"Former United States Central Intelligence Agency director James Woolsey confirmed in Washington this week that the U.S. steals economic secrets 'with espionage, with communications [intelligence], with reconnaissance satellites,' and that there was now 'some increased emphasis' on economic intelligence. He claimed that economic spying was justified because European companies had a 'national culture' of bribery and were the 'principle offenders from the point of view of paying bribes in major international contracts in the world.'" And he says the U.S. government doesn't deliver corporate secrets to U.S. companies - unless it would benefit them. How reassuring. The source is Heise Online (the publishers of c't). The full article is available in English. See also the recent European report Interception Capabilities 2000 (summary), which the former director said was "intellectually honest."
ANY company which suppports Linux should be applauded for it,
BUT
I think it is time to realise that the times when a companies decision to support Linux was driven by some "higher" goal is starting to end.
Linux is becoming a stratetgic platform, companies are realising that getting into the Linux market will be crucial, and it will start returning a profit in the (not to distant) future.
Especially ATI has shown in the past that it does not believe in open support for its product (yes there have been some changes, but only fairly recently).
It seems that there will be a market for a Linux DVD SW player, as countless posts on./ have shown. The first product to appear is certainly going to get a lot of interest. While it is certainly very important to get DVDs onto the Linux platform, it is worrying that it is only possible with COMMERCIAL SW running on a SPECIFIC graphics card. We all know the problems most commercial drivers have, when it comes to rapidly changing kernel releases, SMP, etc.
Secondly it would tie anyone who wants to watch DVD on Linux to ATI.
Thirdly, a working closed source solution might kill (or dampen) any effort to get a more open solution, as commercial partners (which seem to be necessary, given the DVD NDA constraints, etc) might fear that the market is already occupied by ATI.
Overall I believe that ATI is mainly doing a very clever strategic move, which might end up doing more harm than good to Linux as a multimedia platform.
I still hope some company will produce a CARD INDEPENDANT DVD player app, probabely with an open interface, to enable graphics card vendors to add card specific support (not absolutely necessary, as with modern processors MPEG-2 decoding can be done entirely in SW).
IMHO this would be more helpful than any graphics card dependant solution.
And lastly let's not forget that ATI only announced the SDK, I haven't heard any product announcements yet....
Yes, the theory sounds a bit like something from X-files, and the french are known to, well not exactly like the US,
BUT
When people were speculating about ECHOLON a few years back, many also said it was nothing but conspiracy theories (and the first time its existance was officially admitted, was due to a question in the european parliament).
Secondly, it is generally assumed (or known) that the CIA shifted considerable effort into spying out foreign companies, and passing secret business information on to american companies, when the Cold War faded. That even lead to several american diplomats being "asked to leave" Germany (and yes the French do the same, ask Siemens about the ICE/TGV competition...)
Hence I think it would be wrong to immediately dismiss any thought of cooperation between Microsoft and some US Goverment agency.
I admit I don't believe that IBM was forced to accept MS-DOS as part of some master plan by the NSA to spy on the world, but rather that once MS was becoming so succesful, some agency recognized the potential...
And even if there are no backdoors etc, getting detailed inside knowledge about the protocols, formats, api's etc directly from the source would be a great help to NSA, CIA, etc.
After all, spying really IS the business of those organizations, and thinking they are only interested in other spies and terrorists would be quite naive...
Hmm, I think the author missed the REAL advantage of Code Morphing (or Binary Retranslation).
JIT Compilers (or even FX32!) are pieces of software running ABOVE the OS layer. They are executed for a specific application which runs on top of it (that is the application on top of the JIT environment).
Yes there are similarities between Code Morphing and JIT compilation, as each does some profiling to get better performance in frequently executed code areas. It is further true, that on more predictable and orthogonal platforms like the Alpha, all kinds of optimisation depend on knowing the underlying platform.
BUT they only see the code from the application running on top of them.
The processor itself however gets code from MANY SOURCES. The number of processes running on any machine today is quite large. In addition things as shared libraries, OS calls, etc add a pinch of code. Hence the processor still has to perform all kinds of clever things, trying to resolve dependencies, avoiding stall and stalling at times. For modern multiple Issue machines this gets VERY complicated. It slows down cycle times, increases size (and hence speed) takes up space which could be used for better performing optimisations (larger cache, etc), etc.
