I'm still hoping someone will re-think HyperCard as an Internet-optimised operating system, with integrated scriptable modules for creating and viewing webpages, images, email, multimedia, etc.
If it runs as slick as HyperCard, it should become the new basic minimum of computer-literacy, so a creative community would inevitably grow up around it.
Build it on top of Linux and offer it for Internet Appliances, and it could put Microsoft out of business. But wireless and web-services and multiplayer gaming don't seem central to me, at all.
The biggest, world-transforming programming
breakthru will be when we start to teach
computers to behave like people-- which will
be a major step towards understanding who we are.
"The Sims" is a very crude glimpse of this, but
there's almost no 'psychology' in The Sims, and
the sorts of science you learn in psychology
class are almost entirely useless for this
purpose.
So one extreme 'fringe' involves wrestling with
the literary side of behavior, trying to
analyse and classify the real behaviors people
do. My whole website
is devoted to this, including a minimalist
notation-scheme for story-skeletons, an exhaustive analysis of the psychology of romantic love via quotes from love poems (which I view as preliminary research
for a computer game about love), and most
extreme-fringe-y of all, an analysis of James Joyce's "Finnegans Wake" as a deep, systematic model of human emotions.
I read the science headlines every day at NewsHub and it's distressing to see how fake a lot of this 'news' is-- if a scientist craves publicity, there are certain themes they can appeal to that the press seems to get excited about for no good scientific reason (the oldest anything, black holes, asteroid impacts, etc).
I'd be interested to know exactly how different this 'really really complete' genome is from the fakety-fake 'complete' genome they announced a few years back.
The rumor then was that it was the egomaniac Venter's own DNA they were using, so calling it 'the' human genome is another big lie-- one of the most interesting uses for the data is to cross-compare different ethnicities (and different species) and use this to reconstruct the human family tree. So the fact that one person's genome is the first to be sequenced will quickly become insignificant to the overall picture.
Edward Tufte is a demigod in the world of information-design, and he made an interesting case recently that bad PowerPoint design in Boeing's report contributed to the misinterpretation of the analysis. Eg, the way the ppt-slide was laid out almost completely concealed the fact that the test was on a small cube of foam.
I thought the neurosis resulted from HAL being unwilling to admit he made a mistake...?
Re:First, human self-knowledge
on
AI in Sci-Fi
·
· Score: 1
already an AI subfield: genetic algorithms
Yeah... I don't begrudge them the right to try that route. I just think the path of self-knowledge is more direct and logical, and I think it's bizarre that it's discussed so little.
Re:First, human self-knowledge
on
AI in Sci-Fi
·
· Score: 1
"The Sims" is a simulation. There's no self-awareness there.
Was it Descartes who started the whole dualistic-consciousness obsession? I think it's irrelevant-- as you build better Sims, you learn about yourself, and your AI gets closer and closer to realism. Eventually it will seem self-aware, which is all I care about.
Re:First, human self-knowledge
on
AI in Sci-Fi
·
· Score: 1
people have been breeding livestock and plants without understanding of the underlying genetics.
So what's the research-program you're suggesting? Try random combinations of circuitry, and when one shows interesting behavior, duplicate and modify it many times?
I just think that's wishful thinking-- why is the prospect of seriously pursuing self-understanding so intimidating???
Re:First, human self-knowledge
on
AI in Sci-Fi
·
· Score: 1
A very loose-- okay, absurdly loose, valid only in the metaphorical sense-- interpretation of Godel's Theorem would imply that this may be impossible.
But "The Sims" exists, and for all its flaws it's damn impressive! So how much better can it get before your metaphorical-Goedel catch-22 kicks in?
First, human self-knowledge
on
AI in Sci-Fi
·
· Score: 3, Insightful
What's missing in all the sci-fi scenarios is the necessity, before an AI can be built, that humans first understand themselves.
This means that psychology will have to be able to really model human behavior, even (especially!) in the game-like sense that Will Wright's "The Sims" tries to do.
But this will mean we have to learn to detach from our desires enough to view them objectively, and see how they interact-- which is a spiritual practice as much as a scientific one... and also a literary practice, because novelists have been trying to portray human motives objectively for several centuries.
I've been wrestling with these issues for thirty years, and my website is almost entirely devoted to the problem. In particular, see my AI faq and most recently my illustrated 400k timeline of knowledge representation, in the broadest sense of that term.
I think the point is that if you save to their XML specification, you will lose all your document formatting.
