Will machines have feelings in the future ?

[El Greco]

1) By "machines" I mean things built from scratch, not cloned or otherwise tampered with humans.

2) Provided of course that humanity doesn't get destructed before.

3) By "feelings" I mean something exactly like what we mean today by this word.

The rationale is that eventually we will be able to fully analyze the brain and we will also have the technology to build an analogous artificial neuronal network. So we will be able to make a computer that has human-like feelings.

[Eos of the Eons]

Why would we want them to have feelings? I would prefer logic, cold hard logic being the machine that transports us or whatever. I wouldn't want to have it become angry and irrational and taking revenge! Or what about jealousy. Ugh. I've seen some sci fi shows that explore this, and it's always a nightmare.

[El Greco]

Originally posted by Eos of the Eons
Why would we want them to have feelings?

Well, many reasons... An important one is to be able to play realistic multiplayer games without other humans :D

But anyway, if there is such possibility in the future, it's almost certain that humans will do it without asking too many "whys". Humans have been, are, and will always be trying out anything possible.

[Anders]

Sure it will, giving the advancement in HW and especially software. Also, as the brain is in no way immaterial, it’s just a blob of organic tissue that. Also, feelings are created by the brain, so no problem there. I guess that someone is going to try to refute this…

[Paul C. Anagnostopoulos]

Eos said:
Why would we want them to have feelings?
Because we want to see if it's possible.

~~ Paul

[CFLarsen]

Originally posted by Paul C. Anagnostopoulos
Because we want to see if it's possible.

~~ Paul

Then go for recreating Naomi Campbell's butt. Much, much more marketable...

[BillyJoe]

The choices are a bit restricting.
A materialist would say that it is certainly possible.
Whether it will actually ever happen is another question

[El Greco]

Originally posted by CFLarsen
Then go for recreating Naomi Campbell's butt. Much, much more marketable...

That's the idea, but I also want to hear how much she loves me.

[Dymanic]

I find myself unable to answer for lack of a crisp enough definition for 'feelings' (or, for that matter, 'machines'). There are some interesting ethical implications which are often overlooked (these were investigated in some short stories by Stanislaw Lem, back in the early seventies).

[steenkh]

Originally posted by Eos of the Eons
Why would we want them to have feelings?
The Christians believe that God has created us in order to have somebody (after a proper selection procedure on Earth) spending all eternity singing his praise. We could build machines that would love us, and by a similar darwinistic selection process, we could weed out those machines that did not do it properly!

Now, what are feelings? Would we ever really accept it in anyone else than ourselves? I take it that what is really meant is "emotions", but if we look at how humans evaluate animal feelings, we have the same problem.

Do fish feel pain? They certainly act as if they do, but many people, especially fishermen, would claim that fish do not feel pain.

Every owner of a pet dog knows that it can display emotions, but nevertheless I have heard a scientist doing experiments with animals claim that these are just chemical reactions in the brain that have evolved to handle certain situations, but because animals do not have our capacity for thought, we should not think that these are emotions like we have them.

On the other hand, people are fooled by crude computer programs to believe that the programs can think and have emotions.

The question is, what is the definitions of emotions, and can we recognise emotions when we see them?

If we somehow manage to build a really intelligent machine, are we then ethically justified in switching it off at will? And what if the machine is unable to show emotions? Are we then more justified in switching it off than if it can show emotions, or something that looks like it?

One of my Windows machines sings "Daisy, daisy" every time I close Windows!

(Edited to say "ethically" instead of "morally")

[Hellbound]

Originally posted by Eos of the Eons
Why would we want them to have feelings? I would prefer logic, cold hard logic being the machine that transports us or whatever. I wouldn't want to have it become angry and irrational and taking revenge! Or what about jealousy. Ugh. I've seen some sci fi shows that explore this, and it's always a nightmare.

I'd think we'd want certain feelings, at least in the form of program restrictions if nothing else...traits such as compassion, empathy, etc.