This is one of the major headaches for CPU designers these days, as these parts increase unproportionally in complexity when trying to add more functional units, etc.
This is the reason why VLIW (very long instruction words) are used in the next generation processors (i.e Itanuim). In those the compiler does all the hard work, resolving all dependencies stalls, etc, at compile time, and then simply sending one long instruction to the processor, which includes one instruction for every functional unit. This means the processor can run at MUCH Higher speed, as it loses a big overhead complexity.
This is the reason why TRANSMETA uses such a VLIW, it enables them to build fast, small and CHEAP processors which can run at quite high speeds. (I am sure they could do more than 500Mhz, but then cost and power consumption get in the way. And the are much smaller and havent (yet) got the man resources as Intel, AMD)
Now where does code morphing come into play ? I recently saw a few comments asking for access to the VLIW instruction directly, bypassing the code morphing. Well, that would DEFEAT the concept !
The Big Thing about code morphing is that it sees ALL instructions which get executed, in the right order. This includes the library calls, the OS routines, etc. Hence it can do MUCH BETTER optimization than ANY Compiler by definition can do. A compiler can only know about the program it compiles (yes you can do clever profiling, but you still only have the application source to optimize), it cannot optimize ACROSS APPLICATIONS.
The code morphing software has absolut knowledge about what code has executed and what code currently tries to execute. This enables it to do much better optimizations. Furthermore it is SPECIFIC to the underlying processor. It knows everything about the processor and even more important, the processor is designed specifically for it ! (Hence the two versions are internally comletely different, each optimised for the kind of code linux/Win it is executing)
This is what makes the concept so interesting. It adds an additional abstraction layer on top of the actual hardware, allowing the underlying implementation to be always up-to-date and optimized for the application (removing all the complexity of having backwards compatible hardware, ask Intel;-)
Overall, the approach is a MAJOR step forward, IMHO it can be compared to the CISC->RISC transition. It is after all the first implementation, competing with products well understood and optimized for years. It has potential to solve many of the major headaches of modern CPU designs (including COST). Comparing it to JIT is missing the BIG points.
Ahem, the sentence was supposed to read: "As much as it [the problem of repeatability] should be adressed (to make robotics a "serious" science;-> , IMHO.....".
"...but that does not imply that intelligence requires randomness."
I like to think it is one of the necesseties for intelligence, especially for low level intelligence.
Yes, humans might be that advanced as to come up with new ideas, by contemplating a problem long enough, but even for us (limited) random behaviour can bring unexpected rewards. I don't think the first people to discover burned (i.e. roasted) meat tastes pleasant or the primate noticing that using rocks as hammers opens things, did so by reasoning.
I believe acts of randomness lead to unexpected results, which can then be learned and potentially be understood due to intelligence. Hence IMHO they make up a big part of intelligence,- DISCOVERING and acquiring NEW behaviours.
The same applies to robots. If a robot, due to the unpredictability of the sensors, i.e. a freak perception, or due to an intentional random element in its behaviour "discovers" a "new" behaviour which leads to high rewards, the associative structures might be able to reproduce it; the robot seems to behave intelligent.
Hence, I believe that a random and unpredictable element SHOULD be part of any attempt to produce "intelligent" machines.
I agree that I personally thing a subsymbolic AI approach should be the way forward. After all that seems to be a pretty scalable way of achieving "intelligent" behaviour, if we look at animals. Furthermore like you say, it does not require the "creator" to think of everything himself (hmm, I like that...):-)
However the problem of reproduciability also exists (to an extend) with all subsymbolic AI solutions (AFAIK), after all EVERY robot is unique (sensor characteristics, motor, environment....).
As much as it should be adressed (to make robotics are "serious" since;-> IMHO it is part of the question, as "intelligent behaviour" has to contain some sort of randomness and unpredictability.
According to the pressrelease the robots is mainly controlled via a remote control (or PC).
This demonstartes one of the major problems of robotics to date :
Robot Navigation in umodified environments.
