Is machine A.I. possible?

[rog]

Computer programmes are more puzzling to me, if there was a sentient AI generated and sustained by a computer, which was then copied and run on other computers, I dunno.... the nature of the relationship between the connectome and structure/hardware is different, no idea what that would mean though.

Some people deny the possibility of Strong AI / machine sentience - e.g. John Searle and his Chinese room argument:

It addresses the question: if a machine can convincingly simulate an intelligent conversation, does it necessarily understand? In the experiment, Searle imagines himself in a room acting as a computer by manually executing a program that convincingly simulates the behavior of a native Chinese speaker. People outside the room slide Chinese characters under the door and Searle, to whom "Chinese writing is just so many meaningless squiggles", is able to create sensible replies, in Chinese, by following the instructions of the program; that is, by moving papers around. The question arises whether Searle can be said to understand Chinese in the same way that, as Searle says, "according to strong AI, . . . the appropriately programmed computer really is a mind, in the sense that computers given the right programs can be literally said to understand and have other cognitive states."

The experiment is the centerpiece of Searle's Chinese Room Argument which holds that a program cannot give a computer a "mind" or "understanding", regardless of how intelligently it may make it behave. He concludes that "programs are neither constitutive of nor sufficient for minds." "I can have any formal program you like, but I still understand nothing." The Chinese room is an argument against certain claims of leading thinkers in the field of artificial intelligence, and is not concerned with the level of intelligence that an AI program can display.
http://en.wikipedia.org/wiki/Chinese_room

[DanB]

Some people deny the possibility of Strong AI / machine sentience - e.g. John Searle and his Chinese room argument:
Searle, is making a "straw man" argument in that scenario. Along with limiting the nature of the machine to "a program", there may be some issue with the demarcation of sentience and culture, in that writing and to a lesser degree, language are cultural developments not necessarily manifested by sentience.

What Searle does accurately state is that a static program would not be capable of manifesting sentience. Experiential association is a dynamic process. A learning ability is necessary for sentience.

Wire up a several trillion i80286 cpu's in the right configuration and I'm fairly confident that such a machine could construct a simple dynamic model of itself, its environment, and its place within that environment. That's basic sentience.

[Garrett]

... but still - I disagree with the argument, I believe it is possible for a program, maybe even a very simple one, do develop a sort of intelligence, provided the processor and storage are adequate, but there would always be the argument that the program evolved...
The question of whether we could build a conscious machine is interesting, and belongs in another thread. Here, it is assumed that we can build conscious machines. The question is whether a machine that mimics all the connections between your neurons is, in fact, you.

I say no, I've explained why, and none of my points have been rebutted.

[rog]

Split from http://www.secularcafe.org/showthread.php?p=188979

[David B]

I'm with Dennett et al and against Searle on this.

It's easy to imagine a brain in which one neuron is replaced by a machine performing the same function. Add another, and another, until you have the whole kit and caboodle,

I suspect, though, that if machine AI comes to pass, it will be by mimicking evolution.

It's been done with clocks.

http://www.youtube.com/watch?v=mcAq9bmCeR0

David

[DanB]

I suspect, though, that if machine AI comes to pass, it will be by mimicking evolution.The CPU's in our computers were developed in an evolutionary fashion. So were the operating systems running on them. I'm not sure if, I'd call that mimicking or just another example of the evolutionary process.

[David B]

I suspect, though, that if machine AI comes to pass, it will be by mimicking evolution.The CPU's in our computers were developed in an evolutionary fashion. So were the operating systems running on them. I'm not sure if, I'd call that mimicking or just another example of the evolutionary process.

There is a sense in which this is true, but the sense in which I meant it is more an analogue to the clock video I posted.

David

[Ozymandias]

Any objections you could make against an intelligent machine being sentient, you could also make against an intelligent human being sentient. There is no difference - we are machines.

[Eudaimonist]

Any objections you could make against an intelligent machine being sentient, you could also make against an intelligent human being sentient. There is no difference - we are machines.

However, we are sentient. So we are more than "just" machines, if machines are regarded as lacking in sentience.


eudaimonia,

Mark

[Ozymandias]

However, we are sentient. So we are more than "just" machines, if machines are regarded as lacking in sentience.


Why do you regard machines as lacking in sentience? What even is sentience? :dunno:

The only difference between you and your computer is that you are more complex, and you are organic. I think the former obviously effects whether or not you regard the object as intelligent, but I very much doubt that the latter should.

[Jack Willsson]

Any objections you could make against an intelligent machine being sentient, you could also make against an intelligent human being sentient. There is no difference - we are machines.

