You are correct in that software is much more flexible in the ways you describe. Unfortunately, it doesn't work. Even the best simulations are nothing like the real world. Evolution has a nack for exploiting the difference. Which means that the software evolved on virtual robots stands no chance of running on real robots.
These experiments are old news. This has been done before in Brussels by Luc Steels et. al. Read some of that groups papers to understand why software evolution of robots would never work.
(although I can still be a useful exploration of evolving intelligence in its own right).
I hate dead tree books. Sure you can write in
the margins - but I don't like to spoil the book
that way. I much prefer to annotate an e-book,
so I can show or hide my annotations at will.
Every time I move I have to pay for and haul a
large mass of dead trees. Every time I commute
from home to work I have to carry dead trees.
The particular book I usually want it typically
at the wrong end of my commute too. I much
prefer my e-mail library.
Having said that - I have e-book that are in a
propritary format that restricts my feeling of
ownership. I don't like the risk of not being
able to read an e-book if Adobe or MS goes out of
business or decides to make be used particular
s/w or h/w etc. I perfer plain PDF or HTML books
thanks.
Which book would I like to see:
A good book on multi-body dynamics/physics for
interactive uses (using LCP solvers etc.)
There are a few physics books starting to come
out, but they either focus on off-line
computations for off-line rendering (movies etc.),
or they sacrifice too much realism for speed (for
games).
My 5c worth.
>Since we're into car analogies, let me explain in simpler terms...
What's a blinder?
I though it was the things you put on the sides of a horse's eyes do they don't get distracted by things to their sides.If so, that would make it a horse analogy then. ??
Re:Von Neumann Architecture Can't Do It.
on
Arguing A.I.
·
· Score: 1
>I don't think you can get AI working on normal
>Von Neumann Architecture. Sure you could use
>that architecture to simulate the mahcine that
>would work, but hoping to find human-like
>intelligence without using neural networks is,
>IMO, crazy.
Rubbish. Most NNs simulated today are implemented
on just such an architectures. While they
certainly won't be fast enough, that is
orthogonal to their capabilities.
In *principle*, if we understoon the human
brain entirely, we could implemented in todays
PC (although it would run at least trillions of
times slower than the real thing).
>Another requirement would be senses that mimic
>human senses. I'm amazed that people think you
>can simulate human-like intelligence without
>using nearly the exact set of sensory input.
>Dolphins are clearly intelligent creatures,
>but we can't talk to them... and I think it
>has to do with sensory input
Exactly.
>Lastly, you won't be able to program an AI. It
>has to be grown. Human intelligence takes years
>of sensory input, filtering, communication, and
>response analysis to work.
I'd use 'developed' rather than 'grown', but
essentially, yes. But you could still call it
programming.
I can tell you that it isn't the
> deluded scientists and technocrats talking mostly to one another
that has caused the hype. It is the media and to some extent public fanatasy. AI will not be realised in the short to medium term.
Having said that, I believe it *could* be achieved in the medium term if there was a will. AI is like space - while the government doesn't see a need for the benefits (or doesn't understand the potential), progress will crawl along.
Only when the potential becomes evident and there is a political will to realize it, will AI happen (same goes for cheap space access). Pour money in to brain science and AI and it will happen sooner than you think. That is never going to happen though.
This is exactly what persistient OS'es try to avoid. If the whole system is persistient, then not of this suspend/resume voodo is needed because the state of the OS persists for ever anyway.
I, for one, would happily pay 150%, perhaps even 200% for a Linux version of a new game, but NOT if it were released a week after the Windows version. I will not wait that long. I think this is a common attitude.
The 'motor memory' component comes into play when moving your hand down to select the item, not the action of moving up to the menu.
Essentially you brain can associate movement of a particular distance with a function. Hence the function becomes automatic after a time.
('motor memory' functions are probably in the hippocampus, not the motor cortex. But no-one knowns for sure of course).
Of course, MS's window system is a classic example of how to break all of the established rules about GUI design.
If you rool your own, don't copy the mistakes made by MS/Apple/Xerox. GUI research has come along way since those days. Alas, we never see the results because of the entrenched WIMP paridgm - which is very out moded.
In addition, many of the things learnt through WIMP have been successively undone my MS.
For example, menus at the top of the screen (ala MacOS) worked well because a user's motor memory it trained to select items. The functions become motor program - like learning to play a piano.
Unfortunately, this is broken if the menus change (e.g. MS's idea of hiding items and them bringing them back, moving them around etc.). It also doesn't work for menus on window titles - as you need to overload your visual system to select the menu to begin with. Only context and screen top menus retain the original design a pros of menus these days.
