|
Thank's a lot for the link's DrainX!
I'm very interested in the possibilites and limits of AI, and has been doing some simple programming with it.
HeadBangaa, your are right that computers can't be "smarter" than humans when they just follow instructions to perform computations, but there are other possibilities. The only one that I know of is "neural networking" where you simulate the connected cells in a brain.
http://en.wikipedia.org/wiki/Artificial_neural_network
|
On November 02 2007 07:21 jtan wrote:Thank's a lot for the link's DrainX! I'm very interested in the possibilites and limits of AI, and has been doing some simple programming with it. HeadBangaa, your are right that computers can't be "smarter" than humans when they just follow instructions to perform computations, but there are other possibilities. The only one that I know of is "neural networking" where you simulate the connected cells in a brain. http://en.wikipedia.org/wiki/Artificial_neural_network Even then, it still seems like a paradox.
If a programmer designed a mechanism which produces better-than-human results, then the moment the design was completed, the mechanism loses that attribute. Human knowledge was increased, and it's still just a really fast human-mind. An inductive proof seems appropriate here.
I don't see how a virtual neural network could be instantiated without being completely deterministic, unless modern computing itself was turned on its head.
|
On November 02 2007 07:45 HeadBangaa wrote:Show nested quote +On November 02 2007 07:21 jtan wrote:Thank's a lot for the link's DrainX! I'm very interested in the possibilites and limits of AI, and has been doing some simple programming with it. HeadBangaa, your are right that computers can't be "smarter" than humans when they just follow instructions to perform computations, but there are other possibilities. The only one that I know of is "neural networking" where you simulate the connected cells in a brain. http://en.wikipedia.org/wiki/Artificial_neural_network Even then, it still seems like a paradox. If a programmer designed a mechanism which produces better-than-human results, then the moment the design was completed, the mechanism loses that attribute. Human knowledge was increased, and it's still just a really fast human-mind. An inductive proof seems appropriate here. I don't see how a virtual neural network could be instantiated without being completely deterministic, unless modern computing itself was turned on its head. yeah but that's just the thing! It is essentially deterministic, but we could never predict what structure will evolve inside the network in practice.
Imagine that you plug in a camera that films my handwriting. You check the output of the system and when the system output's the right letters, you give a positive signal backwards strengthening the connections used for coming up with that letter. This way the system will keep rearranging itself and improve over time. Here is an example of pretty much what I described, it really works: http://www.codeproject.com/library/NeuralNetRecognition.asp
Also, beeing a materialists and I think that the human mind also is a deterministic in the exact same way.
|
On November 02 2007 07:45 HeadBangaa wrote:Show nested quote +On November 02 2007 07:21 jtan wrote:Thank's a lot for the link's DrainX! I'm very interested in the possibilites and limits of AI, and has been doing some simple programming with it. HeadBangaa, your are right that computers can't be "smarter" than humans when they just follow instructions to perform computations, but there are other possibilities. The only one that I know of is "neural networking" where you simulate the connected cells in a brain. http://en.wikipedia.org/wiki/Artificial_neural_network Even then, it still seems like a paradox. If a programmer designed a mechanism which produces better-than-human results, then the moment the design was completed, the mechanism loses that attribute. Human knowledge was increased, and it's still just a really fast human-mind. An inductive proof seems appropriate here. I don't see how a virtual neural network could be instantiated without being completely deterministic, unless modern computing itself was turned on its head. By including true random numbers in its inputs.
True randoms (at least based on quantum randomness) aren't hard to generate with the proper hardware. Some CPUs have them. Randomness is injected by "noisy" inputs such as microphones and cameras as well.
Anyway, I don't see determinism as relevant. Predictability and power to reach useful conclusions are the relevant considerations in comparing a machine intelligence to a human intelligence.
|
^ I think the distinction between true randomness and predictability is important theoretically, if not pragmatically. edit: I didn't know that "true randoms" could be generated programmatically.. edit2: oh snap, physics to the rescue
|
On November 02 2007 08:13 HeadBangaa wrote: ^ I think the distinction between true randomness and predictability is important theoretically, if not pragmatically. edit: I didn't know that "true randomness" had been achieved. "true randomness" through quantum mechanics is very much in the use and development, but there's still some discussion whether it is truly random or if it depends on local variables. There are some solid counterarguments vs the local variable theory, but some still defend it etc.
|
Computers and brains alike can only be as inwardly non-deterministic as their component particles allow.
