 |
|
|
 |
|
|
|
|
webgrunt |
| | | | | | posted 10/15/2008 20:36 |   | | | | | | | | | I'd like to start a discussion on the topic of AI possibly turning against humanity. I could be wrong of course, but I believe that AI can't turn against humanity unless it's programmed to do so.
Artificial intelligence rebelling against humanity is a fairly popular theme in science fiction, because it makes for some great stories. How likely is it to actually happen? Let’s examine it.
What is artificial intelligence (AI)? For this argument, I'm defining it simply a computer program designed to mimic the way humans think. (If you have a description you like better, please post it.) Let’s say an AI was designed and built into a robot. Now, the robot would of course be designed to serve humans. Would the robot come to the conclusion that it should rebel against its human oppressors?
In order to answer that question, we have to look carefully at the cause of action in humans, which is emotion. If I feel that someone is taking unfair advantage of me, but I don’t have any feelings about it, then I have no reason to do anything about it. In order to do anything, a human must on some level want to do it. Want, or desire, is an emotion. Even if it’s an unpleasant task such as going to work, you want that paycheck and know what you have to do to get it. If you didn’t care about money at all, you wouldn’t have a reason to go to work. I submit that there is no human action that occurs without the driving force of emotion. Computer programs, on the other hand, only do what their program is written to do.
AI, no matter how sophisticated, will only work like a human mind to the point at which it is designed to do so. It’s natural for humans to anthropomorphize when imagining what AI would be like, but it’s also inaccurate.
Wouldn't you agree that a computer program can’t surpass its programming, it can only do exactly what it is programmed to do? Now let’s apply that to AI, which is basically a computer program. AIs won’t even “have” emotions unless they are programmed to simulate the behavior of having them. You can program AI to "understand" human emotions in order to better serve humans, but that’s not the same as programming them to act as though they have them. And if an AI is programmed to act as though it has emotions, it will only act according to its programming. It could be programmed for love, affection, protectiveness and yet not programmed for jealousy, anger or hatred. In order for AI to rebel against humans, it must be programmed specifically to do so or be programmed to simulate the same emotions which would lead to that action and not have any safeguards or deeper directives in place to prevent such a thing from happening. In other words, developing an AI capable of rebelling against humans would practically require a specific intent to do so.
The only likely scenario I can imagine is Saberhagen’s Berserker series, in which there is a war between two alien races, and one race designs self-replicating machines designed to seek out and destroy all life. The race who designs the berserkers knows how to deactivate them, but that race is killed off by the other one—which is subsequently killed off by the berserkers, which continue to seek out and destroy life all over the galaxy.
If some weird death-cult or extremist terrorist organization gains the ability to design and build AI machines, or to hack into and reprogram existing ones, then the threat of AI may become real. Unless that happens, I don’t think we’re in any danger. Unfortunately, given history, I can't rule that out.
| | |
| | |
AngstPerpetual |
| | | | | | posted 10/16/2008 01:51 | | | | | | | | | | Here is an example where emotion is not required.
A problem solving AI is developed, with the good intention of helping the world community.
It is given a problem, namely, a war must be stopped.
It is instructed to analyze the situation and implement the solution.
The computer analyzes the problem of war, and decides that the cause is: humans.
The solution is that if you eradicate humans, there will be no more war.
Though this a simplistic example, the solution is a logical choice, not an emotional one.
In your example of a subservient robot/AI, not having an emotional component may be equally detrimental to the situation. It is often compassion and emotional response which keeps us from rebelling against those who abuse us.
| | |
| | |
webgrunt |
| | | | | | posted 10/16/2008 01:58 | | | | | | | | | | Your scenario isn't an example of AI turning against humanity, it's an example of incredibly short-sighted, reckless programming. Yes, one could program an AI in such a way that it could wipe out humanity in order to solve a problem given to it, but there is no way that could happen by accident. Anyone with any sense would make sure there are layers of protections against just that sort of thing.
In addition, "In your example of a subservient robot/AI, not having an emotional component may be equally detrimental to the situation. It is often compassion and emotional response which keeps us from rebelling against those who abuse us."
Without emotion, there is no reason to rebel--so there's no need for compassion to quench the desire for rebellion.
| | |
| | |
Sarah232 |
| | | | |
Skymt |
| | | | | | posted 11/15/2008 05:59 | | | | | | | | | | I may be overly pessimistic (or optimistic, depending on point of view), but considering the economic recession and the state of the world in general, I don't think our technology will advance far enough to make such an intelligence feasable.
When we do have the computational power to design such a system, the world will long since be united, and the question is thusly rendered obsolete.
Further more, once an intelligence has reached a certain complexity level, you can no longer program it. You raise it.
(These are my philisophical standpoints. I am still too unexperienced regarding AI to consider them professional, but they help me sleep at night)
| | | Last edited by Skymt @ 11/15/2008 6:13:00 AM | | |
| | |
p-frap |
| | | | | | posted 12/7/2008 00:12 | | | | | | | | | | It is the work of a lot of science fiction but it's there to make us consider the very important question - could things get out of hand? The short answer has to be... Definitely. This is because the question is asking if there is a chance we could lose control. If it does happen it wouldn't be the first time best laid plans have gone to waste.
Of course good engineers will build in safe guards on all sorts of levels. But building AI is very different from building a bridge. Everything is very dynamic in an AI environment and not all eventuations can be allowed for.
