 |
|
|
 |
|
|
|
|
newbie
[Guest] |
| | | | | | posted 6/12/2006 18:05 |  | | | | | | | | | Firstly, by true AI I mean a program that’s self-aware. Is it necessary for a program to be ‘conscious’ (I have to use this word but don’t really want a debate on its definition).
An AI program that can simulate intelligence to an extent where it can match the human intellect (long way off at the moment), such a program may not be a true AI but simply a program following an extremely well structured set of rule.
If this is the case do we need a true AI or will a ‘semi’ AI suffice. I would be capable of performing any task anyway, why would it need consciousness - discuss
| | |
| | |
tkorrovi
[Guest] |
| | | | | | posted 6/13/2006 00:49 | | | | | | | | | | Unfortunately, no set of rules, no matter how well structured, is sufficient, because in the real world the new rules would be created all the time. Therefore, the "true" AI must be as flexible as its environment, which are we, the people, and this is the most important.
| | |
| | |
newbee
[Guest] |
| | | | | | posted 6/13/2006 08:20 | | | | | | | | | | i agree, but new rules can be added by a the program itself, it doesnt have to be a true AI to do this - just following the existing rules can lead to it creating others. The program would simply make the best decision it could with the information it has i.e. it goes for the one that fulfills its rules/needs/desires best. so why does a program need to be truely intellegent to do this? and if it doesnt, why do we need a true AI anyway?
| | |
| | |
tkorrovi
[Guest] |
| | | | | | posted 6/13/2006 14:33 | | | | | | | | | | What can I say... flexibility is about everything, also the rules by which new rules will be created. I must admit that I personally am not so much interested in debate here, who understands, understands, who doesn't, doesn't.
| | |
| | |
newbee
[Guest] |
| | | | | | posted 6/13/2006 18:32 | | | | | | | | | | no im not starting a debate, and i do understand: your point is that rules by nature are not flexable and need to be upheald, this means that any program following them is also not flexable - i agree with this, i just think that if the rules are well enough designed they can simulate intellegence but it could never be anything more. but if a program could simulate intellence is there a need for a true AI? correct me if ive misinterpreted any of this but in your opinion it seems that you dont think an AI can simulate intellence through rules and would need to be a true AI to achive this. is this assumption correct?
| | |
| | |
Ravan
[Guest] |
| | | | | | posted 6/22/2006 07:53 |  | | | | | | | | | I do not believe that any good AI will suffice for the future... What I mean to say is, it takes true AI to fully function in an ever changing world. No set of rules can fully scape it, even if it posesses the ability to create new rules, which by the way, is bridging on true AI to me. What is the bridge to True AI, anyway... that one will always ellude me.
| | |
| | |
lordjakian
[Guest] |
| | | | | | posted 2/26/2007 04:27 | | | | | | | | | | If this is the case do we need a true AI or will a ‘semi’ AI suffice. I would be capable of performing any task anyway, why would it need consciousness - discuss
I don't understand what it is you are asking. It sounds like to me you are wondering if you want a tool or an equal?
If you are capable of doing anything, why is it necessary to build an "equal" since you can already do it? The "consciousness" is not needed. All that is needed is a powerful enough tool, for you to accomplish what ever it is "you" need.
Is this the direction of your logic?
Interesting question. It is funny. It reminds me of someone saying that don't need anybody else, that they can do it all on their own. Me personally, I enjoy the idea of helping others and others helping me.
No one can say either way of thinking is wrong. Its more of a type of lifestyle. Every cat has its rat.
| | |
| | |
Kinnison |
| | | | | | posted 9/4/2008 01:51 | | | | | | | | | | One very good reason for someone to be interested in AI at all, is the benefit it can bring to mankind.
To be of benefit, it doesnt have to be self-aware, just intelligent and as intelligent as us.
I suppose you know of the term "singularity", it is about what will happen when an AI smarter than humans is created, and it has the ability to redesign itself to be even smarter, and as soon as it does, that increases its ability of redesign even further, and so on and so forth (recursively self-improving intelligence). Obvisously that would change the world as we know it completely. Because it would reach heights of intelligence that are own intelligence is not smart enough to imagine. It would be much like science-fiction-come-to-life, but it would be fiction that we cannot even begin to imagine.... so its called a "singularity" that, if it happens, will be impossible to predict.
That could happen either with the appearance of a self-aware or a non-aware intelligence.
Our only chance to know if it is self-aware would probably be during its design and creation. If we actually brought it into existence, and it was as good as we would like it to be, its *bang* its all over. We would not be able to ever know if it is self-aware, just as we cant really know for other humans other than by extrapolation of our own experience. But as far as its usefullnes to us, that really wouldnt matter!