Because really, cold hard logic alone, killing off the weak, frail, diseased, unintelligent, or otherwise genetically "inferior" members of our species could be considered "logical", if heartless and cruel. Just an example I can think of :) Assuming we create an actual artificial intelligence, I'm not sure we could achieve a level of autonomy and proactiveness comparable to humans without some form of emotion developing. I've always thought our emotions are the results of combinations of past experiences, pattern searching, fufillment of basic needs (survival instinct, reproductive drive, hunger, etc, analogs of which could be used in artificial life as well), and other factors that arrive out of our ability to think, imagine, and construct alternative scenarios and ideas in our minds.

Of course, I'm not an expert in psychology, robotics, programming, or anything else related to this discussion, so take my comments for what they are ;)

[El Greco]

Originally posted by steenkh
If we somehow manage to build a really intelligent machine, are we then ethically justified in switching it off at will? And what if the machine is unable to show emotions? Are we then more justified in switching it off than if it can show emotions, or something that looks like it?

These questions were about to follow... and if we assume that the "feeling computer" is on a robot, then the questions become more exciting: Will the robot let us shut it off ? Will it try to shut us off ?

I'm so jealous of the people of the future who will face these questions on a less theoretical base...

[jay gw]

I'll tell you how to give computers feelings.

Just write a program that has them do illogical and random things for no reason.

[Eos of the Eons]

Originally posted by Huntsman

Because really, cold hard logic alone, killing off the weak, frail, diseased, unintelligent, or otherwise genetically "inferior" members of our species could be considered "logical", if heartless and cruel. Just an example I can think of :)

If they have no feelings then they have no reason to kill. Radically, they don't care. Or just make sure their programming does not include killing anyone. If it isn't programmed to kill, then how can it except by some whacky accident. We could have all kinds of safety protocols in place.

Also, don't have robots "judging" humans in any way. Then it won't compute some crazy "inferiority".

Just not to bother, it's rather a waste of time. I personally see no benefit to it.

That's just silly ol logical me though. I'm not curious as to whether or not we could get a machine to have emotions/feelings.
Unless you can tell me a really practical application that won't be more complex than it's worth.

[Dymanic]

Originally posted by steenkh

If we somehow manage to build a really intelligent machine, are we then ethically justified in switching it off at will? The first question really is: are we ethically justified in switching it on?


Originally posted by jay gw

I'll tell you how to give computers feelings.

Just write a program that has them do illogical and random things for no reason.I could almost buy that if you included: because you wrote it to do that among the non-reasons it was what it was doing. Still, what does that have to do with feelings?

[Hellbound]

Eos:

Well, my whole point is that I don't think you could have intelligence without some capacity for the machine to do things you don't plan on.

Even instilling rules against certain things might be no guarantee. I just think that emotions are extensively tied up in our ability to adapt and react...i.e.-the characteristics of our intelligence.

And, just to throw out a point, the killings would be entirely logical, and could well fufill a goal of heloing out the human race. An extreme example, yes, but I could tone it down to realistic proportions. Think of an AI-driven spacecraft (or any passenger craft, for that matter). Any sort of risk analysis is going to become difficult. At what point does the loss of human life outweigh the lose of materials or monetary losses? If a spacecraft is damaged or malfunctions, and the only way it can save enough power to get back to earth is to shut down all extra systems, including life support, what choice would it make? If we restrict it's actions, there are consequences either way, and I don't know if we could really claim it was intelligent. Placing the human factor first would mean that vitally important data could be lost. Placing other factors above might mean that the crew would be killed only a short time before rescue could have arrived. As humans we face these situations every day, and we have to use both logic and emotion to reach answers to situations like these. I don't believe you could break it down into a ruleset..and if you did, that ruleset would not be "intelligence". You would defeat the very purpose you desinged AI for...it's ability to make it's own decisions about circumstances.

I just don't think it's possible to have an intelligent machine that operates on pure logic. Computer programs today operate purely on logic, and the general idea is that heuristic systems are not able to emulate intelligence. You have to have areas of uncertainty, "fuzzy logic" and self-learning neural nets, and these areas could complicate any hard and fast rules, as well as meaning any purely logical position could be deviated from. If the machine cannot deviate from pure logic, is it intelligent? I think that's the crux of the question.