Lots of research goes into the area, but self localisation in an umodified home environment, where lots of obstacles exist, some of which keep moving around (i.e. humans, small furniture pieces, etc) is still a big problem.
The same goes for the use of robotics vision. Yes there are some clever schemes to use stereo cameras(to estimate distance) for object detection and avoidance and some help in landmark recognition, but it takes a lot of processing power and is only mildly reliable. Identifying specific people without major constraints (i.e. person has to face the robot from a certain angle without moving) is AFAIK also not reliably possible.
This are the rasons why at the moment all "proper" (i.e. learning, and not remote controlled or having hardwired complex behaviour) commercial robots are toys. If your AIBO moves in the wrong direction, runs into you, or doesn't manage to fetch its pink ball, nobody will complain it doesn't work properly, people will enjoy it, and find it highly amusing.
Reliable, intelligent, robot servants are still a long way off, but my guess is that they surely will exist in 20 years or so and most likely they won't look anything like humanoid at all...
BTW, for anyone interested there is the COG project at MIT, doing research into a humanoid robot, it is VERY impressive.
Take a look at some of the videos to see what they are already capable of. Especially the head demos are impressive.
http://www.ai.mit.edu/projects/cog/video_index.htm l
Here's part of the definition given in the injunction, what constitutes deCSS :
(c) "DeCSS" means any computer program, file or device that may be used to decrypt or unscramble the contents of DVDs that are protected, or otherwise to circumvent the protection afforded, by CSS and that permits the copying of the contents or any portion thereof.
"any device that may be used to decrypt or unscramblw the contents" hmm, isn't that what any DVD player has to do... (and yes I can copy the content to S-VHS or something)
Does that mean that 2600 have to give back all their DVD players;-)
Slashdot Mirror
As noted in this story:
h tml
http://slashdot.org/articles/00/03/12/1910255.s
(Follow the link to the link of the full story)
Here's the headline blurb:
"Former United States Central Intelligence Agency director James Woolsey confirmed in Washington this week that the U.S. steals economic secrets 'with espionage, with communications [intelligence], with reconnaissance satellites,' and that there was now 'some increased emphasis' on economic intelligence. He claimed that economic spying was justified because European companies had a 'national culture' of bribery and were the 'principle offenders from the point of view of paying bribes in major international contracts in the world.'" And he says the U.S. government doesn't deliver corporate secrets to U.S. companies - unless it would benefit them. How reassuring. The source is Heise Online (the publishers of c't). The full article is available in English. See also the recent European report Interception Capabilities 2000 (summary), which the former director said was "intellectually honest."
Scary !
Frank
ANY company which suppports Linux should be applauded for it,
./ have shown. The first product to appear is certainly going to get a lot of interest. While it is certainly very important to get DVDs onto the Linux platform, it is worrying that it is only possible with COMMERCIAL SW running on a SPECIFIC graphics card.
BUT
I think it is time to realise that the times when a companies decision to support Linux was driven by some "higher" goal is starting to end.
Linux is becoming a stratetgic platform, companies are realising that getting into the Linux market will be crucial, and it will start returning a profit in the (not to distant) future.
Especially ATI has shown in the past that it does not believe in open support for its product (yes there have been some changes, but only fairly recently).
It seems that there will be a market for a Linux DVD SW player, as countless posts on
We all know the problems most commercial drivers have, when it comes to rapidly changing kernel releases, SMP, etc.
Secondly it would tie anyone who wants to watch DVD on Linux to ATI.
Thirdly, a working closed source solution might kill (or dampen) any effort to get a more open solution, as commercial partners (which seem to be necessary, given the DVD NDA constraints, etc) might fear that the market is already occupied by ATI.
Overall I believe that ATI is mainly doing a very clever strategic move, which might end up doing more harm than good to Linux as a multimedia platform.
I still hope some company will produce a CARD INDEPENDANT DVD player app, probabely with an open interface, to enable graphics card vendors to add card specific support (not absolutely necessary, as with modern processors MPEG-2 decoding can be done entirely in SW).
IMHO this would be more helpful than any graphics card dependant solution.
And lastly let's not forget that ATI only announced the SDK, I haven't heard any product announcements yet....
