| 26 FEB 2003 at 7:15pm |
StammerGuild Master
Posts : 3894
Joined: 5 JAN 2003
Status : Online |
I'll read it and i'll tell ya my opinion
Resistance is not futile, we're gonna win this thing, humankind is too good, we're not a bunch of under-achievers! We're gonna stand up, and we're gonna be human beings. We're going to get fired up about the real things, the things that matter! Creativity, and the dynamic human spirit that refuses to submit.
|
| 26 FEB 2003 at 8:44pm |
ValGuild Master
Posts : 3472
Joined: 2 OCT 2002
Location: US
Status : Offline | One of my best friend's brother is supposedly a premier mind on AI research in London. This should prove an interesting read.
We can be heroes, just for one day.
|
| 26 FEB 2003 at 8:56pm |
| Deleted User | Hmm... I didn't read the whole artice. Interesting topic though. I think we'll see real AI this century, but it's going to be a collaborative effort by scientists in many fields. First of all, it would be nice if someone could define what intelligence is.
Personally, I think randomness plays a key role in it. If everything's predictable, computable in mathematical terms, it cannot really be intelligence, can it? But let's say a random generator creates unpredictable results and there is a mechanism to detect the outcome, and when the result is good, it is incorporated in the system. That way, the machine could create something not directly defined in its design, and thus produce something that is better than itself. A kind of virtual evolution process. That, I believe, would lead to some form of artificial intelligence...
|
| 26 FEB 2003 at 9:29pm |
MichalNGrand Inquisitor
Posts : 7058
Joined: 14 SEP 2003
Status : Online | Originally Posted By Petter_Holmberg (26 FEB 2003 8:56pm)
Hmm... I didn't read the whole artice. Interesting topic though. I think we'll see real AI this century, but it's going to be a collaborative effort by scientists in many fields.
That's what they were saying in the last century
First of all, it would be nice if someone could define what intelligence is.
From the article:
"My reaction to intelligence is the same as my reaction to pornography," he wrote. "I can't define it but I like it when I see it."
That sounds pretty accurate. Intelligence is sexy
Personally, I think randomness plays a key role in it. If everything's predictable, computable in mathematical terms, it cannot really be intelligence, can it?
That's the big question. If randomness is an essential ingredient then machine intelligence must be fallible. So when a pilot AI makes a mistake and causes a plane crash... who do you sue?
I forgot my sig.
|
| 26 FEB 2003 at 11:48pm |
| Deleted User | Originally Posted By MichalN (26 FEB 2003 9:29pm)
That's what they were saying in the last century Yes, but I'm absolutely sure it's still true! It's just that people didn't truly realize just how complicated it was. Maybe they still don't know half of it...
That's the big question. If randomness is an essential ingredient then machine intelligence must be fallible. So when a pilot AI makes a mistake and causes a plane crash... who do you sue? No, that's not true. A traditional programming language, which you would normally use to make an "intelligent" program behaves in a completely deterministic way. But you can create undeterministic things even with machines. Or more generally, you can create undeterministic things with mathematics, or at least something that behaves like it was undeterministic for all we know. In real life, we really cannot know for sure that such a thing as pure chance exists, because by observing something we always disturb it in some way, so we cannot tell if it would have behaved differently if we hadn't observe it. I'm talking about Schroedinger's cat and all that...
|
| 27 FEB 2003 at 12:04am |
MichalNGrand Inquisitor
Posts : 7058
Joined: 14 SEP 2003
Status : Online | Originally Posted By Petter_Holmberg (26 FEB 2003 11:48pm)
Yes, but I'm absolutely sure it's still true! It's just that people didn't truly realize just how complicated it was. Maybe they still don't know half of it...
Of course the question is, if they were wrong 40 years ago, why should they be right now?
No, that's not true.
Why not?
A traditional programming language, which you would normally use to make an "intelligent" program behaves in a completely deterministic way.
Given stable hardware, and well written software, yes. And since those two are not so usual, not quite deterministic in reality.
But you can create undeterministic things even with machines. Or more generally, you can create undeterministic things with mathematics, or at least something that behaves like it was undeterministic for all we know.
I know all that But that wasn't the point. The point was, if there is some randomness, how do you make sure the AI doesn't make the same mistakes as humans? Not that non-deterministic process would necessarily have to be always correct...
I forgot my sig.
|
| 27 FEB 2003 at 3:32am |
| Deleted User | The use of fractals should help to that. I'm not sure how though.
|
| 27 FEB 2003 at 12:12pm |
| Deleted User | Originally Posted By MichalN (27 FEB 2003 12:04am)
Of course the question is, if they were wrong 40 years ago, why should they be right now? They weren't wrong in thinking that it was a field of science that tapped into many areas outside of computer science. They were wrong to think that AI was simpler to implement than it's turned out to be.
