"Obviously you wouldn't present the statistics to the judge in that way. You'd say, "I have a method for selecting people from a crowd. The people I select are 100x more likely to have committed a crime than people I don't select. Don't you agree that if someone is 100x more likely than the average person to have committed a crime, that that's 'substantial chance'?""
Judges aren't stupid. They are experts on how to interpret exactly these sorts of statistics, and they know that you, as a lawyer, have taken classes and passed exams ensuring that you also know that that argument is BS. You'll be fined.
As far as what % chance constitutes probable cause, the seriousness of the crime is relevant. This isn't catching murderers, it's catching kids with fake IDs; crimes that won't get prosecuted if the DA is feeling busy that day. Probable cause there is going to mean seeing the ID, and not much else.
"Now let's say you have a method of selecting people that allows you to get a 20% guilty rate"
Let's not. All manner of hypothetical systems might hypothetically be legal. The actual system in question is more interesting to me; it is legal only because the searchers are not cops, and are not looking for criminals.
You seem to have left out his first significant invention: the portable infusion pump that is now bolted to every iv stand in the industrialized world. Leaving him in a financial position where I don't think he much cares how impressed you are with his subsequent efforts.
I was going for a one sentence summary. There are of course, several law-school term papers worth of detail if you like. "Substantial chance", "fair probability"... try to tell a judge these mean one chance in a hundred. If he's in a good mood, he'll laugh and pretend you were joking rather than citing you for contempt.
"If those flagged by "the system" turn out to be 100x more likely to be guilty of a crime than those not flagged, then its conceivable that "being flagged" might, in and of itself, meet the criteria for probable cause."
"the system" is officials searching people based on hunches. So 100X more likely to be a criminal than those not flagged seems unlikely.
But that's irrelevant anyway. The "criteria for probable cause" is that a reasonable person would have reason to believe a crime is probably being committed. We know that of those flagged, 1% are found to be commiting a crime. 1 in 100 isn't "probable".
They should rename it as you describe, because that's what it's being used for. Then we can take them to court, because it's obviously illegal.
Those reasons are mythical. C++ is a superset of C (with minor exceptions not relevant here). It cannot be inherently slower, because you can compile the same code and get, at worst, the same machine code.
Some code (including notable portions of the standard library) winds up faster in C++ because templates let the compiler do better optimization. Analagous reasons might let AS bytecode derived from C be faster than bytecode derived from some more typical AS language. But C++ can't lose to C, because it contains it.
This was before "texting"; we thought it was awfully new-fangled that caller id let us drop the "Hello", cutting our word count in half. So yeah, I'm old.
"To judge whether 1% is actually decent, we'd need to know what percentage of *all travelers* are guilty of the offenses they're arresting the 1% for."
No, we need to know what percentage of travelers are terrorists (not many), and how many this is catching (none). Catching terrorist is, and must be, the purpose of the system.
If the purpose of this system is to catch the petty criminals, it's illegal; there's no probable cause. The system is only legal if finding fake IDs is just an inadvertent side-effect of the legitimate purpose of catching terrorists.
The system must be at least one of three things: Ineffective, pointless, or illegal.
"If arrest rate isn't the statistic you would use to determine efficiency what would you use? In fact what else could you measure considering the systems only goal is to spot criminals?"
The goal is not to spot criminals. The goal is to spot terrorists. They haven't spotted any, indicating the system is a failure, the system is pointless, or both.
"Presumably you could possibly use detection rate but that would mean that you would be stopping people, searching them and letting them go even if they had done something (petty) wrong."
As they should be. You didn't have probable cause for the search, and anything you find should be inadmissible. This weird concoction of someone who can search you just because they feel like it, but isn't bound by the rules put on actual cops is a travesty.
"that abstract term [turing machine] is being used abstractly to refer to modern CPUs."
Which nobody saying anything meaningful would do. It's interesting to know if your machine can simulate a turing machine, because it means it can do a certain (very large) set of things that tturing machines can do. It's interesting to know if a turing machine can simulate your machine (probably it can) because then you know your machine cannot do the things we know a turing machine can't do.
"That a lot of the neural networks are being emulated by turing logic is a side issue... and it shows that even an emulated neural net is faster than turing logic at solving certain classes of problems."
So you're saying turing logic emulating a neural net is faster than turing logic. That would be obviously false if "turing logic" meant anything in this context.
"I don't understand why you think we can't talk about performance. For example, we can write turing logic to construct a hyperplane to (for example) look for underpriced real-estate among a hundred columns of property attributes, but any programmer can imagine the levels of if/then logic that would be required for that. A neural net is massively better suited for that kind of work. If we had a contest for who could find the outliers with minimal code/cycles/memory/programming/time, who do you suppose would win?"
