Printable View
I agree with that. I think the fundamental issue is that the machines have to start somewhere. That start will be a machine programmed by a person (it was built many years ago, as a matter of fact).
All resulting machines will have this "seed" forever. There is no escaping it. That seed will become a diminuitive aspect of the machine's intelligence, but it will still be there.
I dont think man will be able to create true artifical intelligence untill we figure out how to "code" a human brain or something similar. Why are we trying to build something that already has been built. Nature aliens ;) have already built the ultimate super computer mabe we should figure out how to make that work to it's full potential. Hell, we got a ton of memory, processor and disk stored in our skulls but I can't even send my neighbor an e-mail with it.
I think it's badly worded, not unfounded. My take on what rjlohan said was that it's goals will be based upon goals set by programmers. That doesn't mean it can't develop new goals, but that the goals are what a programmer thought was important. The programmer could have been wrong which would send the computer careening in the wrong direction. As the computer "learns" from its mistakes, it makes better choices. Better is subjective though and until the end of time comes, there is no way to know if it was perfect or not.Quote:
Originally posted by HarryW
Is this just a completely unfounded statement? It's looking that way to me.
That's fair enough, but does it make a difference what its goals are? Does it even matter if it achieves them? 'Solving the problem' doesn't seem like the important aspect here, it seems like the cognition is the important bit.
I think there is only relative "perfect"
I dont think there is an absolute perfect.
Well, I'm not deep enough to deal with the actual question at hand, but I still like to talk. So I basically just move the subject to where I can talk about it. It's up to the guy who wants the answer to redirect the topic. :)
Ok, badly worded... I should have learned by now not to do 20 things at once, and try and get involved in one of these threads... :p
Yes, my point was that a computer can only work towards goals set by some external programmer. It doesn't have the capability to comprehend that some action bore negative consequences. Because, it has no idea of right or wrong, I guess, to put it one way.
Also, evolution works through birth, which computers certainly don't do. The only way we could even pretend this was possible was that a computer built another computer. But that computer (or one somewhere far above it) was built by humans, and as such, it will have developed no more since that point. Perhaps, it makes some random change to every computer below it, still, the goals it gives those computers could only be ones set by the original human programmer, so we have evolution in one sense - in that the computers may eventually come to some incarnation that achieves some goal 100%, but it is still the original goal, set by humans, and not developed in any way over generations.
I don't think computers will ever have the ability to develop through experiential learning, which is the crux of learning. The computer can be told as many facts as we know, and given as many goals as we want, (disregarding lack of memory space), but none of that will ever change, until another human comes along and makes some alteration.
And that will only ever be relative to humans, not computers.Quote:
Originally posted by Aerials
I think there is only relative "perfect"
I dont think there is an absolute perfect.
1) Yes, I think there is nothing wrong with continuing to develop and improve. You can't stop people from doing it anyway. It seems like an inate drive for us to create.Quote:
Originally posted by QWERTY
HI All.
I am writing a research paper on aritificial intelligence and while the whole thing (the paper that is) is basically done, my teacher asked me to write something about moral/ethical aspects of creating an intelligent machine (and at this point I assume a situation in which we are in fact able to create an intelligent machine). The questions she wants me to answer are things like: 1. Should we create an intelligent machine (in moral/ethical sense)? 2. Would the machines have the same rights as humans do? These are the two biggies I need to answer. If you could share your ideas about it that would be great; it would be even better if you know any books/journals/periodicals/magazines or things of that nature I could reffer to, but mainly I'm looking for what you have to say about it. Thanks for all your help.
Q
2) What rights? Human rights are the rights we've taken for ourselves. If a machine becomes capable of demanding rights and taking measure to ensure they have those rights, then yes. But nobody "gives" rights, except in a legal sense which is entirely artificial. Rights are taken.
Right and wrong are (putting religion aside) relative. There's no absolute. It's just 'what works for most of us'. As for the comprehension of consequences, that's looking like another unfounded statement.Quote:
Originally posted by rjlohan
Yes, my point was that a computer can only work towards goals set by some external programmer. It doesn't have the capability to comprehend that some action bore negative consequences. Because, it has no idea of right or wrong, I guess, to put it one way.