More a philosophical debate, though. I could very well be completely off my rocker :)

Edited to add: The whole idea of AI, what makes it advantageous, id that it could adapt to situations outside it's programming. Thus, what you program it for is not the limit of what it could do, if it is truely intelligent.

[Dymanic]

Originally posted by Huntsman

I don't believe you could break it down into a ruleset..and if you did, that ruleset would not be "intelligence". You would defeat the very purpose you desinged AI for...it's ability to make it's own decisions about circumstances.
It's a bit of a puzzle, isn't it? If you tell a program to do something you didn't tell it to do, and it does that, did it do what you told it to -- or not?

[Dylab]

If feelings are not sesntial for robots to do everything we want them to do then why do we have feelings? I know natural selection doesn't produce the same designs we would create for the same purpose but I can't see any reason why evolution would favor machines with feelings than machines without them unless feelings had a survival benefit.

[Eos of the Eons]

the killings would be entirely logical, and could well fufill a goal of heloing out the human race.

Why would we program it this way in the first place? That is a very general, hugely open "program" or goal. Keep things rather specific.

What kind of tasks to do people have in mind for an "emotional robot"? To be somebody's pal? What kind of tasks require emotion?

Can't something make decisions on circumstances without bothering with emotions? Examine the consequences of each action that could be potentially done, and find the solution best suited before taking action.

Again, what tasks do people have in mind for determining circumstances?

Something that lets me know that a plant needs watering can just measure moisture levels, they don't need to be concerned about the plant dying. That's my worry if I still don't add water. Or one that adds water as needed, it still doesn't need to be concerned about what too much water or not enough would do to the plant. That's for the human to monitor and worry about. Just program the robot to add x amount of water when the moisture reaches x amount in the dirt.

I wouldn't want any robot in charge of anything that requires too much, like babysitting. So what tasks would a robot do that would require feelings or emotions?

[69dodge]

Originally posted by El Greco
The rationale is that eventually we will be able to fully analyze the brain and we will also have the technology to build an analogous artificial neuronal network. So we will be able to make a computer that has human-like feelings.How close must the analogy be?

Are feelings a hardware problem or a software problem?

[El Greco]

Originally posted by 69dodge
Are feelings a hardware problem or a software problem?

I guess both. We have a basic structure (the brain of an infant) that grows according to how we program it (environmental effects). Perhaps the "software" also affects the way the "hardware" grows, at least during the first years of life. But let's assume that we will also be able to replicate the growing procedure, namely build a program that replicates the functions of an infant brain and so it has limited abilities of self-modification and code expansion.

[Brian]

I can't tell. If a computer was as complex as a human brain, I Think!, it would be the same thing as a human brain. Will we last long enough to build such a computer?
Suppose we built a computer and asked it to build random computers like itself.... Maybe, that would mimic nature, maybe it would just be a part of nature?

[bjornart]

Originally posted by steenkh
The question is, what is the definitions of emotions, and can we recognise emotions when we see them?


What he said. I assume Humans have emotions because I think I do, and I seem to be a human. I assume "high order" mammals have similar experiences because of the design we share. I assume insects don't, cause I don't want to feel bad about cutting their limbs off.
I have of knowing whether a machine has feelings or is only simulating them, and to my knowledge no-one else have either. Until someone does, the question of machine emotions is unanswerable.

[Hellbound]

Originally posted by Eos of the Eons
Why would we program it this way in the first place? That is a very general, hugely open "program" or goal. Keep things rather specific.

In this cas,e then, there's no need for AI. If we constrain the actions to specific functions, then it can be covered by a hueristic ruleset.

What kind of tasks to do people have in mind for an "emotional robot"? To be somebody's pal? What kind of tasks require emotion?

I'm not saying that emotions would be added in intentionally. I'm saying that I don't believe true intelligence can be achieved without some analog of emotions being present. Simply because the AI can, by definition, learn, change, and adapt.

Can't something make decisions on circumstances without bothering with emotions? Examine the consequences of each action that could be potentially done, and find the solution best suited before taking action.

