Sorry, I should have added, "Be sure to visit the Division of Paleontology link at the bottom of that site, for a nice collection of close-ups of the fossil."
the story of the RIAA's threat against academic researhers is all over the mainstream media.
Brilliant move, RIAA. What is you SDMI worth now? Where are the anti-DMCA crowd going to turn for PR, and what are they going to mention the next time we have congressional hearings or a court case involving the DMCA? And which side of the fence do you think any remaining waverers are going to come down on?
For semi-adventurous Red Hatters, I notice that there are lots of new GNOME RPMs at ftp://rawhide.redhat.com/, though you're on your own at getting everything installed correctly on your system.
One hint would be to look at the GNOME site to see what components/versions of various things you need to fetch.
> I work with a bunch of guys, that are as far from the traditional geeks as they come.... What they do, is write quality code, develop innovative hardware, and usually do it under budget and ahead of schedule.
I congratulate you, and envy you too. In my experience, professionalism is even rarer than joie de hack is. Over half the IT people I've ever worked around lacked both.
I'll grant that professionalism is more productive than joie de hack, but the latter still tends to be much better than neither. Too many people in IT are clueless, unmotivated timeservers waiting for the next paycheck to show up.
> But it really would be good to see them join together
I figure that in 1 to 3 years we'll have a Third Choice free desktop competitor starting up, with the explicit goal of learning from KDE's/GNOME's mistakes.
> when the electromagnets are fired in a decelerating sequence around the head, gives the wearer the sensation that time is slowing down quite drastically.
Heh heh. In the last episode of Futurama they dug up an old VW "bus" and started using it. When DrZ first got in, he asked "Where's the device that speeds up and slows down time?" Fry replied that it was under the seat, and fetched it out. They only gave a brief glimpse, but I believe it was a water pipe.
> One can't help but notice the overwhelming bias against Intel and towards AMD on slashdot, to the point of zealousness
Yep, some of use are the kind of assholes who expect a product that costs twice as much to actually be better.
> i.e. only noting the benchmarks where AMD is ahead but pretending not to see the benchmarks where Intel is ahead
It's duly noted that the P4 creams the T-bird on Q3 Arena. If I ever set up a dedicated Q3 Arena box, I'll keep that in mind. But I'll also keep that 2x price for the processor and memory in mind, so price:performance might keep me from buying a P4 even for for my dedicated Q3A box (in the unlikely event I ever build one).
> ut Intel will have much higher clock rates in the same price range, and ultimately, equivalent and/or better performance, particularly over the next 1 to 3 years, as Intel puts out CPUs with humongous clock rates
Good for them! Maybe I'll buy one in 1 to 3 years, if your prediction pans out.
More likely I'll be running a 64-bit AMD machine with several gigs of cheap open-architecture memory, though.
OK, he was looking for the neural correlate of religious experience, he found something, and he announced success.
He should be the first person to exercise skepticism toward his own findings. Why not run the experiment on people practicing TM, or absorbed in a programming problem, or getting laid, or any other non-religious activity that brings about extreme focus in the brain, and see whether you get the same effect?
It's way too easy to find some general phenomenon and think you've found the specific phenomenon you've been looking for. That's the danger of focussing your career on a search for something that you "know" must exist.
Maybe he's right, but I'll hold out for a second opinion.
> The only ones that are applicable for me are a) constant computing (folding@home) and b) gaming; both areas where the P4 excels.
Actually, several of the review sites are showing that the AMD beats the Intel in about half the gaming tests.
And that at about half the price, both for processor and for memory, based on what's currently listed on pricewatch.com.
I'm also curious about folding@home. Does it run through a lot of memory? The review sites left me with the impression that the P4 gets its advantage -- when it does get one -- from the bandwidth performance of its horribly expensive RDRAM.
If folding doesn't use a lot of memory, then the Athlon might actually win. (Either way, it probably wins at performance/price by a large margin.)
If you happen to have access to both kinds of machine, I'm sure lots of people here would be interested in a folding benchmark, though.
> The amount of human time required to develop and debug a proper AI, one that makes a significant use of computational resources, is enormous.
IMO, machine learning (ML) is the way to solve this.
And that's what Evans is doing here:
When your monster does something good (or at least something that you want it to keep doing), your Divine Hand literally strokes it; when it does something incorrectly, the same Hand of God smacks it. Eventually -- ideally, anyway -- it grows into an active extension of your will...