Why not?
Given stable hardware, and well written software, yes. And since those two are not so usual, not quite deterministic in reality. They are still deterministic, even if they don't behave the way they were intended to by the programmer. Bugs can create weird results, but they just mean that the programmer made an error, which resulted in a program that does something else. That's not undeterministic behaviour, it's just unexpected behaviour.
I know all that But that wasn't the point. The point was, if there is some randomness, how do you make sure the AI doesn't make the same mistakes as humans? Not that non-deterministic process would necessarily have to be always correct... Ah, very good question! My answer is that I think it should not behave correctly. The whole idea would be for it to behave irrationally all the time, but whenever a freak occurence actually improves perfomance, it would adapt to implement this think into its design. It's just like evolution. A reproductive molecule collects raw material from its surroundings and makes copies of itself. High-energy radiation sometimes changes the layout of the molecule in an undeterministic fasion, which in 99,9% of times results in damages that makes the molecule useless. But in some few cases, the result is a molecule that actually behaves more efficiently than the original one, so it slowly takes over, and so on... A computer program behaving in a similar way could speed up this process greatly. Sure, we humans could sometime give it a nudge in the right direction, but if it really should become intelligent, it would have to do it mostly by its own. It's like the classical religious dilemma: If God has absolute power and uses it to its full extent, he would have designed everything going on around us. We wouldn't even have a free will, because we'd just be his puppets, behaving according to his intentions. This is not a very encouraging thought, so most religious people (that believes in one god) have chosen to believe that God has created us, but then left us the responsibility to create our own future. In the case of AI, we'd be the evermighty gods over our creation, so we'd need to give the AI the power to evolve by itself as well.
Originally Posted By Randdom (27 FEB 2003 3:32am)
The use of fractals should help to that. I'm not sure how though.
That has to be the comment of the day! Right on, Randdom, you might have a point there, but I don't know exactly how either... Fractals are really cool!
|
| 27 FEB 2003 at 4:36pm |
MichalNGrand Inquisitor
Posts : 7058
Joined: 14 SEP 2003
Status : Online | Originally Posted By Petter_Holmberg (27 FEB 2003 12:12pm)
They weren't wrong in thinking that it was a field of science that tapped into many areas outside of computer science. They were wrong to think that AI was simpler to implement than it's turned out to be.
What if they were wrong in thinking that an AI can be implemented in the first place?
They are still deterministic, even if they don't behave the way they were intended to by the programmer. Bugs can create weird results, but they just mean that the programmer made an error, which resulted in a program that does something else. That's not undeterministic behaviour, it's just unexpected behaviour.
Nope. The behaviour is often nondeterministic as in unpredictable and nonrepeatable. This may be caused by interatcion with external events (such as user input) which are by their nature nondeterministic. Then you have hardware glitches caused by RFI, poor cooling and whatnot. These are completely nondeterministic. I'm talking about real life PCs here, not some theoretical "ideal" computers
Ah, very good question! My answer is that I think it should not behave correctly. The whole idea would be for it to behave irrationally all the time, but whenever a freak occurence actually improves perfomance, it would adapt to implement this think into its design. It's just like evolution.
This could work... or it might not
A reproductive molecule collects raw material from its surroundings and makes copies of itself.
You mean cell? Molecules can't reproduce themselves.
In the case of AI, we'd be the evermighty gods over our creation, so we'd need to give the AI the power to evolve by itself as well.
Of course there's the little catch - what if the AIs decide that they'd be much better off without humans?
That has to be the comment of the day! Right on, Randdom, you might have a point there, but I don't know exactly how either...
I think Randdom was under influence when he said that
I forgot my sig.
|
| 27 FEB 2003 at 5:32pm |
| Deleted User | Originally Posted By MichalN (27 FEB 2003 4:36pm)
What if they were wrong in thinking that an AI can be implemented in the first place? Possible, but not probable, if you ask me. It's probably extremely hard, but I think it's possible because I think our intelligence in its most basic level is defined by our physical properties, and that we can develop the technology to emulate or even recreate such physical properties.
Nope. The behaviour is often nondeterministic as in unpredictable and nonrepeatable. This may be caused by interatcion with external events (such as user input) which are by their nature nondeterministic. Then you have hardware glitches caused by RFI, poor cooling and whatnot. These are completely nondeterministic. I'm talking about real life PCs here, not some theoretical "ideal" computers All right, I cannot argue with that. But random user data or hardware glitches doesn't have anything to do with A.I. A.I. A program is still a theoretical entity, entirely defined by the programmer and complying to the laws of the programming language. An intelligent machine would somehow have to program itself, based on decisions not made by a human.
This could work... or it might not Right! I'm just brainstorming. I don't know much about A.I. actually.