There are many many other examples like this. I sugest you find some books on human factors and learn how visual and motor systes work etc.
Grow up.
Actually, I am right about all those points. However, it seems that those faults are not present once the software gets updated.
My cable box says "Comcast" on the left and on the right it says "General Instruments GI".
The serial control method does not work. (Although the cable box has a serial connector, it is for proprietary interfacing)
Obviously, I have the 1.3 version of the software. I'm glad to hear many of my gripes will be fixed in the updated software (I didn't know an update was available).
I just bought the TiVo from amazon.com, and it arrived less than a week ago. I has only dialed into the service twice.
I bought a TiVo because it runs linux and has a good hacker community, can be upgraded etc.
However, while I do like the TiVo, the software could be alot more flexible. There are many little things that could be improved. For example:
1) You can select to record programs by searching by name, but it only searched within a limited horizon of programming (the week or so for which it has the guide). If not found, it can be recorded. For example, you couldn't ask it to record "Mission to Mars" whenever it comes on next.
2) You can't ask it to record programs matching criteria like a particular actor, or keyword in title etc.
3) Once a program has been selected for recording, you can't change the record quality without cancelling it, finding it again via search, channel guide or whatever, and re-selecting to record it.
4) It takes 2-4 hours to process and index the program guide after making a call to TiVo! (what on earth is it doing?)
5) There are a few subtle bugs in the menu display software that sometimes cause display artifacts (rarely though).
6) If you have a partial recording of something that you are also currently recording, it doesn't distinguish between the two - so you can delete the partial until the current recording it complete
7) Sometimes the GUI is slow to respond (I assume the CPU is busy - just evert so slightly underpowered to do everything it needs. Although the record/playback seems to get highest priority - I've noticed no artifacts there)
Many other, small, but annoying things like this.
Perhaps they should Open Source the software and let us./'ers fix and improve it!:)
Apart from those gripes the TiVo is good. I am impressed by the record quality and playback.
OOT: If you have a Comcast General Instrument cable box, the supplied infrared blasters for channel changing are not strong enough and you'll have to but better ones (for ~$30).
I've reciently gotten completely jack of make. After looking around at options I settled on Jam (actually ftjam - the freetype project's mod).
I find it brilliant. You'll have to write your own Jamerules though - check out what the boost.org project have done with it.
Highly recommended.
It ain't necessarily so.
You could move security to be the responsibility of the loader. For example, by only allowing languages without direct memory access (e.g. most except C/C++) and then validating and just-in-time-compiling them you could have a system with security AND no memory protection overhead!
Re:I'm buying one purely for the tiny firewire hd
on
Apple releases iPod
·
· Score: 1
If size matters, then why not buy a Sandisk microdrive from IBM? I believe they come 1GB. They're tiny (they're used in digital cameras).
Of course you'll need a card reader (probably USB I guess). They're cheap enough you could just buy one for each end of the sneaker net.:)
Developers now have a card vendor neutral way to access programmable shaders (pixel and vertex shaders) from DX8.
But does OpenGL1.3 have anything comparable, or do we have to resort to NVidia or ATI extensions?
If that is the case, OpenGL will be hard hit unless a standard vendor neutral extension it added soon.
perhaps someone should write to those scientists who do publish in the publications not meeting the demands, to make them aware of the issues.
I don't think this boycott has the high profile in the scientific community we'd like.
Perhaps an OpenSource Java applet that can view PDF files and an also alternative format. That way, once it becomes widly distributed the alternative format could easily be used.
Perhaps the alternative could be a subset of OpenOffice's XML based open document formats?
Why? Human DNA will look to any aliens like a string of random code - regardless of how advanced they are. Anyone who thinks DNA is a 'blueprint' for a human has been watching too many Hollywood movies.
DNA is a code that describes a *process* that happens to result in a baby human when carried out in a VERY specific environment (a female human!)
So unless we sent a woman along with it - they never be able to make any sense of it at all.
(or perhaps a very detailed knowledge base of a human - which we don't have yet)
Any volunteers?
Is it really task switches, or process switches?
If the overhead is in switching the virtual memory state over, then that is really unnecessary. It's really just a hardware expense to support poor software. If all the apps on an OS were written in a language that didn't allow arbitraty pointer arithmetic (like Java, etc.), and the compiler was trusted (and not buggy) then seperate address spaces between apps (and the kernel) are not necessary.
Don't forget that Solaris is a Mach microkernel OS. (The old SunOS wasn't). Sun had real performance problems when it first came out - but it seems OK now.
