Open Standards should be kept in the most competent hands. In this case OpenGL hasn't had to go through 8 revisions. And the last time I checked, www.opengl.org had a couple of discussions groups hosted, one of which was for requests for new extensions to OpenGL. From any "peon" developer. Quite nice really. Now I just need to write up that proposal for proper bump-mapping (graphics geeks raise your hands if you know what I'm talking about...:)
And the kids who have gone to schools tend to have better socialization skills. Go figure. I do know quite a few bright, talented people who were home schooled. They also have difficulty interacting with people, especially working as a group.
Are the kid's who aren't home schooled dumber? Would I be smarter if I hadn't gone to a public school? This is a silly argument to be involved in, I quit:)
Of course, if you want your Oracle Database to go fast, you jam as much of it as you can into RAM... Downside: when you buy Gigabytes worth of RAM for a DEC/Compaq Alpha running Oracle it's a wee bit pricey:)
After taking a look around the DEMO website, I'm curious to see where you stand on the "Good Old Fashioned AI" vs "Evolutionary" (for lack of a better term) debate. Do you see a need for expert systems or case-based reasoning, or do you expect A.I. approaches like Neural Networks or Genetic Algorithms to eventually supplant GOFAI? Explanation is something that GOFAI is well-suited to. Do you see, for example, Neural Networks being used in explanation of solutions?How?
Re:seems like it's missing a few things
on
Jet3d Game Engine
·
· Score: 1
Real-time rendering of NURBS is fast and easy. Under OpenGL you have access to the GLU libraries and their support for NURBS. There are plenty of parameters to control the amount of geometry you generate for a given NURBS surface. I don't know about Direct3D tho... And you could always do neat things with Gaussian surfaces, they seem a bit better suited to parameterizing the amount of geometry you want to emit while maintaining good detail.
I've built a couple of programs for doing NURBS rendering, and while they certainly FLY on an SGI Octane, they are slower on the PC, but I realize I'm generating literally millions of vertices... If I make the rational decision to relax the quality of the NURBS surface, speed greatly improves -> less geometry. Realize that I was using a grid of approximately 100*100 control points for my NURB...
OpenGL provides better "drivers" on better hardware than what is available from MS. The API is much cleaner (and after taking a quick browse of the OpenAL sight, I rather liked their API: very OpenGL-ish). DirectX, more specifically, Direct3D can't hack it for high-quality high-speed rendering. We briefly toyed with Direct3D for some visualization tools, only to realize just how kludgy the API was and promptly went back to OpenGL. The code is much cleaner. And more importantly, we can run our tools on everything from a beautiful 16-processor SGI to a PC to a Mac to some odd Linux box, to well, one day perhaps a PalmPilot (see that link from last week ). Well, okay, maybe not a PalmPilot....:)
By the way, nice troll on the "exposing geometry features" bit. That has got to be the easiest troll for OpenGL baiters. It's just so obviously wrong that it sucks people right in. It's a bit like arguing with someone who claims gravity doesn't exist: It just hurts your head...
Then again, the number of people with related in-depth technical knowledge responding to a subject in a slashdot article can probably be counted on one hand:)
Very true. But by virtue of being "simple" or "practical", the theory is easier to prove or disprove.
I think it would be far easier to either prove or disprove the theory that consciousness is a macro phenomenon caused by "large" electrical pulses, etc. in the brain. And therefore more worthy of examination. At least to the point where someone goes, "Ah. No, sorry, you're wrong about that, here's why...".
The theory that consciousness arises from some quantum effect strikes me as being a bit like the "ether" theory for light. Theory that's a bit too complicated for a simple phenomena.
Of course, you might not view consciousness as a simple phenomena...
Hey! Do you think you could get a bowl of hot grits to write the occasional opinion piece too? That would be most excellent... Man, I'd love to have a box I could check that would mark my posts as knowingly off-topic...
I'm going to have to agree here (see earlier AC response to this article). Saying the conciousness is the result some quantum effects is ignoring a much simpler existing explanation... And let's not forget about the chemical reactions going on too... Mondo large scale activity.
How did the brain get to the other side of the wall? It ran into the wall 10^23 times till it passed through...
Ah, but >>companies and hey! If there are any SGI folks reading this: please do something about that "bool" keyword! Those compiler switches do nothing for us!
Oh! And having used Motif. I'd really like it to go away. My personal choice would be Qt (as long as it ran on everything under the sun (and them too)).