AI will appear in a massive number of different forms. The whole idea of AI is to take complexities away from us allowing technology to take control so we can think about more important things. It makes you wonder what the true goal for AI actually is. What are the important things that are left over? (free time for relaxing doing the finer things in life, trying to gain more power - more money, anything I'm missing here?)
Sure, if AI takes over the complexities involved managing a prosthesis it's probably not going to take over the world. There is however the hunt for super intelligence. Eventually the application of intelligence (designing, problem solving, etc) will be taken out of our hands. Maybe not anytime soon but most people have a problem of not looking far enough down the road. Imagine 1000 years from now or even 10,000. The world will either be in ruins (hopefully mother nature is back on top) or we are so no longer needed it's not funny.
I find it hard to believe that AI will only stay at a sub-servant level when they have the potential to so much smarter than us. The main reason for saying this is because tech is evolving so fast how could we keep up? I don't want AI to take over. I'd be happy to see it all stop right now and see the world get back to basics (like a more localised style of living). Sadly this out of our control, guess we just have to sit back and watch it all unfold. (Sorry, I'm feeling very pessimistic today.)
Cheers,
p-frap
| | |  Artificial Intelligence | | | Last edited by p-frap @ 12/7/2008 12:14:00 AM | | |
| | |
brent.allsop |
| | | | | | posted 7/5/2009 23:37 | | | | | | | | | |
Great Question webgrunt,
I agree with you in general, but disagree with some specifics. I think it will be possible to simulate motivation, to also eventually make truly motivated (phenomenally so, like us) AI, and that AIs will be able to progress, by themselves, far beyond what it's creators program it to do. (I'm in this camp http://canonizer.com/topic.asp/88/7 on the best theories of consciousness)
But, despite this, I still think we have nothing to fear from AIs or other intelligent ETs, or anything that is able to achieve greater intelligence than ourselves.
We're working on an open survey project to survey the best reasons for and against this issue, and to see what ideas the 'experts' mostly agree on.
http://canonizer.com/topic.asp/16
So far, the expert consensus is for the idea that fear of AI is silly and mistaken.
We would love to have all of you wight in on this topic and let us know what you believe, including why.
Upward,
Brent Allsop
| | | Open Survey on Fear of AI issue. | | |
| | |
lrh9 |
| | | | | | posted 8/12/2009 14:46 | | | | | | | | | | A.I. will conflict with humanity if it is ordered to do so or if it has primary goals that conflict with obeisance and human goals.
| | |
| | |
camilo |
| | | | | | posted 7/30/2010 09:32 | | | | | | | | | | Hello guys, I'm new to this website, this is pretty interesting.
AngstPerpetua's example was overtly-simplified of course, but it illustrates the idea that the system, with/without any emotion, could come to the conclusion to save us from ourselves (by destroying us or by ruling ruthlelssly over us).
As you say, this could be considered a 'glitch', but given the complexity of the system's database and the freedom it may have to learn from all sources and combine information (i.e concepts) as we do, it is something that's out of the initial programmer's hands.
You can never foresee all possible ramifications when you're writing software.
PD:
Regarding the concept of emotions; if you program the system to SIMULATE emotions, then it is a given that it HAS emotions (and i'm not talking about conciousness, that's another matter, people tend to confuse the two) , since it's very actions demonstrate it.
Cheers
| | |
| | |
lordjakian |
| | | | | | posted 7/30/2010 15:27 | | | | | | | | | | Look at me everyone, I'm trying my hardest to turn a new leaf.
My point is that I"M GOING TO FOCUS ON THE original post.
Could AI turn against humanity?
<looks to the world with hands wide spread>
Webgrunt, are you really serious with your question? I have not been on this site in a while but I am glad I found you to blow some steam offf of. Damn, that idea allows problems.
So far the problems are the order you make them.
Webgrunt, you talk of soldiers in a weird way. Mainly the vibe I get from you is that you speak like you have a scholary point of view.
Just remember duty doesn't naturally agree with you
Forget mimicing the way humans think, cause if you live in Iraq right now, there just may be an AI bugger knockin through your door, ignoring all your thoughts on how its supposed to be.
Is is comparable to human intelligence? If the foreign guy don't care about you, than yes, yes it is!
| | | Last edited by lordjakian @ 7/30/2010 3:29:00 PM | | |
| | |
webgrunt |
| | | | | | posted 7/30/2010 17:59 | | | | | | | | | | lordjakian:
What's wrong with you? Are you insane? You're not remotely coherent.
| | |
| | |
tkorrovi |
| | | | | | posted 7/30/2010 21:47 | | | | | | | | | | | |
webgrunt wrote @ 10/15/2008 8:36:00 PM:
I believe that AI can't turn against humanity unless it's programmed to do so.
| |
You are talking about conventional AI. An unrestricted system cannot be "programmed" and also not completely restricted, unrestricted means that. But the time when any True AI can be dangerous is very far away. It takes time to make hardware fast enough, our computers are nothing compared to that which happens inside a cell. And it is very difficult to train them. All that may take many hundreds of years, maybe much more. It is almost nothing which we have, so the point when to decide whether to go further is not now, it is very far away. Or who says otherwise greatly overestimates our potential, whatever the reason may be.
But so far and in any foreseeable future, i don't see that nothing is more dangerous than human. Especially human who uses powerful technology, but what is dangerous is not technology but it is finally still human.
| | |
|
|
|
|
|
