> Do you really believe that there is any agency > in the federal government capable of > competently carrying out any kind of > conspiracy. Every single bureau i've ever > seen has been filled with morons that sit > at their computer and play solitaire all day > long.
See, you've gone and bought their cover story. Those are specially-trained morons who spend years watching government training videos to reach a state of moronhood that will convince even the most skeptical observer. Meanwhile, the real operatives work behind the scenes, unfettered and unobserved, covering up UFO sightings with carefully-worded obfuscation posted on forums like Slashdot. BTW, I am not one of them, so you can trust that what I say is true.
> Biologists define life as a cellular, self- > replicating entity (more or less)
Well, here in the Agency, we don't make any assumptions about alien life. You should see some of the critters we've captured! Uh-oh, I've said too much, gotta go...
Careful about asking for an eval copy from Hummingbird. The web form doesn't mention that they won't send you the CD until they've had a telesales operator call you to confirm the details. After that, someone from a reseller will call to try to get you to buy. Then, a salesperson from Hummingbird will call. I ended up getting four unwanted calls from them, even though I told the first one I didn't want any more calls, and emailed the sales rep the same thing. If you screen all your calls, or don't mind being endlessly bugged by clueless sales people, go ahead and order the eval.
> But wouldn't doing this make your US > citizenship invalid? No, the U.S. allows its citizens to hold dual citizenship. Plenty of U.S. citizens do this. If you have an Irish grandfather, for example, you can apply for and more or less automatically receive Irish citizenship, which you can hold in addition to your U.S. citizenship.
The number of computations it *can* handle is unbounded
That was the point I was trying to make, but I probably stated it ambiguously.
but for any given nondeterministic step it handles some finite number of inputs
True, but I was responding to this:
Isn't "going through a whole buttload of computations in parallel" exactly what non-deterministic machines do?
...and saying that no, that's not what non-deterministic machines do, unless you define a "buttload" as being equal to the number of inputs at any given step, and either varying with each step or being equal to the maximum that is required by a given computation.
the logic of your argument is: if you can't perceive the difference, then there is no difference. but this is a naive view. a limitation in the perceiver does not change the nature of the thing under observation. the limitation is in the lack of perception, not in the lack of difference.
This is itself a naive, or at the very least incomplete, view. When assessing the capabilities of another intelligence, there are philosophical limits to the degree to which we can determine it by such "hard" measurements as, say, brain mass, number of neurons, etc.
The limitation here is not necessarily in the perceiver, but rather in what it is theoretically possible to perceive, and at what level. [For example, you probably won't be able to tell that I'm imagining a pink penguin just by sticking a probe in my brain. And if you stick enough probes in my brain, I'll stop thinking about the pink penguin and start wondering "what are all these probes doing in my brain?" instead, which would spoil the experiment.]
There are good arguments which say that intelligence cannot be assessed by hard physical measurements, and that the only way to assess it is by interacting with the intelligence itself, not with the underlying substrate. You have to ask it questions and see how it responds. This is the basic reason behind the Turing Test.
We do this all the time when we meet new people - while we might make initial assessments based on superficial characteristics such as appearance, assuming we're open-minded, ultimately we're left assessing the person based on what they say or do, and on how they respond to us.
Assessing an intelligent computer will work the same way. If the computer is capable of holding its own in a conversation, or responds well to an intelligence test for which it has not been specifically prepared, or is able to self-direct its activities the way humans do, then it would qualify as intelligent.
And if you disagree with that, watch out, because next thing you know the computers will be marching (or wheeling) through the streets, demanding equal rights and an end to discrimination against silicon-based lifeforms!
Schneier's quote talks about trying *all* guesses in parallel. Presumably the number of guesses such a system can handle is unbounded. However, going through a "whole buttload" of computations in parallel is still deterministic, it would just be "n" times faster than a serial solution, where n equals the size of a buttload.
Is Jon Katz an artificial intelligence? It took me way too long to work it out, but all the signs are there. The liberal use of the latest buzzphrases, to hide lack of actual content. The slightly off-base parotting of well-known concepts, as though he doesn't quite get it.