Slashdot Mirror
You are correct in that software is much more flexible in the ways you describe. Unfortunately, it doesn't work. Even the best simulations are nothing like the real world. Evolution has a nack for exploiting the difference. Which means that the software evolved on virtual robots stands no chance of running on real robots.
These experiments are old news. This has been done before in Brussels by Luc Steels et. al. Read some of that groups papers to understand why software evolution of robots would never work. (although I can still be a useful exploration of evolving intelligence in its own right).
I hate dead tree books. Sure you can write in
the margins - but I don't like to spoil the book
that way. I much prefer to annotate an e-book,
so I can show or hide my annotations at will.
Every time I move I have to pay for and haul a
large mass of dead trees. Every time I commute
from home to work I have to carry dead trees.
The particular book I usually want it typically
at the wrong end of my commute too. I much
prefer my e-mail library.
Having said that - I have e-book that are in a
propritary format that restricts my feeling of
ownership. I don't like the risk of not being
able to read an e-book if Adobe or MS goes out of
business or decides to make be used particular
s/w or h/w etc. I perfer plain PDF or HTML books
thanks.
Which book would I like to see:
A good book on multi-body dynamics/physics for
interactive uses (using LCP solvers etc.)
There are a few physics books starting to come
out, but they either focus on off-line
computations for off-line rendering (movies etc.),
or they sacrifice too much realism for speed (for
games).
My 5c worth.
>Since we're into car analogies, let me explain in simpler terms...
What's a blinder?
I though it was the things you put on the sides of a horse's eyes do they don't get distracted by things to their sides.If so, that would make it a horse analogy then. ??
>I don't think you can get AI working on normal
>Von Neumann Architecture. Sure you could use
>that architecture to simulate the mahcine that
>would work, but hoping to find human-like
>intelligence without using neural networks is,
>IMO, crazy.
Rubbish. Most NNs simulated today are implemented
on just such an architectures. While they
certainly won't be fast enough, that is
orthogonal to their capabilities.
In *principle*, if we understoon the human
brain entirely, we could implemented in todays
PC (although it would run at least trillions of
times slower than the real thing).
>Another requirement would be senses that mimic
>human senses. I'm amazed that people think you
>can simulate human-like intelligence without
>using nearly the exact set of sensory input.
>Dolphins are clearly intelligent creatures,
>but we can't talk to them... and I think it
>has to do with sensory input
Exactly.
>Lastly, you won't be able to program an AI. It
>has to be grown. Human intelligence takes years
>of sensory input, filtering, communication, and
>response analysis to work.
I'd use 'developed' rather than 'grown', but
essentially, yes. But you could still call it
programming.
I can tell you that it isn't the
> deluded scientists and technocrats talking mostly to one another
that has caused the hype. It is the media and to some extent public fanatasy. AI will not be realised in the short to medium term.
Having said that, I believe it *could* be achieved in the medium term if there was a will. AI is like space - while the government doesn't see a need for the benefits (or doesn't understand the potential), progress will crawl along.
Only when the potential becomes evident and there is a political will to realize it, will AI happen (same goes for cheap space access). Pour money in to brain science and AI and it will happen sooner than you think. That is never going to happen though.
yeah, it's missing the pictures. Too bad iTools pictures don't show up for a large segment of the web viewing public. :(
I remember seeing these gadgets in Tokyo in 1995. I don't remember the brand tho. They're really cool.
This is exactly what persistient OS'es try to avoid. If the whole system is persistient, then not of this suspend/resume voodo is needed because the state of the OS persists for ever anyway.
check out this for some examples of what I mean.
I, for one, would happily pay 150%, perhaps even 200% for a Linux version of a new game, but NOT if it were released a week after the Windows version. I will not wait that long. I think this is a common attitude.
The 'motor memory' component comes into play when moving your hand down to select the item, not the action of moving up to the menu.
Essentially you brain can associate movement of a particular distance with a function. Hence the function becomes automatic after a time.
('motor memory' functions are probably in the hippocampus, not the motor cortex. But no-one knowns for sure of course).
Yes, I did preview my post - I'm just tired. Sorry.
Of course, MS's window system is a classic example of how to break all of the established rules about GUI design.
If you rool your own, don't copy the mistakes made by MS/Apple/Xerox. GUI research has come along way since those days. Alas, we never see the results because of the entrenched WIMP paridgm - which is very out moded.
In addition, many of the things learnt through WIMP have been successively undone my MS.
For example, menus at the top of the screen (ala MacOS) worked well because a user's motor memory it trained to select items. The functions become motor program - like learning to play a piano.
Unfortunately, this is broken if the menus change (e.g. MS's idea of hiding items and them bringing them back, moving them around etc.). It also doesn't work for menus on window titles - as you need to overload your visual system to select the menu to begin with. Only context and screen top menus retain the original design a pros of menus these days.