Again, in this case, why the need for AI? And what is the best solution? What value does a human life, or multiple human lives, play in these calculations? How does it determine what is best? And if it's a flat claculation, again, complex rulesets could cover it without the need for AI. If we expect it to create novel solutions and adapt to unforeseen circumstances, then I don't think we'll be able to prevent emotions of some sort being present.

Again, what tasks do people have in mind for determining circumstances?

Something that lets me know that a plant needs watering can just measure moisture levels, they don't need to be concerned about the plant dying. That's my worry if I still don't add water. Or one that adds water as needed, it still doesn't need to be concerned about what too much water or not enough would do to the plant. That's for the human to monitor and worry about. Just program the robot to add x amount of water when the moisture reaches x amount in the dirt.

Yes, but again, this is not AI. This is a very simple ruleset, and using AI for this is like using a bazooka to hunt cockroaches ;)

I wouldn't want any robot in charge of anything that requires too much, like babysitting. So what tasks would a robot do that would require feelings or emotions?

Again, I don't think, necessarily, that anything would require them, although there may be situations were some type of emotion is helpful. However, I don't think you can seperate the emotion and still have intelligence; I believe the ability to learn, adapt, imagine, and create necessarily includes emotions or analogs of what we call emotions. I don't know if we could avoid creating a true AI that was emotionless.

I think we're getting a bit crossed here. My argument is not necessarily that emotions would be required or intentionally programmed, but that they are intrinsically tied into intelligence. And, there are some situations were emotions of a sort could be useful. Also, I think we're not using the same definition of AI. I'm referring to true intelligence...the ability to adapt to novel circumstances, imagine, learn from past experience, extrapolate to future situations, and form abstract concepts. A ruleset, even a complex ruleset, is not intelligent. A robot created that just waters your plants is not intelligent, nor would it need to be. Hope that clears up my position a bit. :)

[Dymanic]

Originally posted by bjornart

I have [no way?] of knowing whether a machine has feelings or is only simulating them...
How can we know whether we have feelings or are only simulating them?

[phildonnia]

Machines already have feelings; just this morning, an ATM expressed its gratitude to me for having used its bank. :)

But semi-seriously, I think that "feelings" are just a much more complex version of my ATM's gratitude.

Would we want them to? Probably not. We don't want factory robots getting bored or drone pilots having moral qualms about killing.

I think that "feelings" are an emergent phenomenon; I expect they will begin to appear (wanted or not) as machines develop greater complexity in their modeling of the world.

[Peskanov]

Huntsman (and all),

A ruleset, even a complex ruleset, is not intelligent.

You are extrapolating all possible rulesets from one you are familiar with!
Just 2 points:
- A ruleset can be "reflexive". In programming, this means that a program can examine itself and change it's own constitution (similar to rules that create new rules). Logic programming languages like prolog (used in AI) are reflexive.
A reflexive system has no known limits (relative to intelligence), and some of the most advanced IA systems are based in changing rulesets (expert systems). We simply don't know what these systems can achieve or even if they have a limitation.
- When you think about rulesets you boil it to the simplest possible relation:
if X happens then let's do Y.
But take in account that when you use a more complex (a bit more only) model:
if X and Y happen's then do Z
,then the results start to be unpredictable and you have to enter in the realm of statistics to try to see what is happening. With that system you can build hierarchic "castles" of logic which weight thousands of similar factors to obtain just one simple result.
And that's not all. Remenber that a ruleset can "feed" itself. The results can be feedback to the system, changing the meaning of previous factors/causes!

And at last compare all that with the brain: the brick of the brain, the neuron, is a little device which weight it's input signals to produce just one true/false result. And the neural nets have these features:
- Logical rules with multiple inputs (neuron)
- Massive use of feedback (neurons are wired backward and fordward)
- Are reflexive (the neurons are re-wired frequently).

[Johnny Pneumatic]

Originally posted by Eos of the Eons
If they have no feelings then they have no reason to kill.

Bzzzzzzzt wrong! Resources. Do you think crocs are ever happy? They might be, but for the sake of this argument please take it as axiomatic that they aren't. Because they don't have jealousy or happyness isn't going to stop a very hungry salt water croc from eating a careless swimmer.

