Huh? the 'start ww3' was just a complete and total fabrication by fbi. there is no and has been no evidence that us missile control systems have ever been accesible from teh civilian side of this planet, much less that you can get to them via network or modem. He didn't even ever get anywhere near classified, never mind actual secret documnets, and actual command systems are sure to be much more protected.
be able to unroll the database state to where it was at the moment of the query
be able to unroll (or recrerate) the random number generator state
Just seed value is not enough - you will also need
to know how many times the generator was asked. Storing every query and related prng turn values will become cumbersome fast.
You understand that this is just the question of picking the metric? Number of rtansistors needed to implement something is a valid metric for electronic systems, as is the number of gates and number of inter-gate connections. There is a lot of literature on managing the increasing complexity (and transistor count) of electronic circuits, maybe you should read up on it?
But of course your own exmaple defeats your argument - multi-page algebra problems are normaly more complex than one liners.
But you don't even need hard AI to be able to get a neurotic computer - very definately not intelligent biological species (no, this is not a stupid joke on windows users) on this planet can and do have neurological problems and can appear neurotic in their behaviour. That level of complexity is well within the reachaes of what could be realisticly achieved if there was a well-funded backer.
Bugs that have been sparyed by a underdosed pest control agent acting on their neural system might be an example.
There are certain classes of algorithms that take random numbers as input and the quality of the randomness directly affects the quality of results received. So unless you keep the random number generator state (which is probably infeasible in many cases) you will no longer have a back door into the problem of 'why you got the result'.
this is untrue - the complexity of computer systems follows (at least so far) Moore's Law, or in other words, the complexity is increasing exponentialy. A modern microprocessor on your desk will have about as many transistors inside itself as the world total sometime in the 1970s - claiming that the system is not more complex than the computer systems used to be is just nonsense.
But you are talking about architecure of the hardware side - and at that, incredibly simple computer - and not the computer as a whole. And your model doesn't stand up to the case where you have a 1/5000000 probability of a floating point op going wrong because you overclocked the processor.
So... not to be rude but I think you are completelly off and have no idea what you are talking about.
May I suggest you read up on say animats and biomimcry? or even older technology like free agents. This is very much about existing experimental AI technology and what some (mis)uses of it might give as a result.
If you want a simple example, consider simple back-propagation neural networks getting stuck on local minima (or oscilating between two) and not finding the global. If such a neural network was used to set the next goal you would effectively get a neurotic computer.
Technology related discussions should not be moderated by people with no clue on the technology at hand, your post should have got at most 1 and not 5 for its score.
The 2GB wall *IS* an arbitary limit imposed by linus and friends. Among all the other things: a) all intel processors since 386DX have been able to address 4GB of memory b) PPros and Xeons have 36bit address bus and can thus address 64GB of memory.
Stallman's folly? You must be jealous
on
Rumours
·
· Score: 1
>If Sprint is so brain dead they will lose because >of this, whether in boycotting customers (me),
As if they cared if some user or other "boycotted them.
>more development and debugging time, bug ridden >products, and so on. Sprint hasn't got anything I >need.
If they have got nothing you need, then you aren't their customer (or wouldn't be) anyways so they have no reason to give a rat's ass about you.
Oh and you claims about more bugs, more time, etc. are just that - just silly claims.
No rumours in general are anything but reliable, but the people at slashdot really should know better than pass on clearly wrong rumours that BIND, sendmail, etc. are covered by GPL, Copyleft or some "FSF" licence.
Considering that: a) Merced is not out b) nobody really knows when it will be out c) nobody knows how well it really will perform, not even intel d) the price difference in R10K and PII comes mostly from volume
I would say that the aricle this is in reply to is pretty much rubbish.
Slashdot Mirror
Huh? the 'start ww3' was just a complete and total fabrication by fbi. there is no and has been no evidence that us missile control systems have ever been accesible from teh civilian side of this planet, much less that you can get to them via network or modem. He didn't even ever get anywhere near classified, never mind actual secret documnets, and actual command systems are sure to be much more protected.
- be able to unroll the database state to where it was at the moment of the query
- be able to unroll (or recrerate) the random number generator state
Just seed value is not enough - you will also need to know how many times the generator was asked. Storing every query and related prng turn values will become cumbersome fast.You understand that this is just the question of picking the metric? Number of rtansistors needed to implement something is a valid metric for electronic systems, as is the number of gates and number of inter-gate connections. There is a lot of literature on managing the increasing complexity (and transistor count) of electronic circuits, maybe you should read up on it?
But of course your own exmaple defeats your argument - multi-page algebra problems are normaly more complex than one liners.
But you don't even need hard AI to be able to get a neurotic computer - very definately not intelligent biological species (no, this is not a stupid joke on windows users) on this planet can and do have neurological problems and can appear neurotic in their behaviour. That level of complexity is well within the reachaes of what could be realisticly achieved if there was a well-funded backer.
Bugs that have been sparyed by a underdosed pest control agent acting on their neural system might be an example.
There are certain classes of algorithms that take random numbers as input and the quality of the randomness directly affects the quality of results received. So unless you keep the random number generator state (which is probably infeasible in many cases) you will no longer have a back door into the problem of 'why you got the result'.
this is untrue - the complexity of computer systems follows (at least so far) Moore's Law, or in other words, the complexity is increasing exponentialy. A modern microprocessor on your desk will have about as many transistors inside itself as the world total sometime in the 1970s - claiming that the system is not more complex than the computer systems used to be is just nonsense.
But you are talking about architecure of the hardware side - and at that, incredibly simple computer - and not the computer as a whole. And your model doesn't stand up to the case where you have a 1/5000000 probability of a floating point op going wrong because you overclocked the processor.
So... not to be rude but I think you are completelly off and have no idea what you are talking about.
May I suggest you read up on say animats and biomimcry? or even older technology like free agents. This is very much about existing experimental AI technology and what some (mis)uses of it might give as a result.
If you want a simple example, consider simple back-propagation neural networks getting stuck on local minima (or oscilating between two) and not finding the global. If such a neural network was used to set the next goal you would effectively get a neurotic computer.
Technology related discussions should not be moderated by people with no clue on the technology at hand, your post should have got at most 1 and not 5 for its score.
The 2GB wall *IS* an arbitary limit imposed by linus and friends. Among all the other things:
a) all intel processors since 386DX have been able to address 4GB of memory
b) PPros and Xeons have 36bit address bus
and can thus address 64GB of memory.
>If Sprint is so brain dead they will lose because >of this, whether in boycotting customers (me),
As if they cared if some user or other "boycotted them.
>more development and debugging time, bug ridden >products, and so on. Sprint hasn't got anything I >need.
If they have got nothing you need, then you aren't their customer (or wouldn't be) anyways so they have no reason to give a rat's ass about you.
Oh and you claims about more bugs, more time, etc. are just that - just silly claims.
No, they will have mail, dns, routing, etc.
All these programs are covered by BSD and not GPL licences.
There isn't much useable GPL-ed network software at all.
No rumours in general are anything but reliable,
but the people at slashdot really should know better than pass on clearly wrong rumours that BIND, sendmail, etc. are covered by GPL, Copyleft or some "FSF" licence.
Considering that:
a) Merced is not out
b) nobody really knows when it will be out
c) nobody knows how well it really will perform, not even intel
d) the price difference in R10K and PII
comes mostly from volume
I would say that the aricle this is in reply to is pretty much rubbish.