I would; that's exactly the sort of problem I write code for all the time. But you can't talk about performance the way you are, because you talk about comparing "turing logic", which would mean code written to run on a computer, or rather, to run on anything you could possibly imagine that would have the ability to perform automated calculations. That's what a Turing Machine is: a simple (imaginary) machine that can be shown to be capable of running any deterministic calculation that it is possible to run.
For our stupid car analogy of the day, you're asking which will get me to the store faster: going down Main street and taking a left on Elm, or a vehicle capable of overland travel?
In this post I will in no way say that you are an idiot. Regardless of what opinions I may or may not have regarding your possible lack of any functioning brain cells, I am definitely not implying here that you are a moronic baboon. You should not take offense at the idea that I think you are clearly a stupid twit who gets in pedanitic arguments on slashdot because nobody will talk to him otherwise due to his disgusting personal habits best left undescribed. I'm definitely not saying that. Under no circumstances should you conclude that I would have any purpose in writing these sentences, or be trying to imply anything, because obviously the only thing I could be implying is your utter mental and moral deficiency; and I'm not saying that. So don't put words in my mouth by pretending anything at all has been said by this post, because that would only be believed by a moron, which I'm not saying you are.
Ahh, I see... but that's a stupid point. Random facts looked up on Wikipedia are almost always correct, so dismissing someone for using it, while offering nothing of your own, is idiotic.
Wikipedia is an excellent source to refer to. It's easy to find the article yourself, and check the citations. The magic source of unimpeachable, well sourced information that you might want does not exist. The best we can hope for is a general reference source is one maintained by a community of rabid pedants constantly shouting "citation needed" all over the place.
Wow, your reading comprehension is just epically bad. You can't even correctly understand the post where someone explains you misunderstood their post.
"If you know enough to ask that question, then you already know the answer, and you're wasting everyone's time."
You're already wasting everyones time by spouting big words you heard in your undergrad AI course when you don't really know what you're talking about.
Nobody does anything on a Turing machine: it's an imaginary device in a thought experiment. Academics note that particular aparatus can simulate a turing machine it must be theoretically capable of all the operations a turing machine is capable of. Saying "Neural networks are... one or two or a hundred orders of magnitude better at their tasks" is meaningless. "Neural network" is an abstract description of a type of algorithm, and you're comparing it's performance to a thought experiment on the limits of computability. It would be ridiculous to compare their performance if it were even meaningful to describe either in terms of performance, which it's not.
When printing presses came in they made copying comparatively easy, and created a totally different situation than the enormous effort needed for hand-copying. Intellectual property laws had to change in response.
The advent of digitally stored information, that can be copied trivially, it creates a totally different situation than printed information, I agree entirely. intellectual property laws will certainly need to change again.
How and into what? I have no idea. But I don't think it makes sense that trivial copying will cause a return to the intellectual property concepts that held sway back in the day when the enormous effort of hand copying was required.
If we ever reach 'star trek' replicator technology, then the only thing of real value will be information; so I expect our societal rules will reflect that.
"And, I kid you not, a big part of what will get us there is the not-yet-established zero-G porn genre."
Zero-G: A fine example of something associated with space that is much more cheaply had in the atmosphere. Build your porn studio into an appropriate airplane and have it fly parabolic trajectories all day for years before you get to the cost of a day in actual orbit.
"we have the technology to send you to meet anyone you could speak to via the Internet."
But I have no desire, time or expectation of meeting them, yet I find it worthwhile to communicate with them.
By analogy, I don't think eventually going there is the only reason to explore space. Even if you do want humans to leave earth, I suggest the best way to advance that now is by sending probes to learn more about it.
"1) Backup Humanity"
In my estimation, long before we are able to establish a self-sufficient off-world colony, we'll have artificial robot bodies anyway; so robotic exploration will be better experience.
"2) Some things you can only learn by going there." I can't think of any that can't be learned faster by sending a probe. Certainly there are a lot of things we could have learned with probes by now if we didn't waste so much money using humans to not learn anything.
"3) Maybe only 'because it's there'" A great reason to explore it; let's not fail by insisting on doing it a stupid way. i.e. spending all our money on people only pretending to explore.
"But propping the industry up by arbitrarily banning people from using their computers from copying particular bits of data in particular ways is demented."
We've long "propped up" the book publishing industry by arbitrarily banning people from using their printing presses to print particular books. It is completely artificial, but after 300 years of it, it doesn't seem to have been entirely destructive.