Reply: What happens when you move and you have your mail forwarded to the new address? Oh wait. That works... A change of address form is a change of address form...
Off-topic from the reply: What happens with all the laws regarding normal mail? Would those apply to your email address? Do they already apply? Curious...
Why shouldn't they look a bit like MS Windows stuff? It's a nice user-interface (note: not the best/worst/etc. okay?:) And it doesn't alienate new users with, dare I say it, a rather garish desktop look...
Besides which, the gnome and kde file managers look and operate an awful lot like Explorer under Win98...hmm... Evil Plot or useful "feature", you decide...
Though the points I'm making are valid: closed (or very tight) control of standards is important to software, and that while it is possible for a user to monkey around with open source software, large projects (such as this) are impossible for any one person to play with and reasonably use.
I can monkey around with the code to Linux, but I wouldn't want to try to fix or change something potentially important in it. It's not worth my time. I think the open source advocates make a mistake in claiming that as beneficial.
I think we can agree that more people use java (applets on web or applications) than use linux. I would propose that there is a proportionally smaller group that would want to fiddle with the guts of Java (not write programs in it) than the numerous people who have supported Linux, and created quite a nifty operating system. One whose growth and modification is tightly controlled.
I think that anyone outside of Sun, or groups not working in close support with Sun are wasting their time on a GPL'd java-like language if they intend it to replace Java. It's great if they want to noodle around with it as a hobbyist, but naive to try and replace it, simply because it hasn't got a license you like.
Java is free, Users (in general) are clueless
on
RMS on Java and GPL
·
· Score: 1
From the Story:
"As long as Sun uses a license that does not permit unlimited technical changes, it will remain unacceptable, and we will keep on building our replacement."
This is the very reason I would object to Sun releasing any of their code. I do not want anyone changing the Java implementation or APIs. I do not want broken but GPL'd packages. I want all of the functionality that Sun has given me, and I want the massive head start they have. I want them to have sole control because there's only "one" of them making changes to the code, not 1000 crack-addled monkeys thinking up "neat" extensions to the language that breaks software.
Oh, another thing :
"So I suggest you start planning for a world in which you, the users of Java, maintain compatibility where you want it through your own free choice, rather than by asking Sun to impose it."
It is best to assume the user is an idiot. This keeps your programs sane, and gives a sane interface to the user. I'll admit to being both a user and an idiot.
The last time I checked, Sun sure didn't charge anything to use their compiler or classes or virtual machines... And they didn't even ask us to help them develop anything. Quite nice of them. They are entitled to their property, and they've been nice enough to make the end tools free. Heck, a lot of people go out and buy compilers! And operating systems! They must be mad! Er, or perhaps those are just opinions in your pocket...
Well, I was bit more interested in the (implied) use of neural networks to simulate behaviour (this is also going to be a bit of a reply to penguincide's article link to the Robokoneko:).
How exactly would you determine when to be, say, "frisky"? Or rather, we have a neural network with a bunch of sensor inputs, some other state data, etc. How do you take that input, run it through a network, and decide that the emotion you'd like Aibo (or Robokoneko) to exhibit is "frisky" --> where you've gone and developed some subsystem (perhaps a neural network) to provide the physical "friskiness" of the robot?
I took a very quick look around http://www.hip.atr.co.jp/~degaris/papers/ and didn't find anything that indicated how they planned to train the network(s) (Although there was quite a bit of interesting stuff to read). One of the bits of correspondence I found questioned how higher level control (not just basic positioning/simple actions) was going to be done, and I didn't really see an answer to it. Again, I just took a quick look around, I don't have the time (today) to read through all the papers...
You wrote :
However, let's assume you do use an absolute list. This -does- get complex, as your output is a time-dependent matrix.
In other words, your error is sampled not just over an instantaneous output, but over a period of time.
Let's take an example. Say your robot dog is trying to walk from one end of a room to another. It encounters an obsticle. It's no use the dog just turning left. Yes, it's no longer obstructed, but it's now going in the wrong direction. BUT, it's ONLY the wrong direction because it didn't turn back on course, once the obstruction was avoided.
In this case, a shortest path algorithm would be much better:) Your net's outputs are driving the velocity/acceleration of motors, so, at any given moment you'd like : the acceleration to be smooth (good to backpropagate errors immediately), and the dog to be heading along a predetermined path (After all, this is what offline training's all about : predetermined data). Hum, this bit's come out all wrong! Anyway, the size/space requirements for most of the input data is pretty small (probably under a gigabyte or so), so offline training is probably viable.