There are many many other examples like this. I sugest you find some books on human factors and learn how visual and motor systes work etc.
I haven't looked at it - but the plugin idea sounds like Forte for Java / aka. Netbeans - no?
Grow up.
Actually, I am right about all those points. However, it seems that those faults are not present once the software gets updated.
My cable box says "Comcast" on the left and on the right it says "General Instruments GI".
The serial control method does not work. (Although the cable box has a serial connector, it is for proprietary interfacing)
Obviously, I have the 1.3 version of the software. I'm glad to hear many of my gripes will be fixed in the updated software (I didn't know an update was available).
I just bought the TiVo from amazon.com, and it arrived less than a week ago. I has only dialed into the service twice.
I bought a TiVo because it runs linux and has a good hacker community, can be upgraded etc.
./'ers fix and improve it! :)
However, while I do like the TiVo, the software could be alot more flexible. There are many little things that could be improved. For example:
1) You can select to record programs by searching by name, but it only searched within a limited horizon of programming (the week or so for which it has the guide). If not found, it can be recorded. For example, you couldn't ask it to record "Mission to Mars" whenever it comes on next.
2) You can't ask it to record programs matching criteria like a particular actor, or keyword in title etc.
3) Once a program has been selected for recording, you can't change the record quality without cancelling it, finding it again via search, channel guide or whatever, and re-selecting to record it.
4) It takes 2-4 hours to process and index the program guide after making a call to TiVo! (what on earth is it doing?)
5) There are a few subtle bugs in the menu display software that sometimes cause display artifacts (rarely though).
6) If you have a partial recording of something that you are also currently recording, it doesn't distinguish between the two - so you can delete the partial until the current recording it complete
7) Sometimes the GUI is slow to respond (I assume the CPU is busy - just evert so slightly underpowered to do everything it needs. Although the record/playback seems to get highest priority - I've noticed no artifacts there)
Many other, small, but annoying things like this.
Perhaps they should Open Source the software and let us
Apart from those gripes the TiVo is good. I am impressed by the record quality and playback.
OOT: If you have a Comcast General Instrument cable box, the supplied infrared blasters for channel changing are not strong enough and you'll have to but better ones (for ~$30).
I've reciently gotten completely jack of make. After looking around at options I settled on Jam (actually ftjam - the freetype project's mod).
I find it brilliant. You'll have to write your own Jamerules though - check out what the boost.org project have done with it.
Highly recommended.
It ain't necessarily so.
You could move security to be the responsibility of the loader. For example, by only allowing languages without direct memory access (e.g. most except C/C++) and then validating and just-in-time-compiling them you could have a system with security AND no memory protection overhead!
If size matters, then why not buy a Sandisk microdrive from IBM? I believe they come 1GB. They're tiny (they're used in digital cameras). :)
Of course you'll need a card reader (probably USB I guess). They're cheap enough you could just buy one for each end of the sneaker net.
Developers now have a card vendor neutral way to access programmable shaders (pixel and vertex shaders) from DX8.
But does OpenGL1.3 have anything comparable, or do we have to resort to NVidia or ATI extensions?
If that is the case, OpenGL will be hard hit unless a standard vendor neutral extension it added soon.
(pls ignore my accidental AC post)
perhaps someone should write to those scientists who do publish in the publications not meeting the demands, to make them aware of the issues.
I don't think this boycott has the high profile in the scientific community we'd like.
Perhaps an OpenSource Java applet that can view PDF files and an also alternative format. That way, once it becomes widly distributed the alternative format could easily be used.
Perhaps the alternative could be a subset of OpenOffice's XML based open document formats?
Why? Human DNA will look to any aliens like a string of random code - regardless of how advanced they are. Anyone who thinks DNA is a 'blueprint' for a human has been watching too many Hollywood movies.
DNA is a code that describes a *process* that happens to result in a baby human when carried out in a VERY specific environment (a female human!)
So unless we sent a woman along with it - they never be able to make any sense of it at all.
(or perhaps a very detailed knowledge base of a human - which we don't have yet)
Any volunteers?
Is it really task switches, or process switches?
If the overhead is in switching the virtual memory state over, then that is really unnecessary. It's really just a hardware expense to support poor software. If all the apps on an OS were written in a language that didn't allow arbitraty pointer arithmetic (like Java, etc.), and the compiler was trusted (and not buggy) then seperate address spaces between apps (and the kernel) are not necessary.
Don't forget that Solaris is a Mach microkernel OS. (The old SunOS wasn't). Sun had real performance problems when it first came out - but it seems OK now.