Just my $0.02
Frank
Mmhhh, somebody decided my post was FLAMEBAIT, whoever did that, feel free to e-mail me to explain....
someonelse thought it was "overrated", well....
Yes, the theory sounds a bit like something from X-files, and the french are known to, well not exactly like the US,
BUT
When people were speculating about ECHOLON a few years back, many also said it was nothing but conspiracy theories (and the first time its existance was officially admitted, was due to a question in the european parliament).
Secondly, it is generally assumed (or known) that the CIA shifted considerable effort into spying out foreign companies, and passing secret business information on to american companies, when the Cold War faded. That even lead to several american diplomats being "asked to leave" Germany (and yes the French do the same, ask Siemens about the ICE/TGV competition...)
Hence I think it would be wrong to immediately dismiss any thought of cooperation between Microsoft and some US Goverment agency.
I admit I don't believe that IBM was forced to accept MS-DOS as part of some master plan by the NSA to spy on the world, but rather that once MS was becoming so succesful, some agency recognized the potential...
And even if there are no backdoors etc, getting detailed inside knowledge about the protocols, formats, api's etc directly from the source would be a great help to NSA, CIA, etc.
After all, spying really IS the business of those organizations, and thinking they are only interested in other spies and terrorists would be quite naive...
Just my $0.02
Frank
Hmm, I think the author missed the REAL advantage of Code Morphing (or Binary Retranslation).
;-)
JIT Compilers (or even FX32!) are pieces of software running ABOVE the OS layer. They are executed for a specific application which runs on top of it (that is the application on top of the JIT environment).
Yes there are similarities between Code Morphing and JIT compilation, as each does some profiling to get better performance in frequently executed code areas. It is further true, that on more predictable and orthogonal platforms like the Alpha, all kinds of optimisation depend on knowing the underlying platform.
BUT they only see the code from the application running on top of them.
The processor itself however gets code from MANY SOURCES. The number of processes running on any machine today is quite large. In addition things as shared libraries, OS calls, etc add a pinch of code.
Hence the processor still has to perform all kinds of clever things, trying to resolve dependencies, avoiding stall and stalling at times. For modern multiple Issue machines this gets VERY complicated. It slows down cycle times, increases size (and hence speed) takes up space which could be used for better performing optimisations (larger cache, etc), etc.
This is one of the major headaches for CPU designers these days, as these parts increase unproportionally in complexity when trying to add more functional units, etc.
This is the reason why VLIW (very long instruction words) are used in the next generation processors (i.e Itanuim). In those the compiler does all the hard work, resolving all dependencies stalls, etc, at compile time, and then simply sending one long instruction to the processor, which includes one instruction for every functional unit. This means the processor can run at MUCH Higher speed, as it loses a big overhead complexity.
This is the reason why TRANSMETA uses such a VLIW, it enables them to build fast, small and CHEAP processors which can run at quite high speeds. (I am sure they could do more than 500Mhz, but then cost and power consumption get in the way. And the are much smaller and havent (yet) got the man resources as Intel, AMD)
Now where does code morphing come into play ?
I recently saw a few comments asking for access to the VLIW instruction directly, bypassing the code morphing. Well, that would DEFEAT the concept !
The Big Thing about code morphing is that it sees ALL instructions which get executed, in the right order. This includes the library calls, the OS routines, etc. Hence it can do MUCH BETTER optimization than ANY Compiler by definition can do. A compiler can only know about the program it compiles (yes you can do clever profiling, but you still only have the application source to optimize), it cannot optimize ACROSS APPLICATIONS.
The code morphing software has absolut knowledge about what code has executed and what code currently tries to execute. This enables it to do much better optimizations.
Furthermore it is SPECIFIC to the underlying processor. It knows everything about the processor and even more important, the processor is designed specifically for it !
(Hence the two versions are internally comletely different, each optimised for the kind of code linux/Win it is executing)
This is what makes the concept so interesting. It adds an additional abstraction layer on top of the actual hardware, allowing the underlying implementation to be always up-to-date and optimized for the application (removing all the complexity of having backwards compatible hardware, ask Intel
Overall, the approach is a MAJOR step forward, IMHO it can be compared to the CISC->RISC transition. It is after all the first implementation, competing with products well understood and optimized for years. It has potential to solve many of the major headaches of modern CPU designs (including COST). Comparing it to JIT is missing the BIG points.