No, we're not talking Terminator or HAL-9000 technology here. A large, FAST neural net isn't difficult to code. Place the software in virtual reality, for a while (simulate as much time as you like), then dump into the robot's computer. Not that complicated to do.
While the flexibility of neural networks is great, they need to be trained (Dead on, on the offline training btw). The bad thing: When training that neural network you need to be working towards a goal.
How do you define that goal? You've got (assuming some sort of feedforward or derivative network) an output layer generating some output vector, and you're comparing it to an established goal vector. How do you define a goal vector for a behaviour? Is the euclidean distance from your output vector to the goal vector a valid metric for determining error? I can see that being very unpleasant to formulate...
Not that it wouldn't be a cool research project...
Slashdot Mirror
Open Standards should be kept in the most competent hands. In this case OpenGL hasn't had to go through 8 revisions. And the last time I checked, www.opengl.org had a couple of discussions groups hosted, one of which was for requests for new extensions to OpenGL. From any "peon" developer. Quite nice really. Now I just need to write up that proposal for proper bump-mapping (graphics geeks raise your hands if you know what I'm talking about... :)
Are the kid's who aren't home schooled dumber? Would I be smarter if I hadn't gone to a public school? This is a silly argument to be involved in, I quit :)
Of course, if you want your Oracle Database to go fast, you jam as much of it as you can into RAM... Downside: when you buy Gigabytes worth of RAM for a DEC/Compaq Alpha running Oracle it's a wee bit pricey :)
After taking a look around the DEMO website, I'm curious to see where you stand on the "Good Old Fashioned AI" vs "Evolutionary" (for lack of a better term) debate. Do you see a need for expert systems or case-based reasoning, or do you expect A.I. approaches like Neural Networks or Genetic Algorithms to eventually supplant GOFAI? Explanation is something that GOFAI is well-suited to. Do you see, for example, Neural Networks being used in explanation of solutions?How?
I've built a couple of programs for doing NURBS rendering, and while they certainly FLY on an SGI Octane, they are slower on the PC, but I realize I'm generating literally millions of vertices... If I make the rational decision to relax the quality of the NURBS surface, speed greatly improves -> less geometry. Realize that I was using a grid of approximately 100*100 control points for my NURB...
By the way, nice troll on the "exposing geometry features" bit. That has got to be the easiest troll for OpenGL baiters. It's just so obviously wrong that it sucks people right in. It's a bit like arguing with someone who claims gravity doesn't exist: It just hurts your head...
Then again, the number of people with related in-depth technical knowledge responding to a subject in a slashdot article can probably be counted on one hand :)
I think it would be far easier to either prove or disprove the theory that consciousness is a macro phenomenon caused by "large" electrical pulses, etc. in the brain. And therefore more worthy of examination. At least to the point where someone goes, "Ah. No, sorry, you're wrong about that, here's why...".
The theory that consciousness arises from some quantum effect strikes me as being a bit like the "ether" theory for light. Theory that's a bit too complicated for a simple phenomena.
Of course, you might not view consciousness as a simple phenomena...
Hey! Do you think you could get a bowl of hot grits to write the occasional opinion piece too? That would be most excellent... Man, I'd love to have a box I could check that would mark my posts as knowingly off-topic...
How did the brain get to the other side of the wall? It ran into the wall 10^23 times till it passed through...
Oh! And having used Motif. I'd really like it to go away. My personal choice would be Qt (as long as it ran on everything under the sun (and them too)).
You 80 column mode is breaking my Gameboy. You have no idea how long it takes to pick out the characters using only a few buttons...
What happens when you move and you have your mail forwarded to the new address? Oh wait. That works... A change of address form is a change of address form...
Off-topic from the reply:
What happens with all the laws regarding normal mail? Would those apply to your email address? Do they already apply? Curious...
But you bought that video card... wonder if the cost of driver (98/NT) developement figured into the sales price of those chips they make...
But SGI's is totally OpenGL compliant. MESA is >>mostly OpenGL compliant (very close but not quite, eh?). Conformance is VERY important.
Besides which, the gnome and kde file managers look and operate an awful lot like Explorer under Win98...hmm... Evil Plot or useful "feature", you decide...
It is a convenient quick-and-dirty translator...
You could also say it's the "pretty" version of postscript...