In humans, yes. Big deal. Surely birth isn't a prime requirement for consciousness?Quote:
Also, evolution works through birth, which computers certainly don't do.
I don't think 'goals' are the pertinent issue here.Quote:
The only way we could even pretend this was possible was that a computer built another computer. But that computer (or one somewhere far above it) was built by humans, and as such, it will have developed no more since that point. Perhaps, it makes some random change to every computer below it, still, the goals it gives those computers could only be ones set by the original human programmer, so we have evolution in one sense - in that the computers may eventually come to some incarnation that achieves some goal 100%, but it is still the original goal, set by humans, and not developed in any way over generations.
Uhh... they already do.Quote:
I don't think computers will ever have the ability to develop through experiential learning, which is the crux of learning.
[/quote] The computer can be told as many facts as we know, and given as many goals as we want, (disregarding lack of memory space), but none of that will ever change, until another human comes along and makes some alteration. [/QUOTE]Learning in humans is not some magical process that happens by accident you know. There is a biomechanical process there, unless you think it's all come from some supernatural being. We already have systems that learn, to a rudimentary (in comparison to humans) level.
Absolutely. I couldn't agree more.Quote:
Originally posted by cafeenman
2) What rights? Human rights are the rights we've taken for ourselves. If a machine becomes capable of demanding rights and taking measure to ensure they have those rights, then yes. But nobody "gives" rights, except in a legal sense which is entirely artificial. Rights are taken.
This is undeniably true. So why don't we concentrate our efforts on making people better? What I mean is more resources for education. I know several teachers that have been in the educational system for a long time that don't get paid squat. Not a lot of incentive to do a better job.Quote:
Originally posted by Aerials
Id like to say taht humans have great potential they just dont use it.
I wasn't referring to right and wrong in a moral sense, but rather what approaches to a problem are appropriate and most effective to achieve the goal.Quote:
Originally posted by HarryW
Right and wrong are (putting religion aside) relative. There's no absolute. It's just 'what works for most of us'. As for the comprehension of consequences, that's looking like another unfounded statement.
Did you read the rest of that paragraph? The point is that evolution works through some small change in a geneset, which is achieved through birth, and that change may or may not prove to be more effective in helping that individual attain some goal. Since computers don't have this capability, excepting my example of a random machine-generated change, which can only be as complex as a human lets it, they can't evolve. And since their goals are completely artificial, and based on human perception, they can't learn. At the most, all we wil get is a bunch of cyborgs that do stuff we let them.Quote:
Originally posted by HarryW
In humans, yes. Big deal. Surely birth isn't a prime requirement for consciousness?
Of course they are. What else is learning? Acquiring knowledge? Well, computers can certainly do that, by us feeding them databases full of facts. The computer can't reason, it is not aware. It can only be those things through human instruction. And that is certainly not evolution.Quote:
Originally posted by HarryW
I don't think 'goals' are the pertinent issue here.
Such as?Quote:
Originally posted by HarryW
Uhh... they already do.
I'm aware of that. I view humans as machines too, and I don't believe in supernatural crap or any sort of god. Unless we somehow develop the technology to replicate all that goes on in the human brain, AI will never happen.Quote:
Originally posted by HarryW
Learning in humans is not some magical process that happens by accident you know. There is a biomechanical process there, unless you think it's all come from some supernatural being. We already have systems that learn, to a rudimentary (in comparison to humans) level.
But if your whole argument is that this is all relative, there is no argument is there? My opinion will be different from yours, and that's the end of it.
Very true. But that's another argument altogether... :pQuote:
Originally posted by cafeenman
This is undeniably true. So why don't we concentrate our efforts on making people better? What I mean is more resources for education. I know several teachers that have been in the educational system for a long time that don't get paid squat. Not a lot of incentive to do a better job.
I have 3 relatives involved in schools out here, and the pay is horrible... one of them is me g/f too - which means I've gotta get a 6-figure salary if I ever want that yacht I've had my eye on... ;)
First off, neural networks combined with back propagation/genetic algorithms can do a pretty good job of learning in a very tight domain, given the framework to do it in.