That's aka Reinforcement Learning. For decision trees, the feedback is the "evidence" that the tree has to explain, so presumably his system saves some/all of the feedback and intermittently updates the decision tree. If you give consistent feedback, it should converge to a point where your monster can guess what the outcome of an action is, and thereby avoid the smacks. As a side effect, it looks like it "knows" what you want. Similarly for perceptrons / neural networks.
The bit about Moore's Law is certainly apropos. I recently ran a genetic algorithm program that searches for good solutions for the travelling salesman problem, and on a late model x86 desktop system the program was evaluating 1000 candidate solutions a second for a 2000 city problem. Our resource-intensive GUI desktops obscure just how fast our desktop supercomputers really are.
Also, contrary to what someone suggested in another thread, games are not the state of the art for AI. You can easily find tons of papers on this kind of stuff with your favorite search engine, and in some cases download the code for the program described in the paper.
That's not to knock it; games will probably be AI's killer app.
> But your comment on Guile got me thinking; probably it is overkill. But perhaps the flexibility of a full programming language would be beneficial for configuration, although it may not be meant to solve the same problems as XML.
> Your post got me to (re-)look into Guile, but I was wondering if you (or anyone) had any more specific thoughts on what formats to use for configuration files, and what in particular you do with Guile that replaces what you would have done with XML.
I'm not a guru at either Guile or XML, and my use of Guile is evolving pretty quickly now that I have started using it regularly. For now, this is how I use it for configuration files.
I have a "data type" that I call a table, which is of the format (key data). The data can be more tables, giving a tree structure to the configuration, or it can be bottom-level data, serving as the leaves in a tree of data. So a simple example of a configuration file for a fictitious game would be:
[Sorry; I had to remove the indentation to get it past Rob's lame-o lameness filter. Lack of indentation really reduces readability.]
(configuration
(difficulty-level 7)
(sides
(good
(description "The Good Guys")
(restrictions no-nukes no-poison)
)
(bad
(description "The Bad Guys")
(restrictions none)
)
(ugly
(description "The Ugly Guys")
(restrictions no-teeth)
)
)
In the "tree" metaphore, configuration is the root, difficulty-level, sides, and graphics are the first level of branches, etc., on down to the leaves where the sub-tables terminate in atomic data.
For simplicity, the following uses pseudocode rather than the actual Guile syntax.
You declare appropriate variables of the Guile SCM type and then use Guile's read to load the configuration file into your program as a Scheme object (without trying to evaluate it as a Scheme expression).
conf=read("~/.mygame/myconfig.scm")
Since you are using tables, you use Guile to define a function lookup(keyname,table) that converts the key-name string into a Guile symbol, and then looks it up in table table:
grap=lookup("graphics",conf) speeds=lookup("animation-speed",grap) tmp=lookup("chase-scenes",speeds) ...do whatever with the value...
tmp=lookup("love-scenes",speeds) ...do whatever with the value...
Your program just runs down the tree like that, looking for whatever data it needs. When you get to the bottom, you use a Guile built-in to convert the data to an integer, string, or whatever your program expects.
Some data can be iterative, too. In the example, sides is a list that you can modify in your config if you want to define more player sides (say, for AI opponents). Your configuration reader just uses your lookup to find sides, and then iteratively loads one side record at a time until you run out of definitions. (In Scheme terms, you process the car with your lookup function, throw away the car, and continue your iteraton on the cdr.)
The lookup function is really easy to define, and you put it in your library directory so all your programs can use it. It just converts the keyname string to a Guile symbol, and then uses the built-in assoc to find that symbol as the key for anything in the cdr of table. If all your data is in the table format, it works to look up anything, working down the tree recursively. For instance, if you wanted the y size and didn't need anything else, you could do:
You can also easily define a recursive check_table that verifies that something you loaded is in fact a table structure, in order to trap errors early if a user has screwed up his config file.
The only things I don't use lookup for are the iteration as described above (but even then I use lookup to find the iterative definition, and then use it again to parse each element in the iterative structure), and to get the bottom level data out of the "leaves", e.g. to parse:
(y 490)
I have a library function get_int that accepts a leaf of the form (key integer), extracts the second element, and converts it from a Scheme integer to a C integer, and similarly for other atomic data types.
Also nice, Guile does garbage collection, so you can use it to splice things together out of the configuration and throw away the husks without having to explicitly collect all the trees of objects that you created.
