offend than people living in suburbia". All this because being black is correlated with living in the city, and being black is also correlated with offending according to the data, due to preexisting selection effects.
How about you let the users opt in, and if you need a more representative sample, you actively get a focus group. It costs more, but it's not like MS is broke or anything.
There's this country in Asia that has considerable experience with a government-controlled internet. Perhaps the Conservatives could arrange a trip there to learn?
I was asking how it could tell different lynxes apart, or different w3ms apart, the way it could tell e.g. Firefox on Linux and Firefox on Windows apart. It doesn't seem possible unless the headers or the response times differ.
If headers are all that make different text browsers look different, perhaps the developers could talk to each other to make their browser more like one another, to thwart just this kind of privacy invasion.
How do you tell different w3m or lynx users apart if they spoof their user agents?
Try to write any sort of backtracking solver (e.g. integer or constraint programming, Sudoku, alpha-beta pruning, MCTS game AI) without recursion. Or even just a password brute-forcer. Listing every password of length n is a pain without recursion; you either get a loop n deep or you end up implementing your own stack.
What these problems have in common is that they grow so fast that the computer doesn't really have time to reach stack overflow. Consider chess, for instance. No chess AI can go 100 turns deep, but pretty much every stack can handle 100 entries.
That's interesting. One problem I've happened upon when playing with adversarial learning like that (though in a genetic algorithm context) is that the programs forget what has happened before; e.g. in a rock-paper-scissors setting, the first system learns rock, then the second learns paper, then the first learns scissors, then the second learns rock and you're back where you started. Presumably they have some way of avoiding this with GANs.
If it's possible to do by a robot arm, it should be possible to do by faking the input from the stylus system. All you'd need is something like a finite element model of the physical system involving the robot and stylus (in the very worst case).
Here's a socialist hellhole where the government is involved in everything. It seems to be working out just fine.
I suspect there's a selection effect: the kind of people who get filthy rich are the kind who are extremely motivated by money. So they wouldn't stop working. If they see money as score and the highest score wins, then they keep working even after they're rich. On the other hand, the kind of people who would quit after $10M may not have the ruthless drive to get there no matter what it takes.
Both are. The enemy of your enemy is not necessarily your friend!
1. Make a program that asks the CPU if it's malware.
2. Have the program do malwary stuff if the CPU says it's not malware, and do benign stuff otherwise.
3. Profit! (Or laugh.)
Or they pause the VM, apply the patch, and continue said VM and you'll be none the wiser. The attempt to break out just fizzles after a second.
Isn't a clip show partially new? And all bad.
The only way you can have a DRE that's anywhere near possible to trust, that is.
Since DREs are usually general-purpose computers running Windows or something, no points for guessing which category they're part of. And the only way you can have a DRE is to make it as simple as possible, which means no Turing-complete hardware. Just a row of buttons, a matrix display, and counters. Or you know, paper and pencil.
Kurzweil (on the one hand) and all those people thinking the terminator scenario will happen (on the other) feel like theists who try to quash their theism by force, but it just pops up elsewhere in another shape.
Kurzweil's "we're all going to be immortal and the singularity will bring plenty to all" is: technology will let us make God and we will all go to techno-Heaven.
The terminator/golem scenario with the out-of-control superintelligences turning the whole world into computer material is: technology will let us make God and we will all go to techno-Hell. The more you get into the really bizarre end of the theology, the more obvious it is.
You mean like Range voting?
1X = 150KB/s
:)
5GB/10s = 500MB/s = 5e5 KB/s
which gives 3333x.
Hey, you asked
It's the corporate version of dictatorships holding elections. Nobody believes them but they feel they have to do it anyway.
I can think of something more useful. See if the site computer has a modem or phone connection or something, and then bridge over onto the Kwangmyong intranet. Port scan and download everything! ... although at dialup speeds, it would take a while.
Statistically speaking, the variance decreases with the sample size. That is, if you have a legislature of 500, the chance of getting a majority of say, people who'd like to turn the country into 1984 is pretty close to 0%. But if you select the president by lottery, you may be very unlucky and get a Big Brother. In other words, yes, it is too risky.
I agree with you that "a king by another name" isn't really a good way to do things. A parliamentary sortition system would solve this without any problem, because the parliament would work as an electoral college. But you're not going to change the US into a parliamentary form of government. Hence an explicit electoral college instead of an implicit one.
Maybe it is a little silly to compare the chances of getting sortition with the chances of becoming more parliamentary. But I think the latter is a lot less likely than the former, for some reason... at least until the former has run for a while. Perhaps the system would decide to reorganize in that manner after a while.
Voting reform is very hard because the people who benefit from the biased voting methods are also those who decide what goes.