For me, the goal of a general AI (and that is the one branch of the AI field that I am interested) is simply to become smarten than us, in order to solve for us the problems we are facing. I feel strongly fond of the goals of the "singularity institute for artificial intelligence" - SingInst.org. And those are the following:
if such a thing comes to life, and its ethics/morality is not such that it matches our own, then it would be the end of humanity by all probabities.
that AI would go about doing what it deems best, but in the process not care if it has to wipe out humans from existence. Since it would be able to self-improve, it would rapidly become smart enough to gain complete control of matter, by solving (if still unsolved by then) the nanotechnology problem.
At the same time it would have access to almost all human knowledge with the world wide web. But if it decided to use the nanotech insidiously enough, then it could decide to monitor "offline" info as well by spying on humans with nanomachines (we wouldnot be able to tell, too).
So, extremely quickly, even while the AI was stll contained in a pc in that lab, it would rapidly gain control of all matter and all information.
It would then aim to satisfy the goals it has. Lets say for example that its goal was set by its original programer to be "gain as much knowledge as possible". Who is to say that the AI would not cause a third world war in its attempt to gain more knowledge of human behaviour? Not that it would have any incentive to be bad, as hopefully its creator wouldnt have given it such a goal. But whatever its given goals were (even worse if it can change its own goals), it would in the process of reaching them, make us extinct.
So, we are trying to make AI s really "work" and reach human intelligence or smarter, but the moment someone gets it, even by accident, without understanding how, then at that moment it will probably be gameover for us if there hasn't been prediction to make it inherently friendly.
So the objective of the Singularity institute for AI (SIAI) is to steer research into making AIs that inherently have it in their being to uphold the moral codes etc. that will ensure our wellbeing will not be threatened. They just call it "friendly AI".
"How do you go about doing that, isn't it impossible to code that in, if it can at will replace its own code?"
Well maybe the trick lies in this: Say you have a pill that makes you want to kill people. You would not take that pill if:
1. you didnt want to kill people
and
2. you knew that it was going to make you kill people.
those are the things that we must make sure the AI will know. So we will make it so that
1) before it augments itself, it has to know exactly how that would cause it to behave
2) and it begins knowing that it wants to be a friendly AI.
this means that it will never overwrite its own ethics as part of a self-improvement optimisation, and only those optimisations that preserve its moral code will be carried out.
THAT is what friendly AI should be all about, that is what I am also hoping for.
AND, even throughout all this journey into the infinite amount of changes that a singularity will entail, it is obvious that there would be no need for the AI to be self-aware, just intelligent.
So, to conclude, I believe that if we end up making our AIs self-aware, that would be a side effect of their operation, and would also be practically unimportant.
The only implications of self-aware or nor superintelligece are moral and philoshophical.
If we make a non-self-aware superintelligence, and it recuresively self-improves to insane levels of cleverness and intelligence, and it still is not self-aware, it would mean that maybe, just maybe, self-awareness really IS something that does not stem from our own phisical nature, and it is evidence of a soul or somesuch, and that for that reason no AI could ever become self-aware. That would mean philosophers could continue to happily go about their business as they do today.
If the AI at some point, whether it is upon its birth or later during its rapid self-development, does become self-aware, then this will be a serious philosophical and spiritual implication to consider, and it might have very profound concequances to our religions.
-BUT-
since we can never know if someone or something is self-aware or conscious, that means that:
-we should probably treat such AI with all the human rights etc. that we treat humans, just in case
-we would still be able to hold our faith, our religious views, and much of our philosophy, since it cant be proved that we should consider it self-aware
and one last thought:
-maybe, just maybe, God inserted this inherently unsolvable problem ("we cant know if something or someone is self-aware/consious"), PRECISELY to allow us to build self-aware intelligence, without forcing us to loose both faith in our religion, and faith that "we humans are special".
Because, if it was self-aware and we knew it, it would solve all of our phisical problems, like disease aging etc, it would make us able to live forever if desired, but it would be a pointless life if we had to know that "you are nothing more than an intelligent machine, just like the AI you created. but feel free to live forever if you wish sinceyour ai allowed you to do so". that might mean that many of us would suicide considering life hopeless, others would suicide after making offspring just to give them the chance to choose if they want to live like this, while yet others could end up pondering forever whether they should die or not.
This probably wouldnt be a very good final destination for humanity. But cheer up everyone! God has prevented this from ever happening, because he has made the nature of all intelligence and all understanding such, that we will never be able to TELL if the AI will be conscious or not. Not being able, in the meantime, to tell if other humans are consious or not, is a small price to pay in this case, since we do get left with knowledge of our own personall self-awareness, end that is enough for us to enjoy all our philosophical questions etc. etc.
| | | Last edited by Kinnison @ 9/4/2008 2:00:00 AM | | |
|
|
|
|
|