I don't know project names for the research projects being done into AI. There are hundreds of them. Not necessarily based on neural networks and genetic algorithms, but on facts, rules and meta rules in a knowledge base, with heuristics involving fuzzy logic for rational decision making. If you go and look, you just might find something. Have you ever done any academic study on AI? If not, perhaps you might consider it.
Basically, you seem to be going down the "nothing can be greater than its creator" route, which is something I just don't agree with. In a way, it seems like wishful thinking to me.
Evolution and learning are two completely seperate issues. Evolution is not learning. Learning is not evolution.
Is learning acquiring knowledge? Well, yes! What else would it be? I really don't see why goals are meant to be so important in this context. You seem insistent that this can only be done by spoon-feeding the system raw data, but I really don't think you have any grounds to base this on. Can't computer systems collect data? Can't they interpret data and get information? Can't they reason with that information and their existing knowledge and infer knew knowledge? I suspect your answer to that last question is 'no'. Perhaps I'll be surprised.
Every system needs to be bootstrapped. Computers aren't going to suddenly become intelligent without our help. Just because they're not at the moment, though, doesn't mean they never will be.
The fact is, humans are intelligent. Where did we come from?
Big Blue, the computer who beat the best human chess player, is not really an example of AI. It was done primarily with number crunching and a Von Neuman Mini-Max algorithm.
That being said, it is interssting to note that none of the programmers on the project were world class chess players. Big Blue played way above the level of the ability of its creators.
The above suggests that future computers might far outstrip the goals and abilities of the programming staff that created them.
The only concept I feel strongly about is that AI will not be attained with any current system architecture. Array procesors with more and faster processors built using current basic design concepts will not hack AI.
I have no problem believing a computer can learn and become smarter. That just makes sense to me. But I have a problem with a non-living entity having self-awareness or "rights."
Sort of like that really bad movie, Bicentenial Man that came out a few years ago. Kind of reminded me of the movie Indecent Proposal where they were obviously trying to create a public reaction and nobody cared.
I think it's an interesting topic, but I don't think machines will ever become anything that people see as human equivalents.
I think we should focus on trying to give birth to true AI, rather than focusing on how to use genetical engineering to advance the development of our own minds.
Not to any useful extent. I have considered it, but I barely have time to scratch me arse these days... :pQuote:
Originally posted by HarryW
First off, neural networks combined with back propagation/genetic algorithms can do a pretty good job of learning in a very tight domain, given the framework to do it in.
I don't know project names for the research projects being done into AI. There are hundreds of them. Not necessarily based on neural networks and genetic algorithms, but on facts, rules and meta rules in a knowledge base, with heuristics involving fuzzy logic for rational decision making. If you go and look, you just might find something. Have you ever done any academic study on AI? If not, perhaps you might consider it.
How do you arrive at that conclusion? I don't think humans are smart enough to make something smarter than humans. I don't think that's the same thing. We don't know squat about the brain. Not nearly enough to begin to 'redo' it. All of those logic and reasoning patterns you have mentioned, I would hazard to guess are just somebody's (or many somebodys) best representation of thought processes. I doubt they are even close, and to develop those things to the extent of recreating a 'brain' seems to me a long long way away, although as I said, I have no AI research to support that. Merely opinion.Quote:
Originally posted by HarryW
Basically, you seem to be going down the "nothing can be greater than its creator" route, which is something I just don't agree with. In a way, it seems like wishful thinking to me.
I never disagreed. I guess we should concentrate on one aspect at a time, huh?Quote:
Originally posted by HarryW
Evolution and learning are two completely seperate issues. Evolution is not learning. Learning is not evolution.
But it is what it does with this knowledge that is the key. Simply 'knowing' facts is no hard task. Any machine can be given 'facts' in some way. But I think learning involves more than gathering information. It also involves developing approaches to using that information. And that, is what I don't think computers will ever be able to do on their own. They work on pre-defined algorithms, or pattterns of thought, they don't develop their own through experience, because they have no inherent way of developing goals, with which they can measure their performance.Quote:
Originally posted by HarryW
Is learning acquiring knowledge? Well, yes! What else would it be? I really don't see why goals are meant to be so important in this context. You seem insistent that this can only be done by spoon-feeding the system raw data, but I really don't think you have any grounds to base this on. Can't computer systems collect data? Can't they interpret data and get information? Can't they reason with that information and their existing knowledge and infer knew knowledge? I suspect your answer to that last question is 'no'. Perhaps I'll be surprised.
