One of my mates had a gag he used to pull on his friends - he'd get their passwords through word assocation. The conversation would go like this:
JG (my friend): "Let's do one of those word association things - you know, I say a word and you say the first thing that comes into your head."
DI (Dumb Idiot): "Oh, okay."
JG: "Chicken"
DI: "Dinner" .. and so on...
JG: "Password"
DI: "********.... oh f**k"
He got it to work a few times - got a few of his mates to say their passwords.
Guess it shows you'll never go broke relying on human stupidity.
--
Spam Echelon, mention Terrorism in all your emails!
The state of a neural network, for example, is a black box from which no real insight can be gained...
Not so. The weights of the nodes inside a neural network can, and are analyzed, to determine the features of the input that the network uses to produce a result. I think this is known as cluster analysis - not my specialty.
magine a neural network whose training set is the raw molecular structure data of different diseases (input) and their cures (output). The network trains on this data until the error reduces to zero...
Cool idea, but not possible, at least by today's knowledge.
For one thing, neural nets do not produce output data of this complexity. A neural network has rigidly defined inputs and rigidly defined outputs. I can't see how you could constrain possible structure of a molecule sufficiently to be used as the output of a neural net.
Secondly, a neural net will not work on any possible set of data you feed into it. There have to be common patterns in the data that the net can draw out. The beauty of neural nets is that the person feeding the data in need not know the exact patterns. If these patterns are not present (and in lots of cases, they are not) the neural net will not converge.
Scientists are right to not trust their jobs to neural nets, because they'll never have to - neural nets are suitable for a particular domain of problem. Like every individual machine learning and AI technique, it is not a magic bullet. Neither are expert systems (a completely different system). AI may dream of walking, talking, learning robots that can learn anything and solve any problem, but what AI is practically about is things like this - any computing problem that has to deal with uncertainty or a lack of definition. They will assist scientists, not replace them.
--
Spam the secret service - include the words terrorism and sedition in every email!
. As far as I know, an autopilot just keeps the plane on a predetermined course - essentially following an invisible track. These satellites have no track, no predetermined course. Instead, they have general objectives that they then fulfil by keeping a certain orbit and certain configuration.
for anyone who read Marvin Minsky's stuff about how useless robot building is, I would point out that things like this are the endproducts of all that robot building.
Dynamix seem to be making a habit of courting a particular graphics card with each of their Tribes releases. For Tribes I, it was designed to run well on Voodoo chipset cards and crash and burn on other cards. I never played Tribes on my old Riva TNT based system - it sucked even after they'd released OpenGL drivers. My friends with Voodoo IIs - no problem.
This time, I've got a GeForce II, so no probs. for me. For the Voodoo users out there, check the readme file with Tribes II - it gives the URL of a site with the voodoo drivers Dynamix recommends for running Tribes II.
--
If I was an evil overlord, I would not tell my Legions of Terror "And he must be taken alive!" The
command will be "And try to take him alive if it is reasonably practical."
http://minievil.eviloverlord.com/lists/overlord.ht ml
However, there are now international treaties which allow corporations to have such laws struck down by bodies such as the World Trade
Organisation. This essentially removes parts of national sovereignty, placing the power in the hands of an undemocratic body with a history of
siding with the multinationals. Moreover, the treaties are often so broad that any laws that impair profitability of local operations may be
overturned.
IANA(Constitutional)L, but I'm pretty sure that the rule in pretty much every country (at least every one rich enough to exert it's will) is that international treaties are not binding law in a country. A country must choose to ratify an international treaty by passing their own legislation.
In Australia (and America too), this is further complicated by the fact that the federal government must respect the law-making powers of the states. In Australia, the federal government has used international treaties as an excuse to override state legislation, but this isn't a hard-and-fast rule. And this still relies on the federal government implementing legislation that implements the treaty.
Australia has no debt to the World Bank, and has no obligation to honour the WTO. Neither does America (and they DON'T always, I'm sure). For all the international organisations and treaties, what happens in a country is down to that countries' government, and the international treaty organisations have no power except through these governments. The UN and other world treaty organisations often criticise certain aspects of Australian law (quite fairly, in my opinion, though they let the US off way too easy) and the Australian PM's response could be boiled down to 'Who's running this country, anyway?'. THis is standard world practice - as an example, consider dubya's treatment of the Kyoto 'agreement'.
I hope that if a multinational company does complain to the WTO about this, the Australian government will just ignore them. Now if Murdoch or Packer complains, we'll bend over. But hell, that's politics for you.
The three note rule - all music is prior art
on
Digital Copyright
·
· Score: 1
I hope this isn't offtopic - it's an idea I've had for a while, and this seems as good a time as any to suggest it.
A long time ago, there was a copyright hearing - someone or other vs. George Harrison . The judge - presumably a complete musical illiterate - ruled that because George Harrison's song used the same 3-note sequence as the plaintiff's song, this was a breach of copyright.
