There is a book called Behavioral Game Theory, I think edited by Camerer (he's the author associated with the title, whether he wrote all the content or not...) A collection of experiments and essentially what they mean (theoretically) in various situations. Very interesting stuff.
One could consider Einstein's thought experiments as not really experiments at all, but logical constructs... I'm not sure what you'd call that, really, but its definitely not "doing experiments and analyzing results to reach conclusions."
On the other hand I'm a just a computer scientist grad disconnected about whether I spend more time running experiments or reasoning about my (sometimes theoretical) constructs...
Not to put too fine a point on it (or to try to talk military/political strategy), but I agree. But I think the problem in Iraq isn't that the soldier shouldn't have non-lethal options. Its that the soldier shouldn't be a cop (special forces, etc. notwithstanding.) Its two different jobs (as you pointed out, different training,) and in an all volunteer army, its best to make the at least common job (special forces, etc. notwithstanding) as easy as possible so that the widest number of people are capable of doing it.
What's interesting here is that you're mentioning cops as an example and he's mentioned soldiers. For cops, lethal weapons are used rarely (or supposed to be...) and subdued means constrained by handcuffs... for a soldier subdued means dead. So from the perspective of cops its an escalation that makes them lazier about how to get a subject into handcuffs (and thus shouldn't be given to police) for soldiers, however, it allows them to avoid killing people (without putting themselves at as much risk as current methods), which is nice in an occupation situation. Seems to me the only problem is how to keep it in the hands of only military for the rest of time.
I only know of one proof for why 1 + 1 = 2, and I've been wondering if there are other proofs. It *is* almost too bad that those abstract concepts aren't taught more at the younger age. I asked some of my nieces and nephews why 2 + 2 = 4 and they essentially showed me the proof on their fingers (although using the whole numbers which makes sense because they haven't really been taught 0 yet...)
Anyway the proof as I know it is this: Define 0 as a number. Define a successor function which takes a number as input and produces a number as output. Then start defining some labels like 1 (doesn't really have to be 1, could be the Symbol formerly known as Prince... just a label... still the same crazy music genius... this, it would be nice if were explained more...) is the Successor of 0, 2 is the Successor of the Successor of 0, 3 and then 4 in the same way. Then finally define + as the following construction: 0 + any number = that any number and S(x) + S(y) = x + S(S(y).
2 + 2 = 4
S(S(0)) + S(S(0)) = S(S(S(S(0)))) by definitions above.
S(0) + S(S(S(0))) = S(S(S(S(0)))) by the second rule of +
0 + S(S(S(S(0)))) = S(S(S(S(0)))) again by the second rule of +
S(S(S(S(0)))) = S(S(S(S(0)))) by the first rule of +
QED
Anyway, ask some 6 year old who knows how to count on their fingers... they'll show you that (holding two sets of fingers on either hand and then counting the "successors" by dropping fingers as they go.)
I think it would even be a value-add for the subscription to get this kind of rehash of any article with > 300 or so comments. Its not like I couldn't go back and read the original, but if I didn't want to... I'd pay.
Re:Nintendo's Wii akin to Chevrolet's Nova?
on
Both Sides of Wii
·
· Score: 1
I sorta imagine a 8 - 10 hour long committee on naming where someone goes "what if..." and then makes the statements about 'i' looking like people and the controller and putting it together is like breaking down the walls... and then finally with shimmery eyes he brings his hands together and says "synergistically" and then committee members so mind numbed from being in the tedium for 8 - 10 hours feel relieved that finally, finally there are enough clever bullet points behind it that they can go home.
Beer is much like wine in one regard, the more you do know about its make and history, the better you'll be at finding finished products you'll enjoy. (And i'll admit its kinda fun to savor it, going "mmm taste the hops, baby" even if it's pretentious.)
That was sort of my first thought on this... is the fact that China made Google censor available by Google search? I'm not trying to be morally ambiguous, but the ends could justify the means (so I want to try to reserve judgment until the "end":)
As people have said, what should google do? The only way to change a civilization is either overtly by war or subvertively by well subversion:) Google most likely thinks they provide better information than either Yahoo or Microsoft. Therefore they should end up thinking they are providing better information to the Chinese people.
Anyway they should be abused for their trite "do no evil" in the face of something which doesn't have a true black or white. However if in the end the Chinese government finally faces up to the rest of the world because the population becomes more enlightened by its access to the internet, it could be a good thing. We might actually get a merging of east and west/old and new cultures. In that case, I hope some people exonerate Google. On the other hand, we may just have world war 3... time will tell.
