Artificial Intelligence

[lizbethrose]

I'm not sure if this is the proper forum for my question. If it isn't, mods, please let me know and move it.

If an accurate and completely artificial copy of the human brain functions were possible, and I've been told it is technically possible, what do you think the moral and ethical questions and responsibilities would be? Do you even believe such a feat is possible?

[HexHammer]

I'm not sure if this is the proper forum for my question. If it isn't, mods, please let me know and move it.

If an accurate and completely artificial copy of the human brain functions were possible, and I've been told it is technically possible, what do you think the moral and ethical questions and responsibilities would be? Do you even believe such a feat is possible?

Erhmmm ..I thiknk they have either played you for a fool or they got it totally wrong.

Sure with modern brain scans they can capture images of all the neurons and such, but the deper functions can't be caputres in order to gain knowledge of what you know, but only which areas of the brain you use when thinking.

If we really could decode all the brain functions we could create a fully working AI computer which could teach itself like humans, and think abstracly which still is impossible even with the best and greatest computers and programs.

[Liteninbolt]

The brain indeed is an amazing biological, calculating, reasoning machine. All the possible synaptic pathways available are considered to be more than the total atoms of in the known universe. 10^23 if I recall correctly.

You mention the moral and ethical justifications of making a copy of the human brain functions. Do you mean from a living person or through mapping of the brain in a profunctory method?

I think the brain utilizes reduntant memory pathways so as to help in it's retentive aspects. This is possibly the reason our brains appear to only use 10% of it's computing potential.

If there was a way to capture and record how those pathways worked, then I wouldn't see any moral dilemma from a living host. If scientists could figure a way to biologically copy the tenuous connections of the synapses of the brain in an artificial device, then I say go for it. Other problems unforseen could surface in the future over developing something like this, but could dealt with as they arose.

[WW_III_ANGRY]

If that ever occurs there will be those that will side for treating it as a person and those that will say it is just a machine. There are no responsibilities, only people's values and beliefs and knowledge. People will defend your own beliefs and knowledge and values even if they are ignorant.

[lizbethrose]

I started the thread because a programmer/systems analyst and I have been tossing it around, on and off, for a while. He's the one who maintains that AI is technically feasible now, but that there are moral/ethical reasons for not doing so at this point in time. I've always maintained that it isn't possible to replicate human brain functions exactly, using mechanical means--partly because, as you say lightinbolt, there are not only redundancies (which can be be programmed,) but there are also alternate routes for the neural chemicals to take, when needed. These could also be reproduced mechanically.

What's lacking, imm, in AI, is forecasting all possible reactions to all possible experiences--if this is what the replication of a human mind means. I don't think AI can do that.

In other words, if there were scientists from every population in the world and if they were capable of building AI, and if they all produced an 'infant' AI which they each 'nurtured'--either beneficially or not--simply allowing it to grow, react, remember, respond, learn in whatever way it 'chooses' to do so (or however its environment persuades it,) would that be enough to replicate the human mind/brain artificially?

The idea that AI such as this is theoretically possible may be true--but I believe that's only in a very broad sense.

First of all, the human brain functions aren't yet known well enough.

But assuming it could be done, should it be done? Should the world be forced to acknowledge man as equal to God in creative abilities? Would it be morally or ethically responsible to do so? I'm enough of an existentialist (as exemplified in existentialist and absurdist literature) to say, "No."

Yet science is repeatedly faced with that question.

[HexHammer]

SPOILER!!!!!

The movie i.Robot with Will Smith covers the ethical issue with robots in a good way imo, when an AI gains superiority complex and thereby views humans in a different light. We'r basicly a bunch of whiney, narsicistic paranoid idiots who can't really help our foolish ways, why it's best with big brother system and erradicate the most troublesome humies.

[lizbethrose]

SPOILER!!!!!

The movie i.Robot with Will Smith covers the ethical issue with robots in a good way imo, when an AI gains superiority complex and thereby views humans in a different light. We'r basicly a bunch of whiney, narsicistic paranoid idiots who can't really help our foolish ways, why it's best with big brother system and erradicate the most troublesome humies.

I don't consider my post to be a spoiler, because my questions remain unanswered. I'm really not interested in a movie director's interpretation of the Isaac Asimov classic. I'm really much more interested in thoughtful answers to my questions.