[Eos of the Eons]

Originally posted by SkepticJ
Bzzzzzzzt wrong! Resources. Do you think crocs are ever happy? They might be, but for the sake of this argument please take it as axiomatic that they aren't. Because they don't have jealousy or happyness isn't going to stop a very hungry salt water croc from eating a careless swimmer.

Robots don't get hungry, and really shouldn't be programmed to view humans as a fuel source. :D

Again, I don't think, necessarily, that anything would require them, although there may be situations were some type of emotion is helpful. However, I don't think you can seperate the emotion and still have intelligence; I believe the ability to learn, adapt, imagine, and create necessarily includes emotions or analogs of what we call emotions. I don't know if we could avoid creating a true AI that was emotionless.

I think we're getting a bit crossed here. My argument is not necessarily that emotions would be required or intentionally programmed, but that they are intrinsically tied into intelligence. And, there are some situations were emotions of a sort could be useful. Also, I think we're not using the same definition of AI. I'm referring to true intelligence...the ability to adapt to novel circumstances, imagine, learn from past experience, extrapolate to future situations, and form abstract concepts. A ruleset, even a complex ruleset, is not intelligent. A robot created that just waters your plants is not intelligent, nor would it need to be. Hope that clears up my position a bit.

Interesting viewpoint, and I think I'm understanding it a bit more. That is quite a claim that one must mull over.

I believe the ability to learn, adapt, imagine, and create necessarily includes emotions or analogs of what we call emotions. I don't know if we could avoid creating a true AI that was emotionless...

they are intrinsically tied into intelligence...

Okay, now you are truly making me think, and I'm intrigued. Do we need robots to be that intelligent? Again, what tasks do you have in mind?

I'm truly wracking my brain for a scenario here. I can't think of one where it would be appropriate.

Thusly, I challenge others to come up with some. This is a higher intelligence we are talking about...what makes us human too, and it scares me that some robot would get some of that. Am I prejudiced?

I mean, I'm glad other animals have emotions, and I don't think we are on some pedestal for possessing what intelligence and emotions we have that make us human, but it scares me to think of a robot with such abilities. I just don't see the benefits, but you state there may be some benefits, some examples would help me wrap my brain around it.

I really feel "ignorant" on this topic since I don't know much about the applications of robots beyond much of what they do now. This actually could be something I could bring up with my husband. He's really into "techonological advances" like this.

Yep, practical application being exploration beyond our solar system. I could see that. I really could. It still scares me though. I guess maybe I'm thinking of those ruthless cylons on Battlestar Galactica. They possess enough emotion to want to get rid of humans, but not enough compassion to live happily alongside humans.

Could we control those robots really well if they are THAT intelligent and possess emotion?

[Johnny Pneumatic]

Originally posted by Eos of the Eons
Robots don't get hungry, and really shouldn't be programmed to view humans as a fuel source. :D

Sure they do, electrical power. Humans use metals, plastics and electrical power that the robots could use so why wouldn't they kill us over the resources? Survival of the most fit; the force behind evolution.

[bjornart]

Originally posted by Dymanic
How can we know whether we have feelings or are only simulating them?

Here's what I wrote first, with the error corrected in the part you quoted... I forgot one word, and deleted another instead of adding the missing one... a couple of crossed wires.

I assume Humans have emotions because I think I do, and I seem to be a human. I assume "high order" mammals have similar experiences because of the design we share. I assume insects don't, cause I don't want to feel bad about cutting their limbs off.

I have no way of knowing whether a machine has feelings or is only simulating them, and to my knowledge no-one else have either. Until someone does, the question of machine emotions is unanswerable.

So my answer is, we can't. But I think I do have them (or at least that's what I tell others;)), and from that I extrapolate that the rest of you, who appear to be put together the same way, also do. But if you were all replaced with Eliza-controlled meat-puppets tomorrow there'd be no way of telling you'd changed.

[Hellbound]

Peskanov:

Although I used the general term rulesets, I was referring to simple rulesets. Even extremely complex static rulesets I would not consider intelligent, because a static ruleset cannot change.