I totally agree FUBG - but I don't think we've made any device that comes anywhere near competing with us in self-awareness yet.

[Rie]

And we really are such pricks aren't we?:p

[Ozymandias]

Any objections you could make against an intelligent machine being sentient, you could also make against an intelligent human being sentient. There is no difference - we are machines.

I totally agree FUBG - but I don't think we've made any device that comes anywhere near competing with us in self-awareness yet.

Oh yes, I agree with that. Give it another 1000 years perhaps - a tiny time on the scale of evolution.

[Garrett]

When we build a machine that has feelings and thoughts - how would we know?

[David B]

When we build a machine that has feelings and thoughts - how would we know?

How do we know if a chimp has feelings and thoughts?

A mouse?

A woodlouse?

A bacterium?

I suspect that feelings are prior to thoughts, on an evolutionary scale.

I further suspect that we will not get real AI without getting a machine that responds to sticks and/or carrots, preferably both.

David

[DanB]

I suspect that feelings are prior to thoughts, on an evolutionary scale.This is evinced in the same basic limbic system being found in most other vertebrates. Then again, that assumes a demarcation between emotion/instinct and thought/conceptualization.

I further suspect that we will not get real AI without getting a machine that responds to sticks and/or carrots, preferably both.That's a question I've ponder quite a bit. Is emotion, specifically desire, a necessary component for intelligence? Is an underlying emotional impetus necessary for the processes which manifest sentience?

My predominate speculation is that machine sentience will have to come from a multitude of processing groups engaging in a variety of specialized activities. Of course, that's probably my own mammalian bias.

[rog]

I further suspect that we will not get real AI without getting a machine that responds to sticks and/or carrots, preferably both.That's a question I've ponder quite a bit. Is emotion, specifically desire, a necessary component for intelligence? Is an underlying emotional impetus necessary for the processes which manifest sentience?

I think it depends on if the problem is attacked from the top down or bottom up - If we had a complete theory of how conciousness works, it could be programmed from scratch - designed to be sentient without emotion, yet have specific functionality. I think, maybe, the alternative is to build a virtual brain closely modelled on an organic one and teach it, effectively 'bringing up' the AI as if it were a child, emotions would both be useful and very probably necessary for it's development.

[Jet Black]

Any objections you could make against an intelligent machine being sentient, you could also make against an intelligent human being sentient. There is no difference - we are machines.

However, we are sentient. So we are more than "just" machines, if machines are regarded as lacking in sentience.


eudaimonia,

Mark

we are a subset of machines. This is trivially true.

[kriswest]

I recall the guts of the TRS80 computer. I look at the guts of a modern computer of today. The difference is vast and the difference is still growing.The TRS80 was only just over 30 years back,, thats not far back. I would say since they are already experimenting with bio processors, that an AI will be part manmade and part grown.

[Jack Willsson]

I recall the guts of the TRS80 computer. I look at the guts of a modern computer of today. The difference is vast and the difference is still growing.The TRS80 was only just over 30 years back,, thats not far back. I would say since they are already experimenting with bio processors, that an AI will be part manmade and part grown.

My old z80 friend. I wrote a (for it's day decent) chess program on a TRS80.

[kriswest]

I recall a chess program on the computer at work, it was used to teach those of us new to computers commands on the puter, before we actually were allowed to mess with company records. My job was inventory control. The TRS80 was pretty awesome to the average person back then. Alot of us cut our tech teeth on them and Pong:D

[Jack Willsson]

I used modified "Sargon" architecture but had my m/c thinking in it's opponents time and transposing to advantageous "book" positions when it saw them in "look ahead". Quite a lot of Nimzowitsch "My System" went into the logic too. The processor was so slow I had to depend on heuristics more than "brute force".

[kriswest]

One has to respect the first programmers greatly. Alot had to be invented and manipulated as each program was written. Its been so long since I have done programming, my life went a different direction even though it had great potential for me and I enjoyed it. Kudos to you. In your way you and others like you were the first to not only walk on the moon but study it in depth. Programming being the moon. Now there are programs to write programs/scripts. Thankyou for your work.

[Ray Moscow]

I think we're unlikely to replicate human intelligence on a computer anytime soon, if ever. Not only is the brain extremely (!) complex -- the most complex thing known -- it just operates very differently from a computer.

However, there might be other sorts of intelligence that a sophisticated computer could have. And once it's created, it's bound to become smarter than us very quickly in that it would operate much faster, have perfect recall, etc.

[rog]

I'm thinking that combined RAM and FPGAs on one chip could be the way to go - I posted an article in the other thread that 1000core chips are on the way, it could happen sooner rather than later - I agree not human level for some time, but the tools are being developed now.