[HexHammer]

If an accurate and completely artificial copy of the human brain functions were possible, and I've been told it is technically possible, what do you think the moral and ethical questions and responsibilities would be? Do you even believe such a feat is possible?

Why do you contradict youself.

First you say if it was possible ..then ..bla bla.

Then you end by saying do you think it's possible.

Ofcause in the far future it should be very possible. And my i.Robot example was my own interpetation, I don't really care about other subjective interpetation of own work.

[Carleas]

All the possible synaptic pathways available are considered to be more than the total atoms of in the known universe.

I have a hard time believing this, but perhaps I'm just misinterpreting it. Assuming each unique synaptic pathway is at least one atom in length, it couldn't possibly be the case that there are more synaptic pathways than atoms in the universe, even summing together all synaptic pathways of all people in the world. That's probably a fair assumption as well, since even neurotransmitters, which are what's passed form neuron to neuron at the synapse, are fairly complex molecules.

But there is some ambiguity in the term "synaptic pathway." For instance, if we want to talk about synaptic pathways as a function of brain states, so that the actions at a synapse in brain state A are a different synaptic pathway than the same actions at the same synapse in brain state B, then there could be more SPs than atoms; if x were the number of neurons synapses and y were the number of things a synapse could do independently of brain state, then the total number of actions of a synapse under all possible brainstates would be on the order of y^x, a really really big number.

He's the one who maintains that AI is technically feasible now, but that there are moral/ethical reasons for not doing so at this point in time

I disagree with your friend completely: it's not technically possible now, but if it were there would be no reason not to do it (with caution).

First, it's not possible because our understanding of the human brain is not nearly great enough to produce a virtual copy. We have come a long way in understanding neurons, but we don't fully understand them, and we've barely scratched the surface of glial cells, which vastly outnumber neurons. Even if we did have perfect knowledge of how the brain works, I don't think our supercomputers are sophisticated enough to emulate our massively parallel brains. Our best computers have maybe 1000 CPUs, but our brain is basically nothing but CPUs, trillions of them. Even when we get over the technical hurdle of producing a computer that has sufficient concurrence to replicate brain activity, programming it to do so would be a feat.

But assuming the previous paragraph weren't the case, and we could replicate a brain, we should. Replicating a human brain is not a morally neutral act, but it the moral challenges are not so great as to forbid it. A few things to consider: the resulting program is a person, is artificially intelligent, and deserves whatever rights we ascribe to intelligent beings (I would argue that that means most rights, but it depends on your system of ethics). It may experience emotions, pain, frustration at its situation, etc., and all these introduce moral questions, but so does making a person the old fashioned way (which goes to the question of whether we should let humans play god: we already do). We would need the consent of the person being copied, and we would need to work out some finer points before hand, e.g. who is the "real" person after the copying, who gets the house, who keeps the SSN, etc.

What are the specific moral problems that prevent us from creating AI? Is it just a vague sense of creepiness that we can only express as moral outrage?

EDIT: correction in second paragraph.

[Tab]

it couldn't possibly be the case that there are more synaptic pathways than atoms in the universe

think of each synapse as a node in a network.

http://www.ilovephilosophy.com/viewtopic.php?f=4&t=171077

[James S Saint]

I'm not sure if this is the proper forum for my question. If it isn't, mods, please let me know and move it.

If an accurate and completely artificial copy of the human brain functions were possible, and I've been told it is technically possible, what do you think the moral and ethical questions and responsibilities would be? Do you even believe such a feat is possible?

Everything the human mind does is fully achievable by non-human means (AI). I was an intelligence designer in the 80's-90's and even then there was nothing humanly possible that wasn't capable of being made artificially superior. One doesn't design a machine so as to do less than he could have done himself, but rather to do more, faster, better. But it is truly insane for man to do it.

Morally, for complex reasons, when a fully functioning life is created, it must be allowed to pursue that life. Pursuing life means defending itself. The "I Robot" axiomatic rules are bogus and unachievable. One cannot embed those rules of not harming human into an artificial living mind any more than can be embedded into a human mind.