It also explains that interminable "Road to Linux" series which never manages to achieve actual use of Linux. Since the Katz software is presumably running on Linux, for Katz, writing about Linux is like exploring one's own subconscious - a task which can only be done incompletely, at best.
Hats off to CmdrTaco for an awesome piece of coding! Is it written in PERL? What hardware does Katz run on? (Hmmm, wonder if that matchbox-sized computer has anything to do with this?)
"it seems that we'd get farther by concentrating on advancing 'traditional' symbolic artificial intelligence, rather than simulating huge neural networks on puny serial hardware... "
Traditional symbolic AI is unlikely to ever achieve much than the traditional applications, e.g. grammar checking, grading essays, playing chess.
You're right about the advantages of the massive parallelism of neural networks. So, we have to build *real* neural networks in hardware (or wetware?), not simulate them on serial hardware (which as you point out is doomed)
"there's the small issue of neural networks being 'opaque' to the creator - all's good when they work, but when they break it's difficult to figure out why."
You've hit the nail on the head about Katz. To report on a subject in a way that adds value, you need a pretty good understanding of the subject - unless you're writing for an audience with an even more limited understanding. Some rudimentary analytical skills also don't hurt (such as, say, the ability to distinguish between problems caused by UPS during shipping, vs. the intent to install Linux on a machine...)
For an example of a writer adding value to a subject (while simultaneously doing a bit of a snow job), check out the article by Michael Lewis (author of "Liar's Poker") in last weekend's New York Times Magazine, about the near-demise and bailout of Long Term Capital Management.
Of course, Lewis probably gets paid more than Katz, but if so there's a reason for that...
Oh please. Yes, let's all ask/. to withhold information from us if it might be the least bit controversial. God knows/.ers wouldn't be able to handle such a thing!
Slashdot Mirror
> in the federal government capable of
> competently carrying out any kind of
> conspiracy. Every single bureau i've ever
> seen has been filled with morons that sit
> at their computer and play solitaire all day
> long.
See, you've gone and bought their cover story. Those are specially-trained morons who spend years watching government training videos to reach a state of moronhood that will convince even the most skeptical observer. Meanwhile, the real operatives work behind the scenes, unfettered and unobserved, covering up UFO sightings with carefully-worded obfuscation posted on forums like Slashdot. BTW, I am not one of them, so you can trust that what I say is true.
> Biologists define life as a cellular, self-> replicating entity (more or less)
Well, here in the Agency, we don't make any assumptions about alien life. You should see some of the critters we've captured! Uh-oh, I've said too much, gotta go...
Careful about asking for an eval copy from Hummingbird. The web form doesn't mention that they won't send you the CD until they've had a telesales operator call you to confirm the details. After that, someone from a reseller will call to try to get you to buy. Then, a salesperson from Hummingbird will call. I ended up getting four unwanted calls from them, even though I told the first one I didn't want any more calls, and emailed the sales rep the same thing. If you screen all your calls, or don't mind being endlessly bugged by clueless sales people, go ahead and order the eval.
> But wouldn't doing this make your US > citizenship invalid? No, the U.S. allows its citizens to hold dual citizenship. Plenty of U.S. citizens do this. If you have an Irish grandfather, for example, you can apply for and more or less automatically receive Irish citizenship, which you can hold in addition to your U.S. citizenship.
I didn't notice this mentioned in earlier comments: here's a U.S. company that sells the Netsurfer chairs:
http://www.officeorganix.com/netsurf1.htm
The Classic goes for $2600 and the Lite for $2365.
They look cool an' all, but I don't think they're at all adjustable??!! A lot of money to pay for a totally fixed-position chair!
That was the point I was trying to make, but I probably stated it ambiguously.
but for any given nondeterministic step it handles some finite number of inputs
True, but I was responding to this:
Isn't "going through a whole buttload of computations in parallel" exactly what non-deterministic machines do?
This is itself a naive, or at the very least incomplete, view. When assessing the capabilities of another intelligence, there are philosophical limits to the degree to which we can determine it by such "hard" measurements as, say, brain mass, number of neurons, etc.