Now, as to your point about rulesets that are reflexive..i.e.-that can change themselves, I agree that they can be intelligent. However, and the basis of my point, is that if the ruleset can change itself then it can ovewrite the safety rules you put in place, or create new rules that go far beyond its original programming or intent.

However, I would argue about the possible "intelligence" of any static ruleset. To me, the major, defining factor in intelligence is the ability to learn, adapt, and react to unforeseen circumstances. An extremely complex static ruleset, quite simply, cannot learn and cannot adapt to the unforeseen as easily. It doesn't develop experience on which it can draw, it would have difficulty creatig relations and correlations between new events and older ones. I don't believe any static ruleset can be consdiered intelligent. Can they perform complex tasks, or even engage in conversations convincingly? Yes, most likely so. But intelligence is directly related to ability to learn, and a ruleset that cannot learn is, by definition, not intelligent.

[uruk]

I voted yes because as software approaches A.I. the simularcum will begin to become indistiguishable from natural.
Anyhoo. The way technology is developing, there will be a time when what is biological and what is technological will not be a clealy defined as it is now.

[Dymanic]

Originally posted by Huntsman

Now, as to your point about rulesets that are reflexive..i.e.-that can change themselves, I agree that they can be intelligent. However, and the basis of my point, is that if the ruleset can change itself then it can ovewrite the safety rules you put in place, or create new rules that go far beyond its original programming or intent.
This way lies madness (http://www.nomic.net/)

"Nomic is a game in which changing the rules is a move. In that respect it differs from almost every other game. The primary activity of Nomic is proposing changes in the rules, debating the wisdom of changing them in that way, voting on the changes, deciding what can and cannot be done afterwards, and doing it. Even this core of the game, of course, can be changed. (Peter Suber, The Paradox of Self-Amendment, Appendix 3, p. 362)"

[Dymanic]

Originally posted by bjornart

I have no way of knowing whether a machine has feelings or is only simulating them, and to my knowledge no-one else have either. Until someone does, the question of machine emotions is unanswerable.I agree. I think there is a fundamental difficulty with this distinction. That's why I didn't vote.

Do human intelligence, creativity, flexibility, and emotions reduce to (very complex, but ultimately rigid) sets of rules, meta-rules, and meta-meta-rules, or is there some additional element? Are our 'feelings' something that is 'simulated' by the application of rules, or are they made 'real' by the addition of some unknown element? Is software really 'approaching AI', or is this like saying that climbing a tree is 'approaching the moon'?

[jay gw]

Do human intelligence, creativity, flexibility, and emotions reduce to (very complex, but ultimately rigid) sets of rules, meta-rules, and meta-meta-rules

Yes. That's why it will be possible to recreate the human brain in a machine.

[Dymanic]

Originally posted by jay gw

---------me-------------
Do human intelligence, creativity, flexibility, and emotions reduce to (very complex, but ultimately rigid) sets of rules, meta-rules, and meta-meta-rules?
------------------------
Yes. That's why it will be possible to recreate the human brain in a machine.Thanks for clearing that up. It seems I have been making much work of a small matter.

[Eos of the Eons]

Originally posted by SkepticJ
Sure they do, electrical power. Humans use metals, plastics and electrical power that the robots could use so why wouldn't they kill us over the resources? Survival of the most fit; the force behind evolution.

Cylons! :D They got rid of humans in order to secure all available resources and make sure nothing could shut them down. Darn things would be the fit survivors.

Evolving AI, scary thought. Would they mimick nature? They aren't "natural".

[Hellbound]

Originally posted by Eos of the Eons
Cylons! :D They got rid of humans in order to secure all available resources and make sure nothing could shut them down. Darn things would be the fit survivors.

Evolving AI, scary thought. Would they mimick nature? They aren't "natural".

You know, I believe this might be one of the best ways to achieve AI..creating a replicating system that inserts mutations into it's progeny, with some sort of selective pressure to remove the less intelligent ones, etc. Airplanes are often designed by a similar process...the computer will simulate a body design, then make semi-random tweaks and changes to the design and test the results. Often, this gives a better aerodynamic shape than a desinger could sit down and come up with.