Its not exactly cool technology, but public transit already solves that problem. Adding the ability to remove the driver would serve to make public transit more efficient.
But if you were able to actually stop driving your car that way, the only difference (other than a huge gain in fuel economy) between an individual car and a bus (or something equiv.) would be the privacy...
You made me think, though... boiling liberals down they really are *always* the ones seeing what's wrong and therefore are the ones who are enacting change. The conservatives are simply worried about the liberals changing things... its like the whole point of calling them liberal or conservative...
What it made me think of was an analogy of a car... a gas pedal and a brake pedal. Both are required if you want to get where your going safely... only the gas is required if you're not so worried about safety... and only the brake is required if you really don't care if you get there.
So assuming liberals are the gas and conservatives are the brake... I hope whoever is steering isn't drunk.
Its sorta sad when I think about it. Apparently the levies in New Orleans would have been a massive undertaking to replace and worked fine for the most part.
Not to arbitrarily bring it up, but I was thinking "what would it take" and it would take a critical attack on this central repository for anyone to realize that a central point of failure is a problem.
Its sad because it seems a policy of reaction is the best we'll ever get. For some reason preemption is too hard to explain.
You reminded me of one of my own "geezer"-ish peeves. People who abuse compile times by trying something (anything), and compiling it to see if that one worked, and if not they just try something else (usually by adding.) I almost wish compilers still took hours to build code so that you had to actually *think* about what it was you were changing or else you wasted an entire hour on each mindless change you did. Watching the uninitiated program today makes me feel like a carpenter watching someone hammer yet another nail into a board that just won't stay up. Scripting languages can exacerbate that problem, I think.
I sometimes think that's related to the influx of people who do it for the money. I think I still count as young blood (under 30.) And I know my friends and I all know the history of the machine (if we haven't necessarily done punch cards, I have respect for the fact that I don't have to carry a stack in a particular order carefully from one end of campus to the other.)
The only people I can think of who wouldn't are a few of the people I know who have learned the technology trade, not grew up with a passion for the machine. (Note when I say friends above i mean the latter. I make friends with similar people.)
So yeah, I think anyone who has a real interest in computation studies knows with some interest how circuits are arranged, how Turing machines work, is at least afraid of the y-combinator, and knows that language fights are dumb.:) I think once the pay starts decreasing again, then things won't be taken for granted *quite* as much.
One mild caveat to all of this, however. Managing complexity means abstracting. As we continue to add complexity there's a point at which some people just won't want to understand how a machine works inside. They blackbox it and move on. Hopefully they'll still get a top-level from it.
There was very little in the way of research into computer-emulating-human intelligence, which is probably a good thing (read: less of a waste of money) considering how little progress the Minsky crowd has made in the past thirty years.
uh... wouldn't that be *why* the Minsky crowd has made little progress? Personally, I found it extremely disappointing how little it seems people spend researching the big questions. (Although in my lab, nearly everyone is doing very useful and cool stuff in cancer... and the senior PhD student may have made a correlation discovery about where dimensia may be located (or localised) in the brain... he used bayes net structure search on MRIs... I think it was MRIs)
Anyway, I read somewhere recently Minsky was trouncing the AI community for being so risk adverse and not struggling for years on the hard problems and instead spending time on getting results. You can't really blame the community though, if you don't show results you don't get funding. Nobody trusts open-ended research...
Oh, and my research is behavioral cloning (trying to get an agent to act like a specific human in some task) which is why it annoys me because there is practically zero research in AI as to how to tell whether I'm cloning well or not (closest thing I can find is the human-computer interaction people...)
That seems to make things sound much more devious...Wouldn't sufficient recompression algorithms render most watermarks void??
The point of most watermarking algorithms is to make it so that by the time you've modified the original enough that the watermark is gone, the original's quality is so poor it doesn't matter. Its very similar to the cryptography idea of making cryptography strong enough for how long you want the information to be secret. A lot of the watermarking schemes store the information as modification to the phase of the image (See Fourier Transform) with the idea being you put a certain amount of energy (amplitude) into your watermark, but just below the point where it modifies the image. Then the number of modifications to the image to destroy that same energy has to be severe (I can't remember the size, but I think its squared.)