The limitation here is not necessarily in the perceiver, but rather in what it is theoretically possible to perceive, and at what level. [For example, you probably won't be able to tell that I'm imagining a pink penguin just by sticking a probe in my brain. And if you stick enough probes in my brain, I'll stop thinking about the pink penguin and start wondering "what are all these probes doing in my brain?" instead, which would spoil the experiment.]
There are good arguments which say that intelligence cannot be assessed by hard physical measurements, and that the only way to assess it is by interacting with the intelligence itself, not with the underlying substrate. You have to ask it questions and see how it responds. This is the basic reason behind the Turing Test.
We do this all the time when we meet new people - while we might make initial assessments based on superficial characteristics such as appearance, assuming we're open-minded, ultimately we're left assessing the person based on what they say or do, and on how they respond to us.
Assessing an intelligent computer will work the same way. If the computer is capable of holding its own in a conversation, or responds well to an intelligence test for which it has not been specifically prepared, or is able to self-direct its activities the way humans do, then it would qualify as intelligent.
And if you disagree with that, watch out, because next thing you know the computers will be marching (or wheeling) through the streets, demanding equal rights and an end to discrimination against silicon-based lifeforms!
Schneier's quote talks about trying *all* guesses in parallel. Presumably the number of guesses such a system can handle is unbounded. However, going through a "whole buttload" of computations in parallel is still deterministic, it would just be "n" times faster than a serial solution, where n equals the size of a buttload.
Is Jon Katz an artificial intelligence? It took me way too long to work it out, but all the signs are there. The liberal use of the latest buzzphrases, to hide lack of actual content. The slightly off-base parotting of well-known concepts, as though he doesn't quite get it.
It also explains that interminable "Road to Linux" series which never manages to achieve actual use of Linux. Since the Katz software is presumably running on Linux, for Katz, writing about Linux is like exploring one's own subconscious - a task which can only be done incompletely, at best.
Hats off to CmdrTaco for an awesome piece of coding! Is it written in PERL? What hardware does Katz run on? (Hmmm, wonder if that matchbox-sized computer has anything to do with this?)
"it seems that we'd get farther by concentrating on advancing 'traditional' symbolic artificial intelligence, rather than simulating huge neural networks on puny serial hardware... "
Traditional symbolic AI is unlikely to ever achieve much than the traditional applications, e.g. grammar checking, grading essays, playing chess.
You're right about the advantages of the massive parallelism of neural networks. So, we have to build *real* neural networks in hardware (or wetware?), not simulate them on serial hardware (which as you point out is doomed)
"there's the small issue of neural networks being 'opaque' to the creator - all's good when they work, but when they break it's difficult to figure out why."
That's what psychologists are for...
"You can't possibly ask Katz to read math books."
LOL!
You've hit the nail on the head about Katz. To report on a subject in a way that adds value, you need a pretty good understanding of the subject - unless you're writing for an audience with an even more limited understanding. Some rudimentary analytical skills also don't hurt (such as, say, the ability to distinguish between problems caused by UPS during shipping, vs. the intent to install Linux on a machine...)
For an example of a writer adding value to a subject (while simultaneously doing a bit of a snow job), check out the article by Michael Lewis (author of "Liar's Poker") in last weekend's New York Times Magazine, about the near-demise and bailout of Long Term Capital Management.
Of course, Lewis probably gets paid more than Katz, but if so there's a reason for that...
How about a limited menu of choices for anonymous user names, e.g.:
Anonymous Coward
Corporate Coward (for posters of sensitive info)
Anonymous OS Bigot (or *nix Bigot, Windows Bigot, Mac Bigot)
Anonymous Foul-Mouthed Jerk
Anonymous Troll
All intended tongue in cheek, of course, hopefully you get the idea. Who knows, some posters might actually choose appropriate categories.
Oh please. Yes, let's all ask /. to withhold information from us if it might be the least bit controversial. God knows /.ers wouldn't be able to handle such a thing!