> I'm sorry to go off on such a rant, but I am SO tired of everything being done in an XML format.
FWIW (not much, really), I did exactly one hobby project in XML. Yech.
I switched to Guile immediately afterward. You can use it just like a ML if that's all you want to do, and you'll find it way easy to parse. As an added bonus, you can embed code in your data/documents/stylesheets if that's what you want. (Watchout for viruses, though.)
> Here's a picture of Cassiopa and Sheba (from Battlestar Galactica)
> Here's a picture of Athena (from Battlestar Galactica)
> Here's a picture of colonel Wilma Deering (from Buck Rogers)
Your catalogue sure has a lot of nice stuff in it, but I don't see the prices anywhere. How much for three (3) of the brunette in white and one (1) of the older blonde?
> Compare the Buck Rogers one with this photo of Natalie (on the right)
Ah, thanks for pointing her out. I thought the one on the left was Natalie, and I couldn't understand why the kiddies were in such a froth over her!
I'm no longer a True Fan after E1, but I'm eagerly awaiting E2 anyway, 'cause I heard that it will include a graphic depiction of the slow dismemberment of Jar Jar Binks.
Slashdot Mirror
Sorry, I should have added, "Be sure to visit the Division of Paleontology link at the bottom of that site, for a nice collection of close-ups of the fossil."
--
You can see a couple of photos of the fossil and a drawing of the reconstructed critter at the American Museum of Natural History's Website.
Kinda cute, doncha think? Probably worse than a cat about tearing up the furniture, though.
--
- it has already permeated the internet, and...
- the story of the RIAA's threat against academic researhers is all over the mainstream media.
Brilliant move, RIAA. What is you SDMI worth now? Where are the anti-DMCA crowd going to turn for PR, and what are they going to mention the next time we have congressional hearings or a court case involving the DMCA? And which side of the fence do you think any remaining waverers are going to come down on?--
For semi-adventurous Red Hatters, I notice that there are lots of new GNOME RPMs at ftp://rawhide.redhat.com/, though you're on your own at getting everything installed correctly on your system.
One hint would be to look at the GNOME site to see what components/versions of various things you need to fetch.
Use at your own risk.
--
> I work with a bunch of guys, that are as far from the traditional geeks as they come. ... What they do, is write quality code, develop innovative hardware, and usually do it under budget and ahead of schedule.
I congratulate you, and envy you too. In my experience, professionalism is even rarer than joie de hack is. Over half the IT people I've ever worked around lacked both.
I'll grant that professionalism is more productive than joie de hack, but the latter still tends to be much better than neither. Too many people in IT are clueless, unmotivated timeservers waiting for the next paycheck to show up.
--
> But it really would be good to see them join together
I figure that in 1 to 3 years we'll have a Third Choice free desktop competitor starting up, with the explicit goal of learning from KDE's/GNOME's mistakes.
--
> the truth is if Intel shipped a computer with a heatsink that doubled as a Grillmaster, I would buy it.
Problem is, it would only cook meat that you bought from the RAMSTEAK consortium.
--
> I always wished that typedefs created new types, instead of behaving like wimpy macros.
Maybe you should consider Ada, which does this and most of the other things people are wishing for everywhere in the responses to this article.
Don't let ESR's lame and factually incorrect entry in his jargon file put you off.
--
> The fact that most scientists...
I thought the big difference was that scientists need power cords to get their work done, whereas musicians need both power cords and power chords.
--
> when the electromagnets are fired in a decelerating sequence around the head, gives the wearer the sensation that time is slowing down quite drastically.
Heh heh. In the last episode of Futurama they dug up an old VW "bus" and started using it. When DrZ first got in, he asked "Where's the device that speeds up and slows down time?" Fry replied that it was under the seat, and fetched it out. They only gave a brief glimpse, but I believe it was a water pipe.
--
> One can't help but notice the overwhelming bias against Intel and towards AMD on slashdot, to the point of zealousness
Yep, some of use are the kind of assholes who expect a product that costs twice as much to actually be better.
> i.e. only noting the benchmarks where AMD is ahead but pretending not to see the benchmarks where Intel is ahead
It's duly noted that the P4 creams the T-bird on Q3 Arena. If I ever set up a dedicated Q3 Arena box, I'll keep that in mind. But I'll also keep that 2x price for the processor and memory in mind, so price:performance might keep me from buying a P4 even for for my dedicated Q3A box (in the unlikely event I ever build one).