I think the root of the confusion goes back to Golfarb's original theory for SGML-- that the styles in a document are secondary to the structures, and should be kept separate.
This has been a religious conviction ever since, despite the fact that most authors are messy and intuitive, and SGML-etc are very, very rigid and unintuitive. The rationalisation is that messy authors can just represent their styles using 'fake' (ad hoc) XML, but if this turns out to be 90% of the real users of MS Office, then I think MS could indeed save valid XML, but it won't be portable in any useful sense.
There are external reasons why poetry has declined in
popularity. Among them: the availability and quality of
artificial light. The natural change of pace and mood that the
waning of the day brought to every generation of humanity
before the twentieth century is gone. With it went a whole
oral tradition of legend, storytelling, and poetry.
I haven't run across this idea before-- is it original?
I can't really imagine a serious poem about tech unless it's anti-tech. For light verse, John Updike no doubt has some things, and there's Nabokov's "The Refrigerator Awakes" [RealAudio]. (In the realm of song lyrics, They Might Be Giants is another likely source.)
Paul Durcan's "Christmas Day" (not online) has a comment that could be Slashdot's motto:
Why do computer programmers always answer
When asked in questionnaires
In Sunday newspapers
What is your idea of Heaven? -
Snorkelling in Acapulco.
Pope Leo XIII wrote a Latin piece on photography in 1867:
[translation]
O miracle of human thought,
O art with newest marvels fraught...
You need a very different set of tests for young sciences, eg the social sciences. My favorite tests:
1. Whether the Usenet newsgroups for that science have healthy ongoing debate (bogus sciences like AI have empty newsgroups with occasional announcements of conferences , or references to literature that isn't even online)
2. How many websites for that science make an effort to explain the basics to average websurfers (AI is dismal for this)
What do you supposedly mean by "symmetry"? Snowflakes are not perfectly symmetric. Anyway, dendrite formation simulation (at the macroscopic scale) is a popular subject in numerical methods. Here's a quick link off google. [dmawww.epfl.ch]
The level of symmetry in snowflakes is infitely greater than that demo. All the little filigrees echo each other, and the 'mistakes' are attributable to fractures.
Molecular vibrational levels are Boltzmann-distributed at equillibrium
That's only for a random fluid, not a crystal in freefall.
James Joyce's Ulysses is supposed to offer a full day's thoughts by Leopold Bloom. I did some calculations and concluded that the overall size of 1.5Mb is about right... so a full lifetime of thoughts is just 37 gigabytes.
If the Library of Congress is 10 terabytes that's less than 300 lifetimes' worth. (Which 300 people should be included?)
Another useful measure is the EB, or Encyclopaedia Britannica, which is about 200Mb. So one LoC = 50,000 EBs = 300 lifetimes.
Because each arm experiences the same conditions, the arms tend to look alike, producing large-scale, intricate, six-fold symmetric snow crystals.
This explanation is obviously handwaving-- the symmetry is perfect (or close to it) over scales of millions of molecules.
I've been arguing since 1980 or so that an ice crystal in freefall is not at absolute zero (obviously) so it must have internal vibrations. This is basically 'noise', but as it echoes thru the ice, it stops looking random and becomes symmetrical, like Chladni patterns on a vibrating plate or drumhead. (Or like the radiating circles from a drip of water into a circular pool, reconverging at an opposite point.) Because these symmetries are present from the first stage of growth, they maintain symmetrical growth.
I don't think the 104.5 degree angle between the hydrogens in water molecules is close enough to 120 to deliver perfect hexagonality-- it's probably due to the geometry of echoes in any disk, because hexagons can be inscribed in circles. (The spinning of the seed probably contributes to the flatness-- growing favors the outside edge of the bulge, otherwise it might be more spherical.)
I don't suppose you could summarise how Loebner changed the rules after that?
My memory is fuzzy because there must have been a couple of rule-changes, but Whalen was obviously allowed to limit the domain of questions, and post-Whalen all entries had to handle every/any domain... I think. (The other rule-change was that to win the big prize they also had to recognise ascii-art-style representations.)
It's a paradigm shift-- instead of looking for complicated 'solutions' that will enhance their status, Whalen took a fresh look at the problem and found a way to deliver useful results with no particularly fancy algorithms.
It's nothing that anyone couldn't already add to a system that needs it.
No one had at the time, and few are even aware of the idea now.