Though the points I'm making are valid: closed (or very tight) control of standards is important to software, and that while it is possible for a user to monkey around with open source software, large projects (such as this) are impossible for any one person to play with and reasonably use.
I can monkey around with the code to Linux, but I wouldn't want to try to fix or change something potentially important in it. It's not worth my time. I think the open source advocates make a mistake in claiming that as beneficial.
I think we can agree that more people use java (applets on web or applications) than use linux. I would propose that there is a proportionally smaller group that would want to fiddle with the guts of Java (not write programs in it) than the numerous people who have supported Linux, and created quite a nifty operating system. One whose growth and modification is tightly controlled.
I think that anyone outside of Sun, or groups not working in close support with Sun are wasting their time on a GPL'd java-like language if they intend it to replace Java. It's great if they want to noodle around with it as a hobbyist, but naive to try and replace it, simply because it hasn't got a license you like.
"As long as Sun uses a license that does not permit unlimited technical changes, it will remain unacceptable, and we will keep on building our replacement."
This is the very reason I would object to Sun releasing any of their code. I do not want anyone changing the Java implementation or APIs. I do not want broken but GPL'd packages. I want all of the functionality that Sun has given me, and I want the massive head start they have. I want them to have sole control because there's only "one" of them making changes to the code, not 1000 crack-addled monkeys thinking up "neat" extensions to the language that breaks software.
Oh, another thing :
"So I suggest you start planning for a world in which you, the users of Java, maintain compatibility where you want it through your own free choice, rather than by asking Sun to impose it."
It is best to assume the user is an idiot. This keeps your programs sane, and gives a sane interface to the user. I'll admit to being both a user and an idiot.
The last time I checked, Sun sure didn't charge anything to use their compiler or classes or virtual machines... And they didn't even ask us to help them develop anything. Quite nice of them. They are entitled to their property, and they've been nice enough to make the end tools free. Heck, a lot of people go out and buy compilers! And operating systems! They must be mad! Er, or perhaps those are just opinions in your pocket...
How exactly would you determine when to be, say, "frisky"? Or rather, we have a neural network with a bunch of sensor inputs, some other state data, etc. How do you take that input, run it through a network, and decide that the emotion you'd like Aibo (or Robokoneko) to exhibit is "frisky" --> where you've gone and developed some subsystem (perhaps a neural network) to provide the physical "friskiness" of the robot?
I took a very quick look around http://www.hip.atr.co.jp/~degaris/papers/ and didn't find anything that indicated how they planned to train the network(s) (Although there was quite a bit of interesting stuff to read). One of the bits of correspondence I found questioned how higher level control (not just basic positioning/simple actions) was going to be done, and I didn't really see an answer to it. Again, I just took a quick look around, I don't have the time (today) to read through all the papers...
You wrote :
However, let's assume you do use an absolute list. This -does- get complex, as your output is a time-dependent matrix.
In other words, your error is sampled not just over an instantaneous output, but over a period of time.
Let's take an example. Say your robot dog is trying to walk from one end of a room to another. It encounters an obsticle. It's no use the dog just turning left. Yes, it's no longer obstructed, but it's now going in the wrong direction. BUT, it's ONLY the wrong direction because it didn't turn back on course, once the obstruction was avoided.
In this case, a shortest path algorithm would be much better :) Your net's outputs are driving the velocity/acceleration of motors, so, at any given moment you'd like : the acceleration to be smooth (good to backpropagate errors immediately), and the dog to be heading along a predetermined path (After all, this is what offline training's all about : predetermined data). Hum, this bit's come out all wrong! Anyway, the size/space requirements for most of the input data is pretty small (probably under a gigabyte or so), so offline training is probably viable.
No, we're not talking Terminator or HAL-9000 technology here. A large, FAST neural net isn't difficult to code. Place the software in virtual reality, for a while (simulate as much time as you like), then dump into the robot's computer. Not that complicated to do.
While the flexibility of neural networks is great, they need to be trained (Dead on, on the offline training btw). The bad thing: When training that neural network you need to be working towards a goal.
How do you define that goal? You've got (assuming some sort of feedforward or derivative network) an output layer generating some output vector, and you're comparing it to an established goal vector. How do you define a goal vector for a behaviour? Is the euclidean distance from your output vector to the goal vector a valid metric for determining error? I can see that being very unpleasant to formulate...
Not that it wouldn't be a cool research project...