You mean cell? Molecules can't reproduce themselves. Yes, there are molecules that can reproduce themselves! This happens all the time in our body, as RNA molecules copy snippets of our DNA molecules. There are even more simple forms of molecules that can generate copies of themselves, and this is how most scientists believe life started in the first place. Some molecule was randomly formed in the "primordial soup" in the ocean somewhere, where water, volcanic chemicals and radiation interacted to "stir around in the pot" so to speak, and it had the capability of copying itself, and started doing that automatically. Sure, it's just a theory, but modern chemistry and physics suggests that it is possible, maybe even probable. I don't know the latest developments in the field...
Of course there's the little catch - what if the AIs decide that they'd be much better off without humans? Hehe, the classic sci-fi story! From 2001 to Terminator, this idea has been around for decades. I don't think it would have to come to that if we thought things over.
|
| 27 FEB 2003 at 5:42pm |
MichalNGrand Inquisitor
Posts : 7058
Joined: 14 SEP 2003
Status : Online | Originally Posted By Petter_Holmberg (27 FEB 2003 5:31pm)
Possible, but not probable, if you ask me. It's probably extremely hard, but I think it's possible because I think our intelligence in its most basic level is defined by our physical properties, and that we can develop the technology to emulate or even recreate such physical properties.
Maybe, maybe not. I'm not convinced either way.
All right, I cannot argue with that. But random user data or hardware glitches doesn't have anything to do with A.I.
Maybe not. Except that random glitches in an AI might have very interesting effects.
Hehe, the classic sci-fi story! From 2001 to Terminator, this idea has been around for decades. I don't think it would have to come to that if we thought things over.
When was the last time humans thought things over?
What if it turns out that for AIs (if they're possible at all) to be useful, they need to have some level of free will? What if it turns out that for an AI to work, it needs to work a lot like the human mind, with all the associated problems?
I forgot my sig.
|
| 27 FEB 2003 at 7:36pm |
| Deleted User | Originally Posted By MichalN (27 FEB 2003 5:42pm)
Maybe not. Except that random glitches in an AI might have very interesting effects. Yes, but most likely it would not be what was driving the AI forwards.
When was the last time humans thought things over? Good point... :
What if it turns out that for AIs (if they're possible at all) to be useful, they need to have some level of free will? What if it turns out that for an AI to work, it needs to work a lot like the human mind, with all the associated problems? Yeah, I think those are important questions too. Maybe that's the case. One day, we may be arguing about AI rights, just like human rights today.
|
| 27 FEB 2003 at 11:39pm |
MichalNGrand Inquisitor
Posts : 7058
Joined: 14 SEP 2003
Status : Online | The second part of the article is up now:
http://www.salon.com/tech/feature/2003/02/27/loebner_part_2/index.html
Here's a fun quote:
In other words, if you read between the lines what you come up with is that one reason "serious" A.I. scientists don't try to mimic human speech anymore is that they discovered they can't do it. Of course, they promised 30 years ago that they would be able to do so "real soon now," but it has turned out to be harder than expected, so now it's dismissed as mere mimicry.
I forgot my sig.
|
| 28 FEB 2003 at 2:47am |
| Deleted User | Originally Posted By MichalN (27 FEB 2003 4:36pm)
I think Randdom was under influence when he said that
ouch!
What are you talking about anyway ??? Do YOU take drugs ?! -big finger pointing at you.
( I deleted this frightful post )
Sorry to interrupt; very interesting and constructive discussion.
|
| 28 FEB 2003 at 2:53am |
MichalNGrand Inquisitor
Posts : 7058
Joined: 14 SEP 2003
Status : Online | Originally Posted By Randdom (28 FEB 2003 2:47am)
ouch!
Well, were you?
Do YOU take drugs ?! -big finger pointing at you.
Very, very, very rarely. And how did you know I wasn't talking about alcohol?
I forgot my sig.
|
| 28 FEB 2003 at 2:59am |
| Deleted User | My answer is that I think it should not behave correctly. The whole idea would be for it to behave irrationally all the time, but whenever a freak occurence actually improves perfomance, it would adapt to implement this think into its design.
Yes, I think that's how every "intelligence form" shapes itself; trials and errors, good and bad conditioning. I see no reason why it couldn't work with for AI.
|
| 28 FEB 2003 at 3:02am |
| Deleted User | Originally Posted By MichalN (28 FEB 2003 2:53am)
Very, very, very rarely. And how did you know I wasn't talking about alcohol?
I'm just very paranoiac must be alcohol (...) side effect
|
| 28 FEB 2003 at 8:17pm |
MichalNGrand Inquisitor
Posts : 7058
Joined: 14 SEP 2003
Status : Online | Originally Posted By Godfather (26 FEB 2003 7:15pm)
I'll read it and i'll tell ya my opinion
So? What is your opinion? Just curious...
I forgot my sig.
|