I think it will go a LONG WAY !
Frank
Ahem, the sentence was supposed to read: ;-> , IMHO .....".
"As much as it [the problem of repeatability] should be adressed (to make robotics a "serious" science
"...but that does not imply that intelligence requires randomness."
I like to think it is one of the necesseties for intelligence, especially for low level intelligence.
Yes, humans might be that advanced as to come up with new ideas, by contemplating a problem long enough, but even for us (limited) random behaviour can bring unexpected rewards. I don't think the first people to discover burned (i.e. roasted) meat tastes pleasant or the primate noticing that using rocks as hammers opens things, did so by reasoning.
I believe acts of randomness lead to unexpected results, which can then be learned and potentially be understood due to intelligence. Hence IMHO they make up a big part of intelligence,- DISCOVERING and acquiring NEW behaviours.
The same applies to robots. If a robot, due to the unpredictability of the sensors, i.e. a freak perception, or due to an intentional random element in its behaviour "discovers" a "new" behaviour which leads to high rewards, the associative structures might be able to reproduce it; the robot seems to behave intelligent.
Hence, I believe that a random and unpredictable element SHOULD be part of any attempt to produce "intelligent" machines.
Frank
Well said.
...) :-)
;-> IMHO it is part of the question, as "intelligent behaviour" has to contain some sort of randomness and unpredictability.
I agree that I personally thing a subsymbolic AI approach should be the way forward. After all that seems to be a pretty scalable way of achieving "intelligent" behaviour, if we look at animals. Furthermore like you say, it does not require the "creator" to think of everything himself (hmm, I like that
However the problem of reproduciability also exists (to an extend) with all subsymbolic AI solutions (AFAIK), after all EVERY robot is unique (sensor characteristics, motor, environment....).
As much as it should be adressed (to make robotics are "serious" since
Frank
According to the pressrelease the robots is mainly controlled via a remote control (or PC).
m l
This demonstartes one of the major problems of robotics to date :
Robot Navigation in umodified environments.
Lots of research goes into the area, but self localisation in an umodified home environment, where lots of obstacles exist, some of which keep moving around (i.e. humans, small furniture pieces, etc) is still a big problem.
The same goes for the use of robotics vision. Yes there are some clever schemes to use stereo cameras(to estimate distance) for object detection and avoidance and some help in landmark recognition, but it takes a lot of processing power and is only mildly reliable.
Identifying specific people without major constraints (i.e. person has to face the robot from a certain angle without moving) is AFAIK also not reliably possible.
This are the rasons why at the moment all "proper" (i.e. learning, and not remote controlled or having hardwired complex behaviour) commercial robots are toys. If your AIBO moves in the wrong direction, runs into you, or doesn't manage to fetch its pink ball, nobody will complain it doesn't work properly, people will enjoy it, and find it highly amusing.
Reliable, intelligent, robot servants are still a long way off, but my guess is that they surely will exist in 20 years or so and most likely they won't look anything like humanoid at all...
BTW, for anyone interested there is the COG project at MIT, doing research into a humanoid robot, it is VERY impressive.
Take a look at some of the videos to see what they are already capable of. Especially the head demos are impressive.
http://www.ai.mit.edu/projects/cog/video_index.ht
Frank
Here's part of the definition given in the injunction, what constitutes deCSS :
;-)
(c) "DeCSS" means any computer program, file or device that may be used to decrypt or unscramble the contents of DVDs that are protected, or otherwise to circumvent the protection afforded, by CSS and that permits the
copying of the contents or any portion thereof.
"any device that may be used to decrypt or unscramblw the contents" hmm, isn't that what any DVD player has to do... (and yes I can copy the content to S-VHS or something)
Does that mean that 2600 have to give back all their DVD players
Frank
Well, I have succesfully fled Germany, but:
;-)
"Hauer" sounds more like someone who frequently beats up other people ("hauen" = old fashioned for "to beat")
Not sure wether that would be a suitable alternative to hacker