Anyway I used the same software as an unpublished study from my department to do the sorta opposite study... how much compression before the image is destroyed. So I can't speak with certainty about the results of that other study, but I think they decided that all modern watermark methods can survive any modern compression (and other standard paintshop mods like crop) and not modify the original image. Maybe if either study gets more thorough they can be combined to show at what level it gets "crappy" (you see why its unpublished:) from both sides.
I like a good sound argument like this. I think you do justice to the non-"right to life" side, but I think this:
The idea that we are killing human life by using the Embryonic Stem Cells will be stood on its ear as soon as we can make a human from a skin cell, or any number of cells not currently involved in reproduction. So is it that the gene is sacred? Or the fact that a cell is involved in reproduction? Is a person created from a skin cell going to be just as valid/human as a "natural born" person.
diverts the point of the argument. It may be the case someday but its also a big leap from anything that's currently possible to assume that reproductive cells *won't* be involved. What I could see occuring is a redefinition of "reproductive", but "what if" doesn't really serve anybody. I think its important to note that a consistent Right to Life argument usually involves saying that once fertilization has occured and that a human will at some point in the future probably form, then its "life." Whether that is correct or valid etc is a debatable point, but it doesn't immediately become invalid when you switch what caused the "at some point in the future."
I think, when you look at life and humanity from a consciousness/value perspective --then it is a lot easier to deal with new advances in science and leads to rights and ethical treatment for new type of life we have not yet considered.
This is a very bad way to argue, just because it makes it easier, doesn't mean one should accept it. That is very nearly the opposite of "principled." Which is usually a good thing to have in a particular ethic.
I disagree that learning is required if you have this:
Of course, part of the problem is also building the AI to act Human. Humans make mistakes, and so should the computer. In warfare, there's always been that element of random chance where you can capitalize on an enemy's mistakes. Take in factos like morale, confidence, etc. It's no fun to play against a perfect oponent all the time:p
with *enough* human like characteristics I think the interaction would be more interesting and be "artificial intelligence" whether its learning or not. It would still feel correct, just not on the second time you play the game (unless its a *very* broad range.)
My research deals with this so I'm biased:) but I think the first step is to increase the range of agent interaction first. As in what you said, make it make mistakes. Make other AIs that make different mistakes and have a uniqueness. Once we get that (and maybe a whole lot more) I think dealing with the complexity of an AI that *learns* from those mistakes will be a little bit easier.
As a stepping stone though I think the games industry should go toward failing like a human. See if you can match the same feel of getting someone pissed off at you online. besides just having the AI call you an "aimbot" or something. Despite the situational irony. Let the AI appear to start acting stupid because its pissed off that some plan failed. If convincing it adds reward as well as giving the AI a failure point to exploit in the game (especially if the failure point is unique for different agents)
Btw, just because its fun to talk about my research, I'm doing behavioral cloning. Trying to get the nuance of different humans into different agents (especially the failure modes). Anybody else doing something like that?
Slashdot Mirror
There is a book called Behavioral Game Theory, I think edited by Camerer (he's the author associated with the title, whether he wrote all the content or not...) A collection of experiments and essentially what they mean (theoretically) in various situations. Very interesting stuff.
On the other hand I'm a just a computer scientist grad disconnected about whether I spend more time running experiments or reasoning about my (sometimes theoretical) constructs...
Not to put too fine a point on it (or to try to talk military/political strategy), but I agree. But I think the problem in Iraq isn't that the soldier shouldn't have non-lethal options. Its that the soldier shouldn't be a cop (special forces, etc. notwithstanding.) Its two different jobs (as you pointed out, different training,) and in an all volunteer army, its best to make the at least common job (special forces, etc. notwithstanding) as easy as possible so that the widest number of people are capable of doing it.
What's interesting here is that you're mentioning cops as an example and he's mentioned soldiers. For cops, lethal weapons are used rarely (or supposed to be...) and subdued means constrained by handcuffs... for a soldier subdued means dead. So from the perspective of cops its an escalation that makes them lazier about how to get a subject into handcuffs (and thus shouldn't be given to police) for soldiers, however, it allows them to avoid killing people (without putting themselves at as much risk as current methods), which is nice in an occupation situation. Seems to me the only problem is how to keep it in the hands of only military for the rest of time.
Anyway the proof as I know it is this: Define 0 as a number. Define a successor function which takes a number as input and produces a number as output. Then start defining some labels like 1 (doesn't really have to be 1, could be the Symbol formerly known as Prince... just a label... still the same crazy music genius... this, it would be nice if were explained more...) is the Successor of 0, 2 is the Successor of the Successor of 0, 3 and then 4 in the same way. Then finally define + as the following construction: 0 + any number = that any number and S(x) + S(y) = x + S(S(y).