> ut Intel will have much higher clock rates in the same price range, and ultimately, equivalent and/or better performance, particularly over the next 1 to 3 years, as Intel puts out CPUs with humongous clock rates
Good for them! Maybe I'll buy one in 1 to 3 years, if your prediction pans out.
More likely I'll be running a 64-bit AMD machine with several gigs of cheap open-architecture memory, though.
--
OK, he was looking for the neural correlate of religious experience, he found something, and he announced success.
He should be the first person to exercise skepticism toward his own findings. Why not run the experiment on people practicing TM, or absorbed in a programming problem, or getting laid, or any other non-religious activity that brings about extreme focus in the brain, and see whether you get the same effect?
It's way too easy to find some general phenomenon and think you've found the specific phenomenon you've been looking for. That's the danger of focussing your career on a search for something that you "know" must exist.
Maybe he's right, but I'll hold out for a second opinion.
--
No, nuns dream of ithyphalic incubi.
Rural priests sometimes dream of sheep, but not electric ones.
--
> The only ones that are applicable for me are a) constant computing (folding@home) and b) gaming; both areas where the P4 excels.
Actually, several of the review sites are showing that the AMD beats the Intel in about half the gaming tests.
And that at about half the price, both for processor and for memory, based on what's currently listed on pricewatch.com.
I'm also curious about folding@home. Does it run through a lot of memory? The review sites left me with the impression that the P4 gets its advantage -- when it does get one -- from the bandwidth performance of its horribly expensive RDRAM.
If folding doesn't use a lot of memory, then the Athlon might actually win. (Either way, it probably wins at performance/price by a large margin.)
If you happen to have access to both kinds of machine, I'm sure lots of people here would be interested in a folding benchmark, though.
--
> This is very cool, and even though it's been around for a while, we don't seem to have run it before.
Since when did that become an issue?
--
IMO, machine learning (ML) is the way to solve this.
And that's what Evans is doing here:That's aka Reinforcement Learning. For decision trees, the feedback is the "evidence" that the tree has to explain, so presumably his system saves some/all of the feedback and intermittently updates the decision tree. If you give consistent feedback, it should converge to a point where your monster can guess what the outcome of an action is, and thereby avoid the smacks. As a side effect, it looks like it "knows" what you want. Similarly for perceptrons / neural networks.
The bit about Moore's Law is certainly apropos. I recently ran a genetic algorithm program that searches for good solutions for the travelling salesman problem, and on a late model x86 desktop system the program was evaluating 1000 candidate solutions a second for a 2000 city problem. Our resource-intensive GUI desktops obscure just how fast our desktop supercomputers really are.
Also, contrary to what someone suggested in another thread, games are not the state of the art for AI. You can easily find tons of papers on this kind of stuff with your favorite search engine, and in some cases download the code for the program described in the paper.
That's not to knock it; games will probably be AI's killer app.
--
> But your comment on Guile got me thinking; probably it is overkill. But perhaps the flexibility of a full programming language would be beneficial for configuration, although it may not be meant to solve the same problems as XML.
f ))));
> Your post got me to (re-)look into Guile, but I was wondering if you (or anyone) had any more specific thoughts on what formats to use for configuration files, and what in particular you do with Guile that replaces what you would have done with XML.
I'm not a guru at either Guile or XML, and my use of Guile is evolving pretty quickly now that I have started using it regularly. For now, this is how I use it for configuration files.
I have a "data type" that I call a table, which is of the format (key data). The data can be more tables, giving a tree structure to the configuration, or it can be bottom-level data, serving as the leaves in a tree of data. So a simple example of a configuration file for a fictitious game would be:
[Sorry; I had to remove the indentation to get it past Rob's lame-o lameness filter. Lack of indentation really reduces readability.]
(configuration
(difficulty-level 7)
(sides
(good
(description "The Good Guys")
(restrictions no-nukes no-poison)
)
(bad
(description "The Bad Guys")
(restrictions none)
)
(ugly
(description "The Ugly Guys")
(restrictions no-teeth)
)
)
(graphics
(size (x 550) (y 490))
(theme penguins)
(images "mytiles.xpm")
(animation-speed
(chase-scenes 12)
(love-scenes 3)
)
)
)
In the "tree" metaphore, configuration is the root, difficulty-level, sides, and graphics are the first level of branches, etc., on down to the leaves where the sub-tables terminate in atomic data.