Check out TRECK
The dismal website design shows how little they appreciate Whalen's insight-- I clicked four different links on the homepage and ended up with ZERO examples of their work. This is absolutely typical of academic-AI websites-- a whole lot of self-congratulation and almost no effort to communicate. (Contrast that with any healthy science, where tutorials aimed at beginners are a dime a dozen.)
To say AI has made no advances because we can't fool people into thinking they're talking to someone
Those words are yours. Academic AI has made a few minor advances, but continues to project itself as possessing arcane, complex secrets that deserve big paychecks.
(Regarding Whalen's Loebner-winner) Well, wasn't that then just a simple, 'one-pass' expert system?
Not an expert-system in any way (those involve a knowledgebase of logical rules). Whalen said he'd gone further than simple keyword-matching, but I never found out how.
It is nothing new although perhaps the only really useful thing that AI research came up with.
The design was new, and clever, and useful.
(And, of course, it has little to do with REAL AI:)
I hope that smiley means you're joking, because that's what the academics claimed, but their arguments were purely self-serving.
Slashdot Mirror
I'm still hoping someone will re-think HyperCard as an Internet-optimised operating system, with integrated scriptable modules for creating and viewing webpages, images, email, multimedia, etc.
If it runs as slick as HyperCard, it should become the new basic minimum of computer-literacy, so a creative community would inevitably grow up around it.
Build it on top of Linux and offer it for Internet Appliances, and it could put Microsoft out of business. But wireless and web-services and multiplayer gaming don't seem central to me, at all.
"The Sims" is a very crude glimpse of this, but there's almost no 'psychology' in The Sims, and the sorts of science you learn in psychology class are almost entirely useless for this purpose.
So one extreme 'fringe' involves wrestling with the literary side of behavior, trying to analyse and classify the real behaviors people do. My whole website is devoted to this, including a minimalist notation-scheme for story-skeletons, an exhaustive analysis of the psychology of romantic love via quotes from love poems (which I view as preliminary research for a computer game about love), and most extreme-fringe-y of all, an analysis of James Joyce's "Finnegans Wake" as a deep, systematic model of human emotions.
The timeline covers all domains, not just medical, but there's lots of related sims, and medical models going back to 800 BC.
I'd be interested to know exactly how different this 'really really complete' genome is from the fakety-fake 'complete' genome they announced a few years back.
The rumor then was that it was the egomaniac Venter's own DNA they were using, so calling it 'the' human genome is another big lie-- one of the most interesting uses for the data is to cross-compare different ethnicities (and different species) and use this to reconstruct the human family tree. So the fact that one person's genome is the first to be sequenced will quickly become insignificant to the overall picture.
Edward Tufte is a demigod in the world of information-design, and he made an interesting case recently that bad PowerPoint design in Boeing's report contributed to the misinterpretation of the analysis. Eg, the way the ppt-slide was laid out almost completely concealed the fact that the test was on a small cube of foam.
I thought the neurosis resulted from HAL being unwilling to admit he made a mistake...?
Yeah... I don't begrudge them the right to try that route. I just think the path of self-knowledge is more direct and logical, and I think it's bizarre that it's discussed so little.
Was it Descartes who started the whole dualistic-consciousness obsession? I think it's irrelevant-- as you build better Sims, you learn about yourself, and your AI gets closer and closer to realism. Eventually it will seem self-aware, which is all I care about.
So what's the research-program you're suggesting? Try random combinations of circuitry, and when one shows interesting behavior, duplicate and modify it many times?
I just think that's wishful thinking-- why is the prospect of seriously pursuing self-understanding so intimidating???
But "The Sims" exists, and for all its flaws it's damn impressive! So how much better can it get before your metaphorical-Goedel catch-22 kicks in?
This means that psychology will have to be able to really model human behavior, even (especially!) in the game-like sense that Will Wright's "The Sims" tries to do.
But this will mean we have to learn to detach from our desires enough to view them objectively, and see how they interact-- which is a spiritual practice as much as a scientific one... and also a literary practice, because novelists have been trying to portray human motives objectively for several centuries.
I've been wrestling with these issues for thirty years, and my website is almost entirely devoted to the problem. In particular, see my AI faq and most recently my illustrated 400k timeline of knowledge representation, in the broadest sense of that term.
These were useful during the Afghan invasion: Lebanon, Jordan, Arab News, Gulf News
I think the root of the confusion goes back to Golfarb's original theory for SGML-- that the styles in a document are secondary to the structures, and should be kept separate.