The age of copying bits on computers is as different from the age of printing copies on presses as that was from the time of monks writing out new copies by hand. It will undoubtedly need different rules, but I find it hard to imagine there won't be any workable legal restrictions at all. In any case, that fight will happen over music, pictures and prose. Whatever protection schemes we devise for those, executable code can throw technological stumbling blocks on top.
So to get back to the only point I was really trying to make in the first place: Free software methods will not be able to come up with ways to fund software development that will be competitive with the straightforward proprietary model for certain sorts of software.
"An industry that exists only because its illegal to make and share copies."
I disagree. It's easy to have proprietary software, even without copyright.
"So set your software up as a web service with crucial proprietary guts as part of the service and keep it on your server"
Good example.
In any case, there would seem to be little reason to think Free software can easily take over development currently funded via shrink-wrap models. You can set up a much cheaper, easier business model if you prevent free copying, whether by legal means or technological ones.
"The result? You can't use their changes with your code in a BSD only licensed version."
One of many possible results I specifically endorsed when I chose BSD, and continue to be entirely happy with.
I'll certainly point out to people that if they send me changes under a BSD license, I may integrate them with my main codebase, saving them a bunch of work integrating changes down the road. So that might be better for them. But if their changes need to be GPL, or proprietary, or whatever, for any reason or desire, that's cool too. I'd rather they used the code and not give back changes than not use the code due to licensing issues. Who knows, maybe they'll have changes they can and want to contribute in the future; they certainly won't if they don't use the code. Besides, when it doesn't cost me anything, helping people is pretty much required by my neo-hippy moral code.
When I choose BSD, I specifically encourage anyone who wants to to use the code in any project where it is helpful to them, regardless of license. Why do you imagine I don't mean it?
Slashdot Mirror
"Obviously you wouldn't present the statistics to the judge in that way. You'd say, "I have a method for selecting people from a crowd. The people I select are 100x more likely to have committed a crime than people I don't select. Don't you agree that if someone is 100x more likely than the average person to have committed a crime, that that's 'substantial chance'?""
Judges aren't stupid. They are experts on how to interpret exactly these sorts of statistics, and they know that you, as a lawyer, have taken classes and passed exams ensuring that you also know that that argument is BS. You'll be fined.
As far as what % chance constitutes probable cause, the seriousness of the crime is relevant. This isn't catching murderers, it's catching kids with fake IDs; crimes that won't get prosecuted if the DA is feeling busy that day. Probable cause there is going to mean seeing the ID, and not much else.
"Now let's say you have a method of selecting people that allows you to get a 20% guilty rate"
Let's not. All manner of hypothetical systems might hypothetically be legal. The actual system in question is more interesting to me; it is legal only because the searchers are not cops, and are not looking for criminals.
That is yet another of those neat and tidy rules nature doesn't always pay much attention to.
You seem to have left out his first significant invention: the portable infusion pump that is now bolted to every iv stand in the industrialized world. Leaving him in a financial position where I don't think he much cares how impressed you are with his subsequent efforts.
I was going for a one sentence summary. There are of course, several law-school term papers worth of detail if you like. "Substantial chance", "fair probability"... try to tell a judge these mean one chance in a hundred. If he's in a good mood, he'll laugh and pretend you were joking rather than citing you for contempt.
"If those flagged by "the system" turn out to be 100x more likely to be guilty of a crime than those not flagged, then its conceivable that "being flagged" might, in and of itself, meet the criteria for probable cause."
"the system" is officials searching people based on hunches. So 100X more likely to be a criminal than those not flagged seems unlikely.
But that's irrelevant anyway. The "criteria for probable cause" is that a reasonable person would have reason to believe a crime is probably being committed. We know that of those flagged, 1% are found to be commiting a crime. 1 in 100 isn't "probable".
They should rename it as you describe, because that's what it's being used for. Then we can take them to court, because it's obviously illegal.
"same reasons that C is faster than C++"
Those reasons are mythical. C++ is a superset of C (with minor exceptions not relevant here). It cannot be inherently slower, because you can compile the same code and get, at worst, the same machine code.
Some code (including notable portions of the standard library) winds up faster in C++ because templates let the compiler do better optimization. Analagous reasons might let AS bytecode derived from C be faster than bytecode derived from some more typical AS language. But C++ can't lose to C, because it contains it.
Nothing that can't be fixed with some sweet pony love.
This was before "texting"; we thought it was awfully new-fangled that caller id let us drop the "Hello", cutting our word count in half. So yeah, I'm old.
Why would you think I'm a coward? Why do you think I think anything in particular? I clearly and explicitly said nothing whatsoever.