If the universe is deterministic, then so is the human brain.
|
Read the most recent Hugo-Award Winning Novel: http://vrinimi.org/rainbowsend.html Rainbows End by Vernor Vinge (He's awesome) It deals with several interesting issues with a society approaching technological singularity.
|
Omg these videos are good. Just ordered some of the recommended readings.
|
|
It always boggles me how exactly the machines will be learning everything. I've been bothered by the idea that they can continually increase their own capability and I'm always wondering why people think we're so close to creating a computer of intelligence comparable to humans. It's not that it is impossible, but there're some things the computer will have to do that I'm not sure we're clear on. To be truly able to advance itself the computer will have to "understand". But how will it "understand"? It has to have some sort of paradigm by which to interpret information. Where will it get that if not from humans? But our interpretations aren't ideal, they're limited by who we are biologically. Can you come up with a superior method of interpreting data from only taking an inferior method to its limits?
|
Currently the number of processing units in the best computers are nowhere near that is in a human brain. But that is taking human brain celluar function as a model. Intelligent computers is foreseeable however because the capacity of computers can increase indefinately with new designs how electronics is scalable, whereas human brain power is limited biologically. I agree though it nothing 'around the corner'.
|
"AI" in the future will just be incredible fast machines, to the point of appearing intelligent. But if they do something we dont want, it will just be a system failure. I think they will still need some programming. Makes me wonder, what the hell is thought proccesing in the bio-world? Is it programmed or some kind of "as we go along" selfprogramming? Someone mentioned magic. I think its magic how some billions of stupid braincells with chemical signalling, somehow via cooperation can produce thoughts, its just so weird I really cant see us humans make a simular system. I might add, Im neigher a believer or atheist but an agnostic, meaning I realise I dont know shit and never will, so might as well quit thinking about it. Lol.
|
With the "Evolution of Clock" topic posted, I think its quite possible that a fast computer in the future can run a simulation combining millions of variables in the world with "survival of the fittest" conditions on vehicles, machines, intelligent beings, buildings.... eventually forseeing the evolutions in the future on almost everything.
This could play a part in technology singularity, an exhaustion method in "evolution simulation" to increase its own "intelligence"
|
On November 03 2007 01:33 noob4ever wrote: "AI" in the future will just be incredible fast machines, to the point of appearing intelligent. But if they do something we dont want, it will just be a system failure.
Not necessarily. http://en.wikipedia.org/wiki/Artificial_neural_network It might take a while to understand, but it's possible for computers to work in about the same ways as brains. It has been done, and a lot of research is going on. The main problem right now is to make networks big and fast enough. A human brain has ~10^11 neorons with ~7000 connections each.
I think they will still need some programming. Makes me wonder, what the hell is thought proccesing in the bio-world? Is it programmed or some kind of "as we go along" selfprogramming? Someone mentioned magic. I think its magic how some billions of stupid braincells with chemical signalling, somehow via cooperation can produce thoughts, its just so weird I really cant see us humans make a simular system. The brain is an extremely complex system, but it's not as hard as you might think to create. The DNA-code that builds the brain is not very long, it basicy describes how to build and connect one neuron and then "build 67482959432 more of those". And what you call selfprogramming is essentially connections between neurons being strengthend weakend or destroyed as a consequence of the strength of the signals passing through them.
I might add, Im neigher a believer or atheist but an agnostic, meaning I realise I dont know shit and never will, so might as well quit thinking about it. Lol. I find that to be a pretty pessimistic world view
|
On November 03 2007 01:33 noob4ever wrote: I might add, Im neigher a believer or atheist but an agnostic, meaning I realise I dont know shit and never will, so might as well quit thinking about it. Lol.
Agnosticism is not a third way. You're either a theist or an atheist.
|
On November 02 2007 04:21 Eniram wrote: You cannot stop judgement day.. Only delay it.
|
On November 03 2007 04:44 cava wrote:Show nested quote +On November 02 2007 04:21 Eniram wrote: You cannot stop judgement day.. Only delay it. Actually the great day of the singularity sounds pretty much like an armageddon when the people in the videos speaks of it. Maybe they can get some research funding from the religios right haha
|
this verges on the turing machene thread, and whether a machine can replicate a human, "Shadows of the Mind" by Penrose is good for this.
As for artificial neural networks (ANN), I at least know they are used in high energy physics and out perform anything there (including humans).
Apparently, there are some stages Kurtwiel gives for replicating brains with computers, and a step before a human brain is a mouse brain. IBM have done this, making a computer big enough to simulate all the connections of a mouse brain.
|
Interesting. What exactly is the application in physics are are talking about? Got any source?
|
|
|
|