Now when I heard this, I wondered why George's lawyer didn't do some research and bring up this argument:
The Even-Tempered musical scale contains the notes C, C#, D, D#, E, F, F#, G, G#, A, A#, B. That's twelve notes.
The number of possible three note combinations is therefore 12**3, or 1728.
The entire public domain of classic music in the even-tempered scale (the basis of all western music) dates from J.S.Bach (before 1700 if I remember right) to 1905. In this time, every single three-note sequence would have been used by some composer. Hell, J.S.Bach wrote a series of fugues in every single key, and each of those fugues probably contained every single three-note sequence you would use in that key, so they were probably all used by him. And almost all rock/pop music today relies on the most basic of the note sequences.
The conclusion is, every three note sequence has probably been public domain since J.S.Bach, and is definitely public domain by now. Thus all music based on the well-tempered scale (i.e. all western music on western instruments) is public domain and holds no copyright.
The downsides with this are that it screws over the artists, and you'd have to find that judge that ruled against George Harrison. But It'd be nice to see someone argue that songs based on classical chord sequences (read: every single pop song today) cannot be copyrighted.
"Human beings, who are almost unique in having the ability to learn from the experience of others, are also remarkable for their apparent disinclination to do so."
-- Douglas Adams
The old problem of AI - AI is the field that contains all the computer problems that haven't been worked out yet. As soon as someone solves an AI problem - it's not AI anymore.
Not only has AI got everything to do with games, games are where some of the best AI research and application happens these days. An example - no-one, to my knowledge, has yet developed a bot for counterstrike that plays as well as an average human team member (though from what I've heard, some now get pretty close, by using Neural Nets and machine learning stuff like that). The reason - counterstrike is TOO HARD for most AI researchers! I've just started a PhD in AI, and I might have a shot at it, but it's a bloody hard problem.
Why would counterstrike be harder than chess? Because chess is organized. Each piece is completely defined, the order of play is completely defined, and the possibilities at each stage are completely defined and are quite narrow. Games like Go have completely defined rules, but because there are so many different possible moves at each stage, the search space is huge, and the computer players are thrashed by any competent human player. And Counterstrike, the rules aren't known - there is no hard-and-fast rule about when to use cover, or what is the optimum range for a sub-machine gun. If a game designer makes a game that simulates the real world, designing believeable AI for that world becomes almost as difficult as building a robot to navigate in the real world.
As a point of interest, one of the main robot-based AI interests at the moment is the world robot soccer championships - teaching robots to play soccer. Trying to create 'intelligent' team behaviour is one of the most interesting, and most researched, problems in AI today.
So in summary, generating (or predicting) realistic opponent behaviour (opponents that behave like people) is as good an AI problem as any.
To get back to topic, I applied to Ratbag Software (Powerslide, Dirt Track Racing) for work, and they advertised regularly at my Uni. They wanted Computer Science and Computer Systems Engineering Graduates who were not only top of the class, but who also knew C++ and had already tried game pogramming. So I'd agree with all the comments about getting a degree, and trying things like mods, bots, etc.
I've got the impression that for this to work, they already know exactly what inputs and outputs a lamprey's brain has in its natural state. From the point of the lamprey, it is not on some wierd acid trip - it is in its natural environment.
Part of the reason they can do this, and the reason they're using lamprey eels, is that the structure of the lamprey eels brain is extremely simple compared to us, or even lower life forms such as insects or record company executives (cheap shot - sorry). To feel pain, they need chemical pain receptors - which would be pretty easy to find and remove on an animal like the lamprey.
Most neuro-scientific research, even with humans, relies on the study of the electrical inputs of the brain and nervous system. Because we know the electrical signal of the brain when it is (for example) having a seizure, we can then modify the brain to prevent that signal. There are some great applications of this research for us - rather than modifying your sight and smell, modifying pain signals (or pleasure signals). Using electrical devices to prevent epilepsy or parkinson's disease.
Slashdot Mirror
all the world's secrets
on one server - cracked - not secret.
Just use peer-to-peer.
--
This post is about Truth, Beauty, Freedom, and above all things, Karma.
One of my mates had a gag he used to pull on his friends - he'd get their passwords through word assocation. The conversation would go like this:
JG (my friend): "Let's do one of those word association things - you know, I say a word and you say the first thing that comes into your head."DI (Dumb Idiot): "Oh, okay."
JG: "Chicken"
DI: "Dinner"
JG: "Password"
DI: "********.... oh f**k"
He got it to work a few times - got a few of his mates to say their passwords.
Guess it shows you'll never go broke relying on human stupidity.
--
Spam Echelon, mention Terrorism in all your emails!
Good post, but some things I disagree with.