Of course, again, this sort of a process leads to the possibility that we would create a seperate intelligence, with its own goals, desires, wants, and needs, that might or might not be amiable towards us.

There's a Sci-Fi book called "The Killing Star" which I thought was a very enjoyable book. It includes an appendix that describes many of the technological ideas mentioned in the story, and summarizes the actual science being done in those areas or applications. Anyway, the twist at the end is related to this, but in case someone wants to read the book and doesn't want to have it spoiled, I'll block it out :)

Aliens end up wiping out all life on Earth, because they realize that Earth is able to make anti-matter powered rockets that can travel at relativistic velocities. Because the craft are, essentially, unstoppable, they pose a threat to any other civilization. Thus, the even slight risk of genocide is unacceptable, so the aliens take steps to wipe us out. Two people are saved, who happened to be in a sub below the ocean when the strike hit, and they're taken to the alien craft. Basically, the alien's machines had taken over, and it was their pure logic to wipe us out before we ever had a chance or inclination to wipe them out.

It's an interesting read. Slow in parts, the story could use some work, but it's interesting for the ideas.

[Johnny Pneumatic]

Originally posted by Eos of the Eons
Evolving AI, scary thought. Would they mimick nature? They aren't "natural".

Anything within the universe is "natural". I think you might be suprised how biological looking robots could someday look. You might have to look at them under an electron microscope to tell the difference. Natural Selection would work on them the same way it does with us. The best forms out perform the less efficient etc. They would have arms races just like life does. Take a look at some of the amazing forms dinosaurs had. One even had a "knife" on the end of its tail and you've probably seen an illustration of the clubbed tail dino.

[Dymanic]

Originally posted by Huntsman

You know, I believe this might be one of the best ways to achieve AI..creating a replicating system that inserts mutations into it's progeny, with some sort of selective pressure to remove the less intelligent ones, etc.

Is it entirely crazy to suggest that such a process is underway already?

It might not even be necessary for us to explicitly strive to create intelligent machines. Machines are being replicated. They are mutating. Humans provide the selective pressure. Unsuccessful designs are eliminated -- not just explicitly (as when a farmer culls a herd) but implicitly (when products are edged out of the marketplace by more successful ones). Individual projects have specific goals, defined by humans. But the process of defining those goals is a hit-or-miss affair; whether they have been achieved -- or whether they were valid in the first place -- is determined implicitly by other humans when they use the machines (or don't). Taking place at a level above that of the individual project is a larger process of evolution of design.

Perhaps something at least indistinguishable from intelligence will emerge more or less spontaneously; a natural(?) consequence of our continued efforts to create machines that interact with humans in flexible and intuitive ways.

[CBL4]

According to Antonio Damasio in "Descartes' Error : Emotion, Reason, and the Human Brain", emotions are necessary for us to act logically. Intelligent people with their brains emotion center damaged, can no longer make logical decisions.

From Amazon:
From Publishers Weekly
Neurologist Damasio's refutation of the Cartesian idea of the human mind as separate from bodily processes draws on neurochemistry to support his claim that emotions play a central role in human decision making.
http://www.amazon.com/exec/obidos/ASIN/0380726475/qid=1097791268/sr=2-3/ref=pd_ka_b_2_3/103-1437002-0655028

I have not read the book but I heard him lecture. As a software engineer who had recently taken an AI class, what I got out of it was that the mind is an incredibly effective pattern matching device and that emotions provide the heuristics behind the brains internal search mechanism.

To put it another way, our brains are much too complicated to do anything that requires a significant search of our all knowledge. Emotions prune the search tree to a manageable level. If some action caused us pain in the past, we generally do not even consider that action again. If some action caused us pleasure in the past, we do not even consider actions that might be better because we "know" the pleasurable action is correct.

Could such an "emotion" algorithm work in a robot? I would assume so. But if we tried to make it emulate our feeling would the robot be feeling or only logically following an unemotional algorithm?

Also, we have a vast history of evolution that led us to our emotion filled brain. Just because we have it does not imply it would be the best solution. Evolution kludges together parts from the past in ineffecient manner.