For simplicity, the following uses pseudocode rather than the actual Guile syntax.
You declare appropriate variables of the Guile SCM type and then use Guile's read to load the configuration file into your program as a Scheme object (without trying to evaluate it as a Scheme expression).
conf=read("~/.mygame/myconfig.scm")
Since you are using tables, you use Guile to define a function lookup(keyname,table) that converts the key-name string into a Guile symbol, and then looks it up in table table:
grap=lookup("graphics",conf)
speeds=lookup("animation-speed",grap)
tmp=lookup("chase-scenes",speeds)
...do whatever with the value...
tmp=lookup("love-scenes",speeds)
...do whatever with the value...
Your program just runs down the tree like that, looking for whatever data it needs. When you get to the bottom, you use a Guile built-in to convert the data to an integer, string, or whatever your program expects.
Some data can be iterative, too. In the example, sides is a list that you can modify in your config if you want to define more player sides (say, for AI opponents). Your configuration reader just uses your lookup to find sides, and then iteratively loads one side record at a time until you run out of definitions. (In Scheme terms, you process the car with your lookup function, throw away the car, and continue your iteraton on the cdr.)
The lookup function is really easy to define, and you put it in your library directory so all your programs can use it. It just converts the keyname string to a Guile symbol, and then uses the built-in assoc to find that symbol as the key for anything in the cdr of table. If all your data is in the table format, it works to look up anything, working down the tree recursively. For instance, if you wanted the y size and didn't need anything else, you could do:
y=lookup("y",lookup("size",(lookup("graphics",con
You can also easily define a recursive check_table that verifies that something you loaded is in fact a table structure, in order to trap errors early if a user has screwed up his config file.
The only things I don't use lookup for are the iteration as described above (but even then I use lookup to find the iterative definition, and then use it again to parse each element in the iterative structure), and to get the bottom level data out of the "leaves", e.g. to parse:
(y 490)
I have a library function get_int that accepts a leaf of the form (key integer), extracts the second element, and converts it from a Scheme integer to a C integer, and similarly for other atomic data types.
Also nice, Guile does garbage collection, so you can use it to splice things together out of the configuration and throw away the husks without having to explicitly collect all the trees of objects that you created.
--
> Your sig getting really pathetic.
Whew. That's one relief. When I saw that an AC had replied, I was afraid I was about to get flamed for recommending Guile over XML.
--
> I'm sorry to go off on such a rant, but I am SO tired of everything being done in an XML format.
FWIW (not much, really), I did exactly one hobby project in XML. Yech.
I switched to Guile immediately afterward. You can use it just like a ML if that's all you want to do, and you'll find it way easy to parse. As an added bonus, you can embed code in your data/documents/stylesheets if that's what you want. (Watchout for viruses, though.)
--
> Here in the city I live in, pay phones were removed because it was alleged that all they do is draw drug dealers and prostitutes.
And sadly, crime actually got worse, because they had unwittingly destroyed Superman's natural habitat.
--
More on SW variants can be found at the IMDB
--
> Here's a picture of Cassiopa and Sheba (from Battlestar Galactica)
> Here's a picture of Athena (from Battlestar Galactica)
> Here's a picture of colonel Wilma Deering (from Buck Rogers)
Your catalogue sure has a lot of nice stuff in it, but I don't see the prices anywhere. How much for three (3) of the brunette in white and one (1) of the older blonde?
> Compare the Buck Rogers one with this photo of Natalie (on the right)
Ah, thanks for pointing her out. I thought the one on the left was Natalie, and I couldn't understand why the kiddies were in such a froth over her!
--
> Y38
Not to be confused with 38@Y, which is the title of a p()rn() flick.
Erm... no pun intended with the "flick" part.
--
> Thursday, April 19, 2001 ... the UNIX clock will read 987654321, which is pretty cool.
OK, but when are we going to get the Slashdot User #987654321 throw-down troll account?
We went from 100000 to 400000 so fast that I wouldn't expect it to take too much longer.
--
> True fans are busy awaiting Episode2
I'm no longer a True Fan after E1, but I'm eagerly awaiting E2 anyway, 'cause I heard that it will include a graphic depiction of the slow dismemberment of Jar Jar Binks.
--