Does my point make sense yet? Regardless of whether you agree with it?
I'd be very reluctant to sy 'never'. But I guess I have above. I think we can have a simulation of AI that will do whatever it was we wanted it to do - probably choose the next target for a military strike... :rolleyes:Quote:
Originally posted by HarryW
Every system needs to be bootstrapped. Computers aren't going to suddenly become intelligent without our help. Just because they're not at the moment, though, doesn't mean they never will be.
But, because computers can't develop their own goals, as I said, I don't think we will ever have true AI. That is opinion, nothing more.
Don't start on that... :pQuote:
Originally posted by HarryW
The fact is, humans are intelligent. Where did we come from?
Any particular reason?Quote:
Originally posted by Nucleus
I think we should focus on trying to give birth to true AI, rather than focusing on how to use genetical engineering to advance the development of our own minds.
:p
Well, I just don't agree with your opinion. I see a lot of potential for AI. Milestones for what constitutes AI get set, are achieved and then get redefined and moved further away all the time. It seems people are very attached to the concept of 'intelligence' and 'consciousness' and just don't want machines to be the same as them. Personally, I think the reason people have such a hard time defining things like 'self-awareness' is that they are very busy making sure that humans are special.
Nope, I just blurt out random comments sometimes, it kind of like a refined version of Turrets syndrome; don't pay any attention to them as they usually add nothing to the discussion. :)Quote:
Originally posted by rjlohan
Any particular reason?
:p
Blip to 20 years later...AI robots take over the home...
"Today's feature on Tom's Hardware Guide....Overclocking Your Maid" ;) ;)
That's just bull****. Just because I don't think we can do it doesn't mean I don't want it to happen. In reality, I don't really care. I think we've got more important things to deal with first. Automating the whole world isn't going to do any good for us (humans) at this point in time. But that's a social issue, and probably won't change either, so why not go for AI.Quote:
Originally posted by HarryW
Well, I just don't agree with your opinion. I see a lot of potential for AI. Milestones for what constitutes AI get set, are achieved and then get redefined and moved further away all the time. It seems people are very attached to the concept of 'intelligence' and 'consciousness' and just don't want machines to be the same as them. Personally, I think the reason people have such a hard time defining things like 'self-awareness' is that they are very busy making sure that humans are special.
My definition of self-awareness would be something along the lines of being aware you exist, and can make choices. It may not be the same as yours or the next person, but I've got no problem finding one.
You arrive at some pretty ridiculous conclusions based on some completely different remarks.
LOLQuote:
Originally posted by jpbtennisman
Blip to 20 years later...AI robots take over the home...
"Today's feature on Tom's Hardware Guide....Overclocking Your Maid" ;) ;)
:D
I agree with rjlohan's response to this comment of yours. I think your are trying to reduce our arguments to us not wanting it to happen. I don't believe it can happen. I don't know enough about the implications of it happening to care if it happens or not or to try to impede it for that matter.Quote:
Originally posted by HarryW
Well, I just don't agree with your opinion. I see a lot of potential for AI. Milestones for what constitutes AI get set, are achieved and then get redefined and moved further away all the time. It seems people are very attached to the concept of 'intelligence' and 'consciousness' and just don't want machines to be the same as them. Personally, I think the reason people have such a hard time defining things like 'self-awareness' is that they are very busy making sure that humans are special.
How many people do you know that are trying to "find" themselves? I know a few and consider them all to be pretty flakey. These are the kind of people that spend all their time totally self-involved and never really achieve any kind of "higher consciousness." In my mind, they are "lost."
The rest of the people I know simply don't worry about it because it's not possible to find the eternal truths they are looking for. So as a hobby, it's interesting. As a way of life, it's a waste of time.