This has been a religious conviction ever since, despite the fact that most authors are messy and intuitive, and SGML-etc are very, very rigid and unintuitive. The rationalisation is that messy authors can just represent their styles using 'fake' (ad hoc) XML, but if this turns out to be 90% of the real users of MS Office, then I think MS could indeed save valid XML, but it won't be portable in any useful sense.
I haven't run across this idea before-- is it original?
Paul Durcan's "Christmas Day" (not online) has a comment that could be Slashdot's motto:
Pope Leo XIII wrote a Latin piece on photography in 1867: [translation]
Some gleanings from my weblog: landing-gear crisis, Chuck-E-Cheese, auto repair
Poetry heals too-- you might try Solace.
You need a very different set of tests for young sciences, eg the social sciences. My favorite tests:
1. Whether the Usenet newsgroups for that science have healthy ongoing debate (bogus sciences like AI have empty newsgroups with occasional announcements of conferences , or references to literature that isn't even online)
2. How many websites for that science make an effort to explain the basics to average websurfers (AI is dismal for this)
I looked it up for my simulations timeline-- it was for Doom II, created c1996: Wired article.
The level of symmetry in snowflakes is infitely greater than that demo. All the little filigrees echo each other, and the 'mistakes' are attributable to fractures.
Molecular vibrational levels are Boltzmann-distributed at equillibrium
That's only for a random fluid, not a crystal in freefall.
(Why do geeks so often get defensive when you point out gaps in science's perfection?)
Crystal growth and dendrite-formation are well-understood subjects within physical chemistry.
Not symmetries that are maintained over distances of a millimeter or more.
Suggesting that all water molecules in a snowflake crystal vibrate in harmony in a state of equillibrium violates the laws of thermodynamics.
Which law is that? Conservation of Chaos? ;^/
If the Library of Congress is 10 terabytes that's less than 300 lifetimes' worth. (Which 300 people should be included?)
Another useful measure is the EB, or Encyclopaedia Britannica, which is about 200Mb. So one LoC = 50,000 EBs = 300 lifetimes.
This explanation is obviously handwaving-- the symmetry is perfect (or close to it) over scales of millions of molecules.
I've been arguing since 1980 or so that an ice crystal in freefall is not at absolute zero (obviously) so it must have internal vibrations. This is basically 'noise', but as it echoes thru the ice, it stops looking random and becomes symmetrical, like Chladni patterns on a vibrating plate or drumhead. (Or like the radiating circles from a drip of water into a circular pool, reconverging at an opposite point.) Because these symmetries are present from the first stage of growth, they maintain symmetrical growth.
I don't think the 104.5 degree angle between the hydrogens in water molecules is close enough to 120 to deliver perfect hexagonality-- it's probably due to the geometry of echoes in any disk, because hexagons can be inscribed in circles. (The spinning of the seed probably contributes to the flatness-- growing favors the outside edge of the bulge, otherwise it might be more spherical.)
My memory is fuzzy because there must have been a couple of rule-changes, but Whalen was obviously allowed to limit the domain of questions, and post-Whalen all entries had to handle every/any domain... I think. (The other rule-change was that to win the big prize they also had to recognise ascii-art-style representations.)
It's a paradigm shift-- instead of looking for complicated 'solutions' that will enhance their status, Whalen took a fresh look at the problem and found a way to deliver useful results with no particularly fancy algorithms.
It's nothing that anyone couldn't already add to a system that needs it.
No one had at the time, and few are even aware of the idea now.
Check out TRECK
The dismal website design shows how little they appreciate Whalen's insight-- I clicked four different links on the homepage and ended up with ZERO examples of their work. This is absolutely typical of academic-AI websites-- a whole lot of self-congratulation and almost no effort to communicate. (Contrast that with any healthy science, where tutorials aimed at beginners are a dime a dozen.)
To say AI has made no advances because we can't fool people into thinking they're talking to someone
Those words are yours. Academic AI has made a few minor advances, but continues to project itself as possessing arcane, complex secrets that deserve big paychecks.
Not an expert-system in any way (those involve a knowledgebase of logical rules). Whalen said he'd gone further than simple keyword-matching, but I never found out how.
It is nothing new although perhaps the only really useful thing that AI research came up with.
The design was new, and clever, and useful.
(And, of course, it has little to do with REAL AI :)
I hope that smiley means you're joking, because that's what the academics claimed, but their arguments were purely self-serving.