"Wow, trolls are out in force today."
Hello, pot, I'm kettle. Nice to meet you.
With the advent of caller ID and a sufficiently consistent agenda, one friend and I went a year or so having the same conversation once a week:
*ring*
Guy 1: Go.
*click*
I had to call him back if I wanted to say anything different.
"To judge whether 1% is actually decent, we'd need to know what percentage of *all travelers* are guilty of the offenses they're arresting the 1% for."
No, we need to know what percentage of travelers are terrorists (not many), and how many this is catching (none). Catching terrorist is, and must be, the purpose of the system.
If the purpose of this system is to catch the petty criminals, it's illegal; there's no probable cause. The system is only legal if finding fake IDs is just an inadvertent side-effect of the legitimate purpose of catching terrorists.
The system must be at least one of three things: Ineffective, pointless, or illegal.
Personally, I don't care which.
"If arrest rate isn't the statistic you would use to determine efficiency what would you use? In fact what else could you measure considering the systems only goal is to spot criminals?"
The goal is not to spot criminals. The goal is to spot terrorists. They haven't spotted any, indicating the system is a failure, the system is pointless, or both.
"Presumably you could possibly use detection rate but that would mean that you would be stopping people, searching them and letting them go even if they had done something (petty) wrong."
As they should be. You didn't have probable cause for the search, and anything you find should be inadmissible. This weird concoction of someone who can search you just because they feel like it, but isn't bound by the rules put on actual cops is a travesty.
"that abstract term [turing machine] is being used abstractly to refer to modern CPUs."
Which nobody saying anything meaningful would do. It's interesting to know if your machine can simulate a turing machine, because it means it can do a certain (very large) set of things that tturing machines can do. It's interesting to know if a turing machine can simulate your machine (probably it can) because then you know your machine cannot do the things we know a turing machine can't do.
"That a lot of the neural networks are being emulated by turing logic is a side issue... and it shows that even an emulated neural net is faster than turing logic at solving certain classes of problems."
So you're saying turing logic emulating a neural net is faster than turing logic. That would be obviously false if "turing logic" meant anything in this context.
"I don't understand why you think we can't talk about performance. For example, we can write turing logic to construct a hyperplane to (for example) look for underpriced real-estate among a hundred columns of property attributes, but any programmer can imagine the levels of if/then logic that would be required for that. A neural net is massively better suited for that kind of work. If we had a contest for who could find the outliers with minimal code/cycles/memory/programming/time, who do you suppose would win?"
I would; that's exactly the sort of problem I write code for all the time. But you can't talk about performance the way you are, because you talk about comparing "turing logic", which would mean code written to run on a computer, or rather, to run on anything you could possibly imagine that would have the ability to perform automated calculations. That's what a Turing Machine is: a simple (imaginary) machine that can be shown to be capable of running any deterministic calculation that it is possible to run.
For our stupid car analogy of the day, you're asking which will get me to the store faster: going down Main street and taking a left on Elm, or a vehicle capable of overland travel?
In this post I will in no way say that you are an idiot. Regardless of what opinions I may or may not have regarding your possible lack of any functioning brain cells, I am definitely not implying here that you are a moronic baboon. You should not take offense at the idea that I think you are clearly a stupid twit who gets in pedanitic arguments on slashdot because nobody will talk to him otherwise due to his disgusting personal habits best left undescribed. I'm definitely not saying that. Under no circumstances should you conclude that I would have any purpose in writing these sentences, or be trying to imply anything, because obviously the only thing I could be implying is your utter mental and moral deficiency; and I'm not saying that. So don't put words in my mouth by pretending anything at all has been said by this post, because that would only be believed by a moron, which I'm not saying you are.
Ahh, I see... but that's a stupid point. Random facts looked up on Wikipedia are almost always correct, so dismissing someone for using it, while offering nothing of your own, is idiotic.
Wikipedia is an excellent source to refer to. It's easy to find the article yourself, and check the citations. The magic source of unimpeachable, well sourced information that you might want does not exist. The best we can hope for is a general reference source is one maintained by a community of rabid pedants constantly shouting "citation needed" all over the place.
"No my reply was aimed at you."
Wow, your reading comprehension is just epically bad. You can't even correctly understand the post where someone explains you misunderstood their post.
"If you know enough to ask that question, then you already know the answer, and you're wasting everyone's time."
... one or two or a hundred orders of magnitude better at their tasks" is meaningless. "Neural network" is an abstract description of a type of algorithm, and you're comparing it's performance to a thought experiment on the limits of computability. It would be ridiculous to compare their performance if it were even meaningful to describe either in terms of performance, which it's not.