CBL

[Peskanov]

Huntsman,

Now, as to your point about rulesets that are reflexive..i.e.-that can change themselves, I agree that they can be intelligent. However, and the basis of my point, is that if the ruleset can change itself then it can ovewrite the safety rules you put in place, or create new rules that go far beyond its original programming or intent.

Oh, ok then. I thinked you were ruling out all models based on rulesets.

Dynamic,

Perhaps something at least indistinguishable from intelligence will emerge more or less spontaneously; a natural(?) consequence of our continued efforts to create machines that interact with humans in flexible and intuitive ways.

This idea has been proposed many times by good thinkers;
In the seventies, Stanislaw Lem wrote a short tale ( an essay disguissed as tale in reality) called Golem XIV about an sponteous intelligence emerged from a militar project. This is probably the best narration I have read about this idea, by far.
BTW, in the tale the AI has no personality although it can simulate several.

CBL4,

Also, we have a vast history of evolution that led us to our emotion filled brain. Just because we have it does not imply it would be the best solution. Evolution kludges together parts from the past in ineffecient manner.

Agree 110%. I think it would be a bit naive to expect a close similarity between a natural and an artifical intelligence, unless the artificial one is built on purpose to match our brain.
Usually artificial systems lack the abundant redundancy found in then living organisms which acomplish similar functions.

[CBL4]

Originally posted by Peskanov
I think it would be a bit naive to expect a close similarity between a natural and an artifical intelligence, unless the artificial one is built on purpose to match our brain.Of course, the one device that we know work intelligently is the human brain. It does make sense to copy it to a large degree.

I also remember a quote from somewhere, "Why would want to design a machine to be like a person. If we want a person, there are already billions of them out there."

CBL

[Eos of the Eons]

Originally posted by CBL4
Of course, the one device that we know work intelligently is the human brain.
CBL

Are you sure of that? :D :D

We do have our definition of intelligence. Could there be a better design? What could that possibly be? Are we capable of imagining something better, more efficient, with less chance of "flaws" (depression, schizophrenia, out of control anger, ADD)?

I have heard that our model of intelligence includes genes that cause schizophrenia, schizophrenics have too many of them...too much of a good thing...

I haven't heard of much follow up on this. Can we make something capable of "extreme intelligence" that includes the necessary emotions that won't have some fatal flaw?

I'm trying to imagine this, and I don't know if I'm capable. I feel so inadequate :D

[El Greco]

Originally posted by Eos of the Eons
Can we make something capable of "extreme intelligence" that includes the necessary emotions that won't have some fatal flaw?

Without doubt. For one, when we will have made the "brain on a chip" thing, that brain will be immortal. Imagine a human living for thousands of years and collecting emotions and experiences. It will also be possible to combine the experiences of more than one of these brains. And all these without even trying to think what would happen if we try to "improve" on the design of the human brain.

[Peskanov]

CBL4,

Of course, the one device that we know work intelligently is the human brain. It does make sense to copy it to a large degree.

I also remember a quote from somewhere, "Why would want to design a machine to be like a person. If we want a person, there are already billions of them out there."

It could be a quote from Minsky, it looks his thinking.
There are tons o reasons to try to build a copy of a machine we don't understand, and one of the most important is that helps us to understand more about itself.
BTW, it's already being done for medical reasons. I have pointed in several threads about AI that an artifical hippocampus is already in test phase.

Eos,

We do have our definition of intelligence. Could there be a better design? What could that possibly be? Are we capable of imagining something better, more efficient, with less chance of "flaws" (depression, schizophrenia, out of control anger, ADD)?

We do have that posibility right now with any kind of animals, including humans. You just have to use selective breed, like in dogs.
In fact some crazy libertarians want to do just that in places like "new-utopia" :D
This "directed evolution" is used in AI systems in the form of artificial eco-systems where you discard the silly agents and clone and mutate the smarter ones.
As usual is a big problem of ethics...

[Eos of the Eons]

Thanks guys.

I will try to have some faith in fellow humans and their abilities when it comes to technology, even this kind of technology. Erase my fears with facts, and I will be all for this.