2 + 2 = 4
S(S(0)) + S(S(0)) = S(S(S(S(0)))) by definitions above.
S(0) + S(S(S(0))) = S(S(S(S(0)))) by the second rule of +
0 + S(S(S(S(0)))) = S(S(S(S(0)))) again by the second rule of +
S(S(S(S(0)))) = S(S(S(S(0)))) by the first rule of +
QED
Anyway, ask some 6 year old who knows how to count on their fingers... they'll show you that (holding two sets of fingers on either hand and then counting the "successors" by dropping fingers as they go.)
Even worse are the transcendental numbers! :) I would totally pay unlimited amounts of money to someone who designed me Chaitin's Omega on a chip! :)
Don't forget the Boost libraries! Its so fancy, its like programming java (but I'm too blue collar for that :)
I think it would even be a value-add for the subscription to get this kind of rehash of any article with > 300 or so comments. Its not like I couldn't go back and read the original, but if I didn't want to... I'd pay.
There's a short-hand for this "copy and paste" based approach to writing... I believe some people call it "editing" :) ... /. whiners are funny.
This quote always pissed me off when I hear it. Its stupid. We should stand on their shoulders, not sit at their feet.
This is getting infinitely rediculous.
in fact, "none of wii is as dumb as all of wii"
I'd rather have a signed picture of cowboy neal.
Beer is much like wine in one regard, the more you do know about its make and history, the better you'll be at finding finished products you'll enjoy. (And i'll admit its kinda fun to savor it, going "mmm taste the hops, baby" even if it's pretentious.)
As people have said, what should google do? The only way to change a civilization is either overtly by war or subvertively by well subversion :) Google most likely thinks they provide better information than either Yahoo or Microsoft. Therefore they should end up thinking they are providing better information to the Chinese people.
Anyway they should be abused for their trite "do no evil" in the face of something which doesn't have a true black or white. However if in the end the Chinese government finally faces up to the rest of the world because the population becomes more enlightened by its access to the internet, it could be a good thing. We might actually get a merging of east and west/old and new cultures. In that case, I hope some people exonerate Google. On the other hand, we may just have world war 3... time will tell.
Its not exactly cool technology, but public transit already solves that problem. Adding the ability to remove the driver would serve to make public transit more efficient.
But if you were able to actually stop driving your car that way, the only difference (other than a huge gain in fuel economy) between an individual car and a bus (or something equiv.) would be the privacy...
Dude, you are *way* to lucid for /.! :)
You made me think, though... boiling liberals down they really are *always* the ones seeing what's wrong and therefore are the ones who are enacting change. The conservatives are simply worried about the liberals changing things... its like the whole point of calling them liberal or conservative...
What it made me think of was an analogy of a car... a gas pedal and a brake pedal. Both are required if you want to get where your going safely... only the gas is required if you're not so worried about safety... and only the brake is required if you really don't care if you get there.
So assuming liberals are the gas and conservatives are the brake ... I hope whoever is steering isn't drunk.
Not to arbitrarily bring it up, but I was thinking "what would it take" and it would take a critical attack on this central repository for anyone to realize that a central point of failure is a problem.
Its sad because it seems a policy of reaction is the best we'll ever get. For some reason preemption is too hard to explain.
You reminded me of one of my own "geezer"-ish peeves. People who abuse compile times by trying something (anything), and compiling it to see if that one worked, and if not they just try something else (usually by adding.) I almost wish compilers still took hours to build code so that you had to actually *think* about what it was you were changing or else you wasted an entire hour on each mindless change you did. Watching the uninitiated program today makes me feel like a carpenter watching someone hammer yet another nail into a board that just won't stay up. Scripting languages can exacerbate that problem, I think.
The only people I can think of who wouldn't are a few of the people I know who have learned the technology trade, not grew up with a passion for the machine. (Note when I say friends above i mean the latter. I make friends with similar people.)
So yeah, I think anyone who has a real interest in computation studies knows with some interest how circuits are arranged, how Turing machines work, is at least afraid of the y-combinator, and knows that language fights are dumb. :) I think once the pay starts decreasing again, then things won't be taken for granted *quite* as much.
