Jump to content

  • Log In with Google      Sign In   
  • Create Account


Creating A Conscious entity


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
131 replies to this topic

#121 Anonymous Poster_Anonymous Poster_*   Guests   -  Reputation:

0Likes

Posted 04 March 2005 - 07:03 AM

My advice: forget about creating a real conscious entity because conscious is a conceptual label that we apply to things and has nothing to do with what those things really are, whatever they are.

Create a harsh, dynamic world. Then create a thing that can survive in that world, subject to its laws. As someone previously mentioned, some sort of evolving neural net would do - the proof is that that's precisely the origin of our own brains. And we can do this already. The only difference between what we can simulate now, and hyper-complex biological neural nets is brute force: better, faster hardware. What you call it in the end, conscious or not, doesn't matter. Does it have a mind's eye? You can't know, by definition.

Sponsor:

#122 GMScribe   Members   -  Reputation: 122

Like
0Likes
Like

Posted 05 March 2005 - 01:48 PM

Sooo mny posts on this topic.
Sorry if i repeat anything but i'm drunk and it's 2am, so i wont read everything until the morning.

It's theorised our brains work on a quantum level of handling data, allowing us to proccess so much and yes it is a lot, at once.

This would mean the way our brains probably work is no more complex than a linked-list style setup, as in theory, a quantum system would proccess linked lists incrediably fast.

Because our brains work on a chemical level, it's quite posiable this is true.

As for sorting of data (i read the 1st post lol), we use chemical bonds that degrade the less we recall and use data in our brains, allowing most needed data to be picked up first. this also gives room for bonds to new data as old data becomes useless (maintaining routine). Ofcourse, we do filter the information. Depending on a person's personality, we chose to listen to different things etc, but we remember most things in immediate time, so i'm presuming most sorting of data happens in sleep where memory is stored for long-term usage (makes sense). it's the same as copying from memory to hard drive, rather than back the program up and waste space, only keep what will be needed next time the program runs.

Hope that's helpful, i need sleep, night night.

#123 Anonymous Poster_Anonymous Poster_*   Guests   -  Reputation:

0Likes

Posted 06 March 2005 - 05:23 AM

I can see a real problem there. Communication. language. It neither is an application of a fixed set of rules nor a nomenclature. The best example of this are both sense of humor and irony. Imagine yourself typing "You bet!" to a computerized Artificial intelligence. How is this AI going to know if you are straightforward or ironic? Because of a statistical analysis? Nonsense! if the statistics are that this sentence has mostly been used in an ironic sense, then using it in a straightforward context can only lead the AI to misunderstand you. It also works the other way around. And imagine someone trying to interact with the AI using incorrect language, badly spelled language, or insults? How is any artificial intelligence going to recognize a badly spelled word for its true meaning? Have any of you out there tried misspelling something on purpose on WORD? Have seen how many possibilities of replacement words one gets for any word? How many times does the computer land good on first try?

So what to do if the communicaiton is NOT reliable in human language? try and get the input in machine language? Then the man input might be wrong, and the understanding modified. How many coders out there have passed endless hours trying to figure out "why on earth is this going wrong? where have I screwed things up?" You all know what i mean. With a "simple" program of something less than seven thousands lines of command, it quickly becomes something extremely annoying to discover the tiny eror where "," became ";"...
So what you would need would be the AI to both CREATE the code it would use AND WRITE the lines of it, in order for it to be completely sure of what it wrote. It would of course need cross-referencing, data-mining, extrapolation from previously acquired data, auto-modification of the database, auto-modification of interpretations, and deletion when needed, and so on and so forth.

When this problem of comprehension of the language used by the AI is done over, try and tackle the one of comprehension. how would you get a computer to understand what is a chair? Would you begin with the meaning of the word, and give it a definition? then you would end up also giving definitions of the words used to give the definition, and so on and so forth until you could no longer understand the validity of the interrogations. Since the definitions would be devoid of meaning, you would also end up trying to explain the intended use of what you are trying to describe. (A chair is a small, flat and approximately horizontal surface landed on four feet that has a surface on the back to relieve the person who sits on it from muscular pressure in the small of the back. The small, flat