People don't know anything about anything. We just like to think we do. But people are the ones who are going to create these machines so they will be very far from perfect. I think the very idea of a machine becoming "self-aware" in fundamentally flawed. Unfortunately, I'm not really able to express why.
I put it in the same category as:
The chicken or the egg
Perpetual motion
Creating a generator that powers a motor that runs the generator
Travelling in time and meeting yourself
I would like to point out that there are respected scientists out there who argue that consiousness itself is uncomputational. Roger Penrose believes, taking much from Godel that there are things that consiousness can know that cannot be proved logically.
I think Guv was right when he said that our current architcecture is inadequate for the job. We will need to formulate a new logic that is capable of deriving all truth without contradicting itself. A form of logic that is so inconceivable to us now that it defies the imagination.
This is the sort of thing that makes me think people are just resistant to it. Why would you think it's impossible unless you think consciousness is some kind of mystical gift? It smacks of faith to me. If not faith, then denial.Quote:
Originally posted by cafeenman
I think the very idea of a machine becoming "self-aware" in fundamentally flawed. Unfortunately, I'm not really able to express why.
I don't mean to be presumptuous about the roots of your opinions, but you just don't seem to back them up with any reasonable evidence. Just more opinion. That's fine if you just say it's your opinion and nothing more, but it irks me somewhat to see people make blanket statements about a field I feel they possibly haven't explored properly and present a pseudo-logical argument for their opinion that ends up, on further investigation, being based purely on more opinion.
Making choices is easy, but how do you test for being aware you exist? This is the kind of target that gets set, and when it seems to have been achieved, the goalposts get moved or the objectives redefined by people who are basically never going to leave them alone because (IMO) they feel the need to set themselves apart from other machines.Quote:
being aware you exist, and can make choices
Harry
Read "The emporer's new mind" by Roger Penrose. He provides a logical objection to the possiblity of AI ever achieving true consiousness and his arguements do not rest on the notion of dualism.
Harry you would look so much better without the beard, and after a diet ... just a little advice....
Here's a brief summary taken from here that describes his position:
n.b. "Shadows of the mind" was his first book, "Emporer's new mind" the sequel.Quote:
In his book Shadows of the Mind, Roger Penrose suggests that deep problems in artificial intelligence, physics, and the philosophy of mind are closely connected. He presents a detailed argument, using Gödel's theorem, for the conclusion that human thought cannot be simulated by any computation. This leads him to the conclusion that physics is noncomputable, and he presents suggestions about how noncomputability may enter into a theory of quantum gravity. Finally, he argues that this may take effect at the level of the mind through quantum collapse processes in microtubules, protein structures found in the skeleton of a neuron.
That is my point. Everything that ANYBODY has to say regarding this is nothing more than opinion, theory and hypothesis, regardless of what research has been done. If there were "facts" as you call them, then this thread would be entirely unnecessary. You ask me to provide reasonable evidence. How do you prove something isn't?Quote:
Originally posted by HarryW
I don't mean to be presumptuous about the roots of your opinions, but you just don't seem to back them up with any reasonable evidence. Just more opinion. That's fine if you just say it's your opinion and nothing more, but it irks me somewhat to see people make blanket statements about a field I feel they possibly haven't explored properly and present a pseudo-logical argument for their opinion that ends up, on further investigation, being based purely on more opinion.
I very strongly DO NOT believe in the mystical. Again, if you read what I read, I said that I do not believe that it can happen. I also do not believe there is a God. That doesn't make me resistent. It means I don't believe it. Can I say that any clearer???? I do not believe it. If I'm wrong and someone can show me proof then I will believe it. It's not resistence. It's disbelief. They are not the same thing.
So please explain to me the path from where we are now to when these machines become self-aware. What is the moment that defines that event?
Cafeenman
Have I finally made it on to everyone's "ignore" lists now? Read Penrose. He attempts to prove that it is impossible.Quote:
How do you prove something isn't?
That's a little unfair of you. If Harry knew the path, he would probably have already achieved it.Quote:
So please explain to me the path from where we are now to when these machines become self-aware. What is the moment that defines that event?