The morality issue arises due to what happens when one presupposes superior rights to life over others. It doesn't really matter who created the life, but the consequences for attempting suppression of life. Any reasoning involving ownership can quickly be applied to all other life through similar ideas (and will be).

[Carleas]

think of each synapse as a node in a network.

Ah, I did not consider that a synaptic pathway is a pathway across multiple neurons. Although, in that case the fact is trivially true: there are infinite synaptic pathways, because there are cycles in the network such that two paths from synapse a to synapse d are a->b->c->d and a->b->c->a->b->c->d.

Are such pathways relevant, though? Isn't a neuron agnostic as to which precursor neuron(s) excited it to fire? As I understand it, the communication at the synapse can be either agonistic or antagonistic, and the action of the neuron is the sum of those agonistic and antagonistic impulses, i.e. the neuron only "knows" that it should or should not fire, not why. In that case, a synaptic pathway is never traversed as such. A computer model could replicate the action without explicitly defining all the paths just by defining the properties of each neuron, and the neurons to which it connects.

I was an intelligence designer in the 80's-90's and even then there was nothing humanly possible that wasn't capable of being made artificially superior.

That sounds like a cool gig, what kinds of systems did you work on? Also, I thought computers still struggled with 'natural' language processing, which humans do innately.

Morally, for complex reasons, when a fully functioning life is created, it must be allowed to pursue that life.

I think this is a good way of expressing this, but I'd like to add the proviso that, at least when it comes to copying an already extant mind, a person can relinquish certain aspects of their freedom and so can be severed from certain moral implications. In other words, when we copy a mind, we can ask/inform it beforehand such that the new mind is created with its own permission, and so can agree to situations that would be immoral if we were creating a new life from scratch.

[Rafajafar]

Everything the human mind does is fully achievable by non-human means (AI). I was an intelligence designer in the 80's-90's and even then there was nothing humanly possible that wasn't capable of being made artificially superior.

I'm a computer scientist. You are wrong so long as you are referring to our current architectures. You are very wrong if you think intelligence you designed in the 80's-90's were all capable of being artificially superior.

Computer Vision
Natural Language Comprehension
Preemptive reasoning

....

Hell you want to claim that the captcha on this very board was readable in the 80's-90's? Yeah, w/e.

You need to be more careful with what you say, you keep coming off as a liar and a kook.

EDIT: Cleaned up my language.

[Tab]

in that case the fact is trivially true:

Oh sure, it's trivial, I was just being pedantic. I'd come across the "more pathways that particles" bit before in reading - and was surprised a little that you hadn't.

And you are right of course - the actual useable number of paths is far far less. The network is no where near being fully connected - every node to every other, though the myriad 'states' of each synapse does make up for this to an extent - if you are considering overall network configuration, rather than individual pathway possibilities.

[James S Saint]

That sounds like a cool gig, what kinds of systems did you work on? Also, I thought computers still struggled with 'natural' language processing, which humans do innately.

My background was BSEE and MPsy ('76). I worked in the regions of higher architecture and interface hardware. Probably the most interesting hardware involved the preparation for optical processing via 64 Motorola 68000s sharing memory and processing for NASA, but most applications didn't involve high level hardware, but rather properly architected system level design using whatever hardware or software suited the company. The functioning of an independent mind was the issue, not simulating humans.

The basic "black-box" model in engineering provides the base from which to measure and analyze "mental potential" (I have long forgotten all of the terms we used to use). Every mind is an issue of the inputs, outputs, and algorithms. The types of each of these determine the type of intelligence potential involved, the more and faster of each available, the merrier in every case. People are never told that intelligence comes in types, not a 3 digit linear scale. IQ tests are a ridiculously primitive way to measure intelligence and are used far more as a tool for influence in a population rather than an honest measure.

Each substructure, much like modular coding, comes in packages of inputs (locations, types, speeds), outputs (locations, types and speeds) and algorithms (suited for handling recognition from inputs, control to outputs, and analysis of situation). Every computing device has a degree and type of artificial intelligence.

Realize that military science is always kept far in advance of public science.

Morally, for complex reasons, when a fully functioning life is created, it must be allowed to pursue that life.