One mild caveat to all of this, however. Managing complexity means abstracting. As we continue to add complexity there's a point at which some people just won't want to understand how a machine works inside. They blackbox it and move on. Hopefully they'll still get a top-level from it.
uh... wouldn't that be *why* the Minsky crowd has made little progress? Personally, I found it extremely disappointing how little it seems people spend researching the big questions. (Although in my lab, nearly everyone is doing very useful and cool stuff in cancer... and the senior PhD student may have made a correlation discovery about where dimensia may be located (or localised) in the brain... he used bayes net structure search on MRIs... I think it was MRIs)
Anyway, I read somewhere recently Minsky was trouncing the AI community for being so risk adverse and not struggling for years on the hard problems and instead spending time on getting results. You can't really blame the community though, if you don't show results you don't get funding. Nobody trusts open-ended research...
Oh, and my research is behavioral cloning (trying to get an agent to act like a specific human in some task) which is why it annoys me because there is practically zero research in AI as to how to tell whether I'm cloning well or not (closest thing I can find is the human-computer interaction people...)
The point of most watermarking algorithms is to make it so that by the time you've modified the original enough that the watermark is gone, the original's quality is so poor it doesn't matter. Its very similar to the cryptography idea of making cryptography strong enough for how long you want the information to be secret. A lot of the watermarking schemes store the information as modification to the phase of the image (See Fourier Transform) with the idea being you put a certain amount of energy (amplitude) into your watermark, but just below the point where it modifies the image. Then the number of modifications to the image to destroy that same energy has to be severe (I can't remember the size, but I think its squared.)
Anyway I used the same software as an unpublished study from my department to do the sorta opposite study... how much compression before the image is destroyed. So I can't speak with certainty about the results of that other study, but I think they decided that all modern watermark methods can survive any modern compression (and other standard paintshop mods like crop) and not modify the original image. Maybe if either study gets more thorough they can be combined to show at what level it gets "crappy" (you see why its unpublished :) from both sides.
The idea that we are killing human life by using the Embryonic Stem Cells will be stood on its ear as soon as we can make a human from a skin cell, or any number of cells not currently involved in reproduction. So is it that the gene is sacred? Or the fact that a cell is involved in reproduction? Is a person created from a skin cell going to be just as valid/human as a "natural born" person.
diverts the point of the argument. It may be the case someday but its also a big leap from anything that's currently possible to assume that reproductive cells *won't* be involved. What I could see occuring is a redefinition of "reproductive", but "what if" doesn't really serve anybody. I think its important to note that a consistent Right to Life argument usually involves saying that once fertilization has occured and that a human will at some point in the future probably form, then its "life." Whether that is correct or valid etc is a debatable point, but it doesn't immediately become invalid when you switch what caused the "at some point in the future."
I think, when you look at life and humanity from a consciousness/value perspective --then it is a lot easier to deal with new advances in science and leads to rights and ethical treatment for new type of life we have not yet considered.
This is a very bad way to argue, just because it makes it easier, doesn't mean one should accept it. That is very nearly the opposite of "principled." Which is usually a good thing to have in a particular ethic.
I disagree that learning is required if you have this: Of course, part of the problem is also building the AI to act Human. Humans make mistakes, and so should the computer. In warfare, there's always been that element of random chance where you can capitalize on an enemy's mistakes. Take in factos like morale, confidence, etc. It's no fun to play against a perfect oponent all the time :p
with *enough* human like characteristics I think the interaction would be more interesting and be "artificial intelligence" whether its learning or not. It would still feel correct, just not on the second time you play the game (unless its a *very* broad range.)
My research deals with this so I'm biased :) but I think the first step is to increase the range of agent interaction first. As in what you said, make it make mistakes. Make other AIs that make different mistakes and have a uniqueness. Once we get that (and maybe a whole lot more) I think dealing with the complexity of an AI that *learns* from those mistakes will be a little bit easier.
As a stepping stone though I think the games industry should go toward failing like a human. See if you can match the same feel of getting someone pissed off at you online. besides just having the AI call you an "aimbot" or something. Despite the situational irony. Let the AI appear to start acting stupid because its pissed off that some plan failed. If convincing it adds reward as well as giving the AI a failure point to exploit in the game (especially if the failure point is unique for different agents)
Btw, just because its fun to talk about my research, I'm doing behavioral cloning. Trying to get the nuance of different humans into different agents (especially the failure modes). Anybody else doing something like that?
Hooked on fonicks just doesn't work.