Exactly my point simonm. If it could be proven then it would be done.
Pretty much anything a person says is opinion. It's entirely unnecessary and irritating to listen to people who preface every thought with "In my opinion." The only time someone wants another to do this is when they don't agree. For some people agreement = fact, disagreement = other person's opinion vs. my fact.Quote:
I don't mean to be presumptuous about the roots of your opinions, but you just don't seem to back them up with any reasonable evidence. Just more opinion.
So far I haven't seen a shred of compelling evidence that it's possible to create a self-aware machine. It's all been conjecture and, to reverse your statement, an irrational desire for it to be possible vs. actual data supporting it.
Cheers Simon, I'll try to remember to check it out. I'll be back to uni for my final year in September, and this is the sort of stuff that is likely to be the subject of my dissertation, so all input is welcome.
As Simon pointed out, it's really quite unlikely that I know the answer to this :rolleyes: There is an issue with what self-awareness is though. If you believe that we are just complex machines, then you have already proven to yourself that intelligence is possible. It's the 'artificial' bit people aren't happy with. If what Simon says Penrose proposed is true, then what would an intelligent system based on those principles be called? My gut feeling is that many people would claim it somehow wasn't 'artificial' enough to be AI.Quote:
So please explain to me the path from where we are now to when these machines become self-aware. What is the moment that defines that event?
OK, so I assume from your posts that you believe this is possible. I do not. There is not a way to prove it either way. At least I don't have an argument that will prove it is not possible. Do you have one proving it is? To me proof is actually doing it.Quote:
Originally posted by HarryW
B]As Simon pointed out, it's really quite unlikely that I know the answer to this :rolleyes:[/B]
Harry
Then you'll definitely want to read some Penrose. He is one of the leading "objectors" to the development of self-aware AI.Quote:
Cheers Simon, I'll try to remember to check it out. I'll be back to uni for my final year in September, and this is the sort of stuff that is likely to be the subject of my dissertation, so all input is welcome.
I beleive that to make a self-aware machine, it would have to have the capacity to step "outside" of itself and be able to conceptualise about itself in the third party. For any logical, analylitic and algorythmic device, this kind of self-consideration is impossible as it leads to an infinite regress. Such a machine needs to make an intuitive leap to grasp the concept of itself. And it is a leap because the gap is cannot be systematically bridged.
Here is one of Penroses arguments regarding the uncomputability of the human consiousness:
Quote:
(1) Assume my reasoning powers are captured by some formal system F (to put this more briefly, "I am F"). Consider the class of statements I can know to be true, given this assumption.
(2) Given that I know that I am F, I know that F is sound (as I know that I am sound). Indeed, I know that the larger system F' is sound, where F' is F supplemented by the further assumption "I am F". (Supplementing a sound system with a true statement yields a sound system.)
(3) So I know that G(F') is true, where this is the Gödel sentence of the system F'.
(4) But F' could not see that G(F') is true (by Gödel's theorem).
(5) By assumption, however, I am now effectively equivalent to F'. After all, I am F supplemented by the knowledge that I am F.
(6) This is a contradiction, so the initial assumption must be false, and F must not have captured my powers of reasoning after all.
(7) The conclusion generalizes: my reasoning powers cannot be captured by any formal system.
I reckon a good starting point to define the self-awareness is when the computer does something completely unexpected that can in no way be traced to any human input. Not something it learned as a result of human-defined algorithms, or any step down the ladder from there, but something completely unexpected.
And since that's never gonna happen, we will never see true intelligence in a machine.
Unless, as I said, someone replicates a human brain, but if we're goin' there, why not just clone a human, and work on some sort of training scheme to make them effectively our slave... oh no - that's been outlawed, right? :p
rjlohan
What about some sort of computer based on quantum effects? The quantum world is probablilistic, right? So, by definition, such a system may demonstrate unpredictable behaviour.Quote:
I reckon a good starting point to define the self-awareness is when the computer does something completely unexpected that can in no way be traced to any human input. Not something it learned as a result of human-defined algorithms, or any step down the ladder from there, but something completely unexpected.
And since that's never gonna happen, we will never see true intelligence in a machine.