The state of a neural network, for example, is a black box from which no real insight can be gained...
Not so. The weights of the nodes inside a neural network can, and are analyzed, to determine the features of the input that the network uses to produce a result. I think this is known as cluster analysis - not my specialty.
magine a neural network whose training set is the raw molecular structure data of different diseases (input) and their cures (output). The network trains on this data until the error reduces to zero...
Cool idea, but not possible, at least by today's knowledge.
For one thing, neural nets do not produce output data of this complexity. A neural network has rigidly defined inputs and rigidly defined outputs. I can't see how you could constrain possible structure of a molecule sufficiently to be used as the output of a neural net.
Secondly, a neural net will not work on any possible set of data you feed into it. There have to be common patterns in the data that the net can draw out. The beauty of neural nets is that the person feeding the data in need not know the exact patterns. If these patterns are not present (and in lots of cases, they are not) the neural net will not converge.
Scientists are right to not trust their jobs to neural nets, because they'll never have to - neural nets are suitable for a particular domain of problem. Like every individual machine learning and AI technique, it is not a magic bullet. Neither are expert systems (a completely different system). AI may dream of walking, talking, learning robots that can learn anything and solve any problem, but what AI is practically about is things like this - any computing problem that has to deal with uncertainty or a lack of definition. They will assist scientists, not replace them.
--Spam the secret service - include the words terrorism and sedition in every email!
No. It's a glorified
- pilot
. As far as I know, an autopilot just keeps the plane on a predetermined course - essentially following an invisible track. These satellites have no track, no predetermined course. Instead, they have general objectives that they then fulfil by keeping a certain orbit and certain configuration.for anyone who read Marvin Minsky's stuff about how useless robot building is, I would point out that things like this are the endproducts of all that robot building.
"Robotics is bunk; work in my field", says Minsky.
Just funding envy?
Marv simulates all, says Roboticists should tell
him all the factors,
then simulations will do all the work of a
physical robot.
Marv, the theorist, has missed or forgot the rules of
good engineering,
And roboticists are interested in more than
the theory of mind.
The outside world is complex, and iterative
models do not work
as well as Marv thinks (read some chaos theory to
appreciate this)
To build functional systems (and that is our job)
nothing substitutes
For construction of physical prototypes that work in the real world.
(Would you drive a car that was only tested in
Marv's simulator?
And The Wright Brothers, did not model their first plane,
and yet they could fly!)
Besides, what if the human-observed factors are
not that important?
We postulate that the analysis of mind
The understanding of the mind's environment,is connected with
and to neglect the
environment, or to reduce it, is to waste
the useful data.
Marv: "Roboticists produce no new theories, so
they are wastes of time"
But NASA has not used your stuff for its space probes,
has it now Marvin?
And Biologists do not use your work to help
them understand life.
Our work does not make as many discoveries,
as we have chosen
problems that are more than toy worlds, hypothesis,
and formal systems.
So in summary, if you want signs of Robots
being useful Marvin
Wait until I send my squad of robots around
to kick your behind!
--
Hey, if you mention terrorism, bombs and sedition in your sig, your email will be read by thousands of people!
Dynamix seem to be making a habit of courting a particular graphics card with each of their Tribes releases. For Tribes I, it was designed to run well on Voodoo chipset cards and crash and burn on other cards. I never played Tribes on my old Riva TNT based system - it sucked even after they'd released OpenGL drivers. My friends with Voodoo IIs - no problem.
This time, I've got a GeForce II, so no probs. for me. For the Voodoo users out there, check the readme file with Tribes II - it gives the URL of a site with the voodoo drivers Dynamix recommends for running Tribes II.
--t ml
If I was an evil overlord, I would not tell my Legions of Terror "And he must be taken alive!"
The command will be "And try to take him alive if it is reasonably practical."
http://minievil.eviloverlord.com/lists/overlord.h
However, there are now international treaties which allow corporations to have such laws struck down by bodies such as the World Trade Organisation. This essentially removes parts of national sovereignty, placing the power in the hands of an undemocratic body with a history of siding with the multinationals. Moreover, the treaties are often so broad that any laws that impair profitability of local operations may be overturned.
IANA(Constitutional)L, but I'm pretty sure that the rule in pretty much every country (at least every one rich enough to exert it's will) is that international treaties are not binding law in a country. A country must choose to ratify an international treaty by passing their own legislation.
In Australia (and America too), this is further complicated by the fact that the federal government must respect the law-making powers of the states. In Australia, the federal government has used international treaties as an excuse to override state legislation, but this isn't a hard-and-fast rule. And this still relies on the federal government implementing legislation that implements the treaty.