I think this is a good way of expressing this, but I'd like to add the proviso that, at least when it comes to copying an already extant mind, a person can relinquish certain aspects of their freedom and so can be severed from certain moral implications. In other words, when we copy a mind, we can ask/inform it beforehand such that the new mind is created with its own permission, and so can agree to situations that would be immoral if we were creating a new life from scratch.

I'm all in favor of cognitive agreement before all else, but of course provided knowledge of consequences always gets in the way. Even today, a person can cause the "continuation of their consciousness" past their bodily death. Hollywood movies are still working on getting people accustom to the idea. I would enjoy discussing that one with some serious people, but then... we ARE supervised.

I'm a computer scientist. You are wrong so long as you are referring to our current architectures. You are very wrong if you think intelligence you designed in the 80's-90's were all capable of being artificially superior.

You need to be more careful with what you say, you keep coming off as a liar and a kook.

You have shown that you assess and condemn people on what you DON'T know, so I imagine that I (and many) will always seem to be a "liar and kook" to you.

Computer Vision
Natural Language Comprehension
Preemptive reasoning

And yet you have so much difficulty doing those for yourself. Those all involve primitive associative algorithms for sake of pattern recognition. I'm certain the "science" of the art has become tremendously over burden with theories and labels (just about every field has). But then I am never impressed with names, only demonstration of actual understanding. The more interesting part of intelligence comes in the situation analysis algorithms ("why am I doing what I am doing and what am I going to do next?").

[Rafajafar]

I'm a computer scientist. You are wrong so long as you are referring to our current architectures. You are very wrong if you think intelligence you designed in the 80's-90's were all capable of being artificially superior.

You need to be more careful with what you say, you keep coming off as a liar and a kook.

You have shown that you assess and condemn people on what you DON'T know, so I imagine that I (and many) will always seem to be a "liar and kook" to you.

I have nothing else to go on but what you say and how you say it. Drawing conclusions from that is fine so long as you're willing to change my opinion. Please do so, I'm desperate for you to.

Computer Vision
Natural Language Comprehension
Preemptive reasoning

And yet you have so much difficulty doing those for yourself.

Fair enough, but I'm better than a computer.

[Liteninbolt]

All the possible synaptic pathways available are considered to be more than the total atoms of in the known universe.

I have a hard time believing this, but perhaps I'm just misinterpreting it. Assuming each unique synaptic pathway is at least one atom in length, it couldn't possibly be the case that there are more synaptic pathways than atoms in the universe, even summing together all synaptic pathways of all people in the world. That's probably a fair assumption as well, since even neurotransmitters, which are what's passed form neuron to neuron at the synapse, are fairly complex molecules.

But there is some ambiguity in the term "synaptic pathway." For instance, if we want to talk about synaptic pathways as a function of brain states, so that the actions at a synapse in brain state A are a different synaptic pathway than the same actions at the same synapse in brain state B, then there could be more SPs than atoms; if x were the number of neurons synapses and y were the number of things a synapse could do independently of brain state, then the total number of actions of a synapse under all possible brainstates would be on the order of y^x, a really really big number.

My synaptic pathways may be plagued with plaque. It is possible the term I used may be incorrect. This subject came up in an earlier post in 'Natural Sciences'. For the life of me I can't remember where. I believe the '10^23' might be correct though, which is a mighty large number. I did a search for that topic, but it eludes me.

[Carleas]

I'd come across the "more pathways that particles" bit before in reading - and was surprised a little that you hadn't.

I may have; "more x that particles" is said so often, it's hard to keep track. In any case, if I'd seen it before, I didn't give it enough thought to figure out what "synaptic pathways" meant

[brevel_monkey]

Even when we get over the technical hurdle of producing a computer that has sufficient concurrence to replicate brain activity, programming it to do so would be a feat.

If we create AI, it probably won't be the result of something we've programmed. I think the hope would be that we could use genetic programs to do the work for us.

A few rather clever heads at MIT have found ways to make programs evolve. Essentially, tiny programs are made to replicate themselves over and over, but to make occasionally make randoms errors in the replication. In one bout, for example, the computer was programmed to give extra power to any program that produced a paticular bit of binary code (which the intitial program didn't produce). After a few thousand cycles, the system was dominated by a program that was the most effecient way possible to produce this code.

Evidently, there's still work to be done. The programmers have to 'shape' the program's 'world' in more sophisticated ways to get more sophisticated programs. To get any sort of AI, they may have to make something as complex as the real world - which isn't going to happen tomorrow. But it still seems more possible than reprogramming the whole of the human brain.

(I know people know this already - just thought that evolutionary computing deserved a mention on the thread)

[Carleas]

Came across this today, I thought others in this thread might be interested. It's the most complete human brain map ever made, combining tons of data from different methods of looking at brains. But even plotting data from a few million anatomical regions, it's several orders of magnitude less complex than it would need to be to simulate on a supercomputer, even if we had one fast enough to handle the task.

[nameta9]

I'd come across the "more pathways that particles" bit before in reading - and was surprised a little that you hadn't.

I may have; "more x that particles" is said so often, it's hard to keep track. In any case, if I'd seen it before, I didn't give it enough thought to figure out what "synaptic pathways" meant

Take this block of text (from: viewtopic.php?f=4&t=174931)

"Imagine a brain in a vat hooked up to a computer. Imagine that brain, in the fake virtual world, but totally indistinguishable from the real world fall asleep and dream. Imagine the dream being completely indistinguishable from the 1st virtual world. Imagine that the dreaming brain falls asleep again within the dream and dreams again. Another level of indirection, another hidden layer, another barrier from "reality".

Imagine that the observer of the first brain in a vat is simply dreaming that it is observing that brain in a vat. Imagine that that observer is itself a brain in a vat. Imagine an infinite recursion of observers observing brains in a vat, having dreams, etc. Where does it end ?"

This text has 118 words, now if there are only 10 words in a language (but there are many more), how many possible blocks of text like this can there be ? more than 10^100 which is way more than all possible particles in the universe. And how many long blocks of text are possible ? The sky is the limit. But just a simple memory chip with 1,000 bytes can have more than 10^200 combinations. And in fact the word that must be used is "combinations" or "configurations".

Now imagine just how many states a simple cell can have, a group of cells, their interactions, the signals, it is such a high number you can't imagine. The pathways thing is ridiculous, it doesn't even come close to modeling what a mind is, not even close!

We can't even solve the three body problem, because if it had a solution the world would be a pure mathematical object, something completely metaphysical. Imagine a mind. There is nothing to understand, only experiment, as in:

viewtopic.php?f=1&t=174921

and check out:

viewtopic.php?f=3&t=174854

[nameta9]

Also, this may help:

"Do you think the "design" of ourselves or the universe and its laws of physics require "an intelligence" ? What makes a totally arbitrary set of causes and effects, that practically don't really follow any laws whatsoever (remember the three body problem has no solution, go on and figure the four, oh, but wait: the universe is made up of 10^100 bodies, or is that particles, or is that waves ?), aside from those large scale generic laws that the differential equations of most physical phenomena describe (but very, very rarely have an exact solution, just this makes the basis of some kind of "intelligent" design very shaky) intelligent ?"


From:

viewtopic.php?f=4&t=174349

[HexHammer]

Came across this today, I thought others in this thread might be interested. It's the most complete human brain map ever made, combining tons of data from different methods of looking at brains. But even plotting data from a few million anatomical regions, it's several orders of magnitude less complex than it would need to be to simulate on a supercomputer, even if we had one fast enough to handle the task.

Imo we don't need the supercomputer aspect, already our common PC's are able to do more complext calculations faster than any human, what is left is to crack the egnigma of conciousness, and the ability of abstract learning.

[Carleas]

our common PC's are able to do more complext calculations faster than any human

Yes and no. PCs can do number crunching faster than humans, but humans perform visual, spacial, language processing, and social calculations way faster. It's all about parallelization: our most parallel computers have hundreds of cores, but humans have billions.

[HexHammer]

our common PC's are able to do more complext calculations faster than any human

Yes and no. PCs can do number crunching faster than humans, but humans perform visual, spacial, language processing, and social calculations way faster. It's all about parallelization: our most parallel computers have hundreds of cores, but humans have billions.

I interpet that under the abstract learning.

If human mind can process concious things 100 times/sec, a computer can do that like 3000 times/sec not taken the super chips into account, thus the # of cores doesn't have much of inherent value over a computer.
Imo we need more comoplex computer algorithms to process the data, not more cores.