You're already wasting everyones time by spouting big words you heard in your undergrad AI course when you don't really know what you're talking about.
Nobody does anything on a Turing machine: it's an imaginary device in a thought experiment. Academics note that particular aparatus can simulate a turing machine it must be theoretically capable of all the operations a turing machine is capable of. Saying "Neural networks are
When printing presses came in they made copying comparatively easy, and created a totally different situation than the enormous effort needed for hand-copying. Intellectual property laws had to change in response.
The advent of digitally stored information, that can be copied trivially, it creates a totally different situation than printed information, I agree entirely. intellectual property laws will certainly need to change again.
How and into what? I have no idea. But I don't think it makes sense that trivial copying will cause a return to the intellectual property concepts that held sway back in the day when the enormous effort of hand copying was required.
If we ever reach 'star trek' replicator technology, then the only thing of real value will be information; so I expect our societal rules will reflect that.
"When it is clear that there is money to be made"
There isn't; clearly.
"...and the governments get out of the way..."
Any evidence they're in the way?
"And, I kid you not, a big part of what will get us there is the not-yet-established zero-G porn genre."
Zero-G: A fine example of something associated with space that is much more cheaply had in the atmosphere. Build your porn studio into an appropriate airplane and have it fly parabolic trajectories all day for years before you get to the cost of a day in actual orbit.
"But IS there money to be made?"
No, of course not. It's fantastically expensive to get to, and there is literally nothing there.
"we have the technology to send you to meet anyone you could speak to via the Internet."
But I have no desire, time or expectation of meeting them, yet I find it worthwhile to communicate with them.
By analogy, I don't think eventually going there is the only reason to explore space. Even if you do want humans to leave earth, I suggest the best way to advance that now is by sending probes to learn more about it.
"1) Backup Humanity"
In my estimation, long before we are able to establish a self-sufficient off-world colony, we'll have artificial robot bodies anyway; so robotic exploration will be better experience.
"2) Some things you can only learn by going there."
I can't think of any that can't be learned faster by sending a probe. Certainly there are a lot of things we could have learned with probes by now if we didn't waste so much money using humans to not learn anything.
"3) Maybe only 'because it's there'"
A great reason to explore it; let's not fail by insisting on doing it a stupid way. i.e. spending all our money on people only pretending to explore.
"So you disagree by agreeing with me?"
I had no choice; you said two opposite things.
"But propping the industry up by arbitrarily banning people from using their computers from copying particular bits of data in particular ways is demented."
We've long "propped up" the book publishing industry by arbitrarily banning people from using their printing presses to print particular books. It is completely artificial, but after 300 years of it, it doesn't seem to have been entirely destructive.
The age of copying bits on computers is as different from the age of printing copies on presses as that was from the time of monks writing out new copies by hand. It will undoubtedly need different rules, but I find it hard to imagine there won't be any workable legal restrictions at all. In any case, that fight will happen over music, pictures and prose. Whatever protection schemes we devise for those, executable code can throw technological stumbling blocks on top.
So to get back to the only point I was really trying to make in the first place: Free software methods will not be able to come up with ways to fund software development that will be competitive with the straightforward proprietary model for certain sorts of software.
"An industry that exists only because its illegal to make and share copies."
I disagree. It's easy to have proprietary software, even without copyright.
"So set your software up as a web service with crucial proprietary guts as part of the service and keep it on your server"
Good example.
In any case, there would seem to be little reason to think Free software can easily take over development currently funded via shrink-wrap models. You can set up a much cheaper, easier business model if you prevent free copying, whether by legal means or technological ones.
"The result? You can't use their changes with your code in a BSD only licensed version."
One of many possible results I specifically endorsed when I chose BSD, and continue to be entirely happy with.
I'll certainly point out to people that if they send me changes under a BSD license, I may integrate them with my main codebase, saving them a bunch of work integrating changes down the road. So that might be better for them. But if their changes need to be GPL, or proprietary, or whatever, for any reason or desire, that's cool too. I'd rather they used the code and not give back changes than not use the code due to licensing issues. Who knows, maybe they'll have changes they can and want to contribute in the future; they certainly won't if they don't use the code. Besides, when it doesn't cost me anything, helping people is pretty much required by my neo-hippy moral code.
When I choose BSD, I specifically encourage anyone who wants to to use the code in any project where it is helpful to them, regardless of license. Why do you imagine I don't mean it?
There is no reason classified information can't be public domain.
If it's public domain, you don't have to worry about getting in trouble for copyright infringement. They can still get you for treason.