Australia has no debt to the World Bank, and has no obligation to honour the WTO. Neither does America (and they DON'T always, I'm sure). For all the international organisations and treaties, what happens in a country is down to that countries' government, and the international treaty organisations have no power except through these governments. The UN and other world treaty organisations often criticise certain aspects of Australian law (quite fairly, in my opinion, though they let the US off way too easy) and the Australian PM's response could be boiled down to 'Who's running this country, anyway?'. THis is standard world practice - as an example, consider dubya's treatment of the Kyoto 'agreement'.
I hope that if a multinational company does complain to the WTO about this, the Australian government will just ignore them. Now if Murdoch or Packer complains, we'll bend over. But hell, that's politics for you.
I hope this isn't offtopic - it's an idea I've had for a while, and this seems as good a time as any to suggest it.
A long time ago, there was a copyright hearing - someone or other vs. George Harrison . The judge - presumably a complete musical illiterate - ruled that because George Harrison's song used the same 3-note sequence as the plaintiff's song, this was a breach of copyright.
Now when I heard this, I wondered why George's lawyer didn't do some research and bring up this argument:
The Even-Tempered musical scale contains the notes C, C#, D, D#, E, F, F#, G, G#, A, A#, B. That's twelve notes.
The number of possible three note combinations is therefore 12**3, or 1728.
The entire public domain of classic music in the even-tempered scale (the basis of all western music) dates from J.S.Bach (before 1700 if I remember right) to 1905. In this time, every single three-note sequence would have been used by some composer. Hell, J.S.Bach wrote a series of fugues in every single key, and each of those fugues probably contained every single three-note sequence you would use in that key, so they were probably all used by him. And almost all rock/pop music today relies on the most basic of the note sequences.
The conclusion is, every three note sequence has probably been public domain since J.S.Bach, and is definitely public domain by now. Thus all music based on the well-tempered scale (i.e. all western music on western instruments) is public domain and holds no copyright.
The downsides with this are that it screws over the artists, and you'd have to find that judge that ruled against George Harrison. But It'd be nice to see someone argue that songs based on classical chord sequences (read: every single pop song today) cannot be copyrighted.
"Human beings, who are almost unique in having the ability to learn from the experience of others, are also remarkable for their apparent disinclination to do so." -- Douglas Adams
The old problem of AI - AI is the field that contains all the computer problems that haven't been worked out yet. As soon as someone solves an AI problem - it's not AI anymore.
Not only has AI got everything to do with games, games are where some of the best AI research and application happens these days. An example - no-one, to my knowledge, has yet developed a bot for counterstrike that plays as well as an average human team member (though from what I've heard, some now get pretty close, by using Neural Nets and machine learning stuff like that). The reason - counterstrike is TOO HARD for most AI researchers! I've just started a PhD in AI, and I might have a shot at it, but it's a bloody hard problem.
Why would counterstrike be harder than chess? Because chess is organized. Each piece is completely defined, the order of play is completely defined, and the possibilities at each stage are completely defined and are quite narrow. Games like Go have completely defined rules, but because there are so many different possible moves at each stage, the search space is huge, and the computer players are thrashed by any competent human player. And Counterstrike, the rules aren't known - there is no hard-and-fast rule about when to use cover, or what is the optimum range for a sub-machine gun. If a game designer makes a game that simulates the real world, designing believeable AI for that world becomes almost as difficult as building a robot to navigate in the real world.
As a point of interest, one of the main robot-based AI interests at the moment is the world robot soccer championships - teaching robots to play soccer. Trying to create 'intelligent' team behaviour is one of the most interesting, and most researched, problems in AI today.
So in summary, generating (or predicting) realistic opponent behaviour (opponents that behave like people) is as good an AI problem as any.
To get back to topic, I applied to Ratbag Software (Powerslide, Dirt Track Racing) for work, and they advertised regularly at my Uni. They wanted Computer Science and Computer Systems Engineering Graduates who were not only top of the class, but who also knew C++ and had already tried game pogramming. So I'd agree with all the comments about getting a degree, and trying things like mods, bots, etc.
Hope this has been suitably enlightening.
--
The Milky Bars are on Me.
I've got the impression that for this to work, they already know exactly what inputs and outputs a lamprey's brain has in its natural state. From the point of the lamprey, it is not on some wierd acid trip - it is in its natural environment.
Part of the reason they can do this, and the reason they're using lamprey eels, is that the structure of the lamprey eels brain is extremely simple compared to us, or even lower life forms such as insects or record company executives (cheap shot - sorry). To feel pain, they need chemical pain receptors - which would be pretty easy to find and remove on an animal like the lamprey.
Most neuro-scientific research, even with humans, relies on the study of the electrical inputs of the brain and nervous system. Because we know the electrical signal of the brain when it is (for example) having a seizure, we can then modify the brain to prevent that signal. There are some great applications of this research for us - rather than modifying your sight and smell, modifying pain signals (or pleasure signals). Using electrical devices to prevent epilepsy or parkinson's disease.