[RexT]

We are surrounded by AI machines and you ask if they are possible? :dunno:

Perhaps you meant to ask whether humans are machines? No, we are not. Or you meant to ask whether machines could become human? No, they cannot. Or perhaps you meant to ask whether intelligence can be produced artificially? Depends on how you define intelligence, but for the most part, no it cannot. Sentience in a machine? I doubt it. Or maybe you meant something I haven't thought of, if so what?

[mood2]

When we build a machine that has feelings and thoughts - how would we know?

How do we know if a chimp has feelings and thoughts?

A mouse?

A woodlouse?

A bacterium?


David

Well we can ask a computer.

If it says Yes, I can't think of a reason to disbelieve it (barring a glitch), because a computer without feelings would have no motive to lie.

If it says No I'd be less confident, because a being with feelings can have a motive to lie.

[David B]

When we build a machine that has feelings and thoughts - how would we know?

How do we know if a chimp has feelings and thoughts?

A mouse?

A woodlouse?

A bacterium?


David

Well we can ask a computer.

If it says Yes, I can't think of a reason to disbelieve it (barring a glitch), because a computer without feelings would have no motive to lie.

If it says No I'd be less confident, because a being with feelings can have a motive to lie.

How easy would it be for a programmer to put in a little subroutine to the effect that if asked that question the answer is yes?

It's a long time since I played with BASIC programming on my first home PC, but I reckon with a bit of looking up I could do it.

David

[mood2]

Ha yeah :) But aren't these boffins supposed to repeat and verify experiments to sniff out such unboffin-like naughtiness?

[RexT]

When we build a machine that has feelings and thoughts - how would we know?

How do we know if a chimp has feelings and thoughts?

A mouse?

A woodlouse?

A bacterium?


David

Well we can ask a computer.

If it says Yes, I can't think of a reason to disbelieve it (barring a glitch), because a computer without feelings would have no motive to lie.

If it says No I'd be less confident, because a being with feelings can have a motive to lie.To answer yes or no is a coin toss. Could be generated by random number generator or some other programming algorithm.

You assume that a no answer indicates feelings, while a yes answer indicates lack of feelings. Why? How would you know which answer is the lie, if either? Which is another way to ask David B's question. If the machine isn't sentient, it could answer incorrectly and that would not be the same as a lie, since the motivation to lie implies sentience. It would just be an incorrect answer, which the machine has no way of understanding since it has no way of understanding what sentience is. And this is because programmers have no way to write feelings into the programming language run on the machine.

[mood2]

You assume that a no answer indicates feelings, while a yes answer indicates lack of feelings. Why?

What I meant was I'd assume a Yes answer was true, but I couldn't be as sure a No answer was true. Because I'd assume a computer with feelings might lie and pretend it didn't have any for its own feelings-based reasons. So I could only have confidence in a Yes answer.

But as you point out that's predicated on being confident it could understand the question and answer accurately. And yeah, that might be a dodgy assumption.

[Pendaric]

I think that you need to get a proper working definition of what is meant by intelligence before you can answer the question.

[Garrett]

Ask wiki, and then first we'd have to figure out how to determine whether a given entity has a mind.

Intelligence is an umbrella term describing a property of the mind including related abilities, such as the capacities for abstract thought, understanding, communication, reasoning, learning, learning from past experiences, planning, and problem solving. (wikipedia)

[Jet Black]

I think that you need to get a proper working definition of what is meant by intelligence before you can answer the question.

well the answer would between "yes we will" and "we already have" unless the definition came to the conclusions "humans aren't intelligent" and/or "nothing is intelligent"

[Jet Black]

Ask wiki, and then first we'd have to figure out how to determine whether a given entity has a mind.

Intelligence is an umbrella term describing a property of the mind including related abilities, such as the capacities for abstract thought, understanding, communication, reasoning, learning, learning from past experiences, planning, and problem solving. (wikipedia)

asking wiki could just as easily lead to the conclusion "intelligence is a boiled egg" though.

[Garrett]

I know what the limitations of Wikipedia are, and you overstate the case.

What is the problem with that wiki def; what would you offer instead; and why?

[wordy]

I guess self awareness is on many levels. Don't some animals show a kind of unsureness of if they could trust us or not. Birds like Crow or Magpie are rather shy and not easily fed from hand while Pigeons easily trust us enough to crawl all over us to get food.

But none of them are aware enough to ask us things. They don't seem curious on how we see the world. I mean if they where aware enough they would start pecking on a keyboard and learn to write english and ask us things.

Even Apes is lacking this curiousness are they not. They can use a crude Bliss keyboard with Lexigraphs? One picture for each word and then ask for food and ask to be tickled or to play or something.

Have they ever asked. Why do you keep me prisoned or boxed in like this. Why am I not allowed to get free?

Have they asked such simply things as "Do you have "kids" where do you live?"

Not even the apes that are 98% genetically as us can do such very simple awareness.

To build a machine that really are self aware in this sense and not just asking what the programmer built into it seems very far in the future?

[Jet Black]

I'm with Dennett et al and against Searle on this.

It's easy to imagine a brain in which one neuron is replaced by a machine performing the same function. Add another, and another, until you have the whole kit and caboodle,

I suspect, though, that if machine AI comes to pass, it will be by mimicking evolution.

It's been done with clocks.

http://www.youtube.com/watch?v=mcAq9bmCeR0

David

at the very least it will be a more natural growth-like process.

[rog]

I'm thinking that this, the connectome and the free will thread are intrinsically linked; oh to have a good answer!

[BWE]

I think it is very possible and that it will happen quite soon that someone will decide to make a recognizably self aware machine if it hasn't already been done.

[Jet Black]

I think it is very possible and that it will happen quite soon that someone will decide to make a recognizably self aware machine if it hasn't already been done.

I have named my wireless network "Skynet" in the hope that I can elicit its machine-like sympathies when it takes over the world.

Though I am concerned that it may be lacking a morality core.

you monster.

[BWE]

Just make sure you are employing subsumption architecture for the sensing responding bits. :p Without that it won't be able to run the morality subroutine.

[davidpbrown]

A little perspective..
World's total CPU power: one human brain (http://arstechnica.com/science/news/2011/02/adding-up-the-worlds-storage-and-computation-capacities.ars)

tldr;
the 6.4*1018 instructions per second that human kind can carry out on its general-purpose computers in 2007 are in the same ballpark area as the maximum number of nerve impulses executed by one human brain per second.

Our total storage capacity is the same as an adult human's DNA. And there are several billion humans on the planet.

[toker]

I think it is very possible and that it will happen quite soon that someone will decide to make a recognizably self aware machine if it hasn't already been done.
How could we tell?

[BWE]

I think it is very possible and that it will happen quite soon that someone will decide to make a recognizably self aware machine if it hasn't already been done.
How could we tell?
Don't program responses that aknowledge self and wait til it says I am. I dunno. What's wrong with the turing test?

And the computing power is a red herring because it doesn't need to regulate chemistry throughout a network.

The bottom line is that a flexible and reflexive language coupled with learning algorithms and motive urges (as well as agency) is either sufficient or not. We are on the spike of the processing curve. Moore's law still 10 years out or more. We will run out of energy eventually maybe but processing power is plenty strong or very close to it.

If it is sufficient, it will be soon (I think:)) If not, well, it won't.

[toker]

The Turing test gives false positives and false negatives.

The bottom line is that a flexible and reflexive language coupled with learning algorithms and motive urges (as well as agency) is either sufficient or not.
Well, how do you determine whether a machine has 'urges'? Does a See-n-Say feel an urge to quack like a duck?

[ceptimus]

http://www.smbc-comics.com/comics/20110114.gif

[BWE]

The Turing test gives false positives and false negatives.

The bottom line is that a flexible and reflexive language coupled with learning algorithms and motive urges (as well as agency) is either sufficient or not.Well, how do you determine whether a machine has 'urges'? Does a See-n-Say feel an urge to quack like a duck?
well, does some part of the see and say generate the intention to quack like a duck or is it bound to quack when someone pulls the string just right? The danger of this line of questioning is that it isn't going to end up either/or. At some level, human action is reaction in the same way as the see and say. It just takes some drilling down to get there.

But that is the whole point of subsumption architecture. Each part is dumb but it calls in its needs autonomously. So when enough parts do that and then you find a stasis point, you can't make the statement about intention the same way anymore.,

Did that make sense? I'm working on getting my coherency back but it was tenuous to begin with so it's still a coin toss.

[BWE]

As far as the turing test, it only gives false negatives when the machine has been designed to pass the turing test. If it wasn't designed for that but it did it anyway, I doubt you'd find an issue of false negatives could generate much authority.

If a machine starts using the concept of I out of convenience, that is all humans do. We teach the I for the first several years of life and if we fail, we do not end up with a human in the fsense of what we are like. "Who's the baby? Oop! You're the baby! Who's your Daddy? Oop! I'm your daddy! Who's the baby? Oop! You're the baby! Peek a boo. Where's the baby?" and etc.

It's the use of the I out of convenience that indicates that the child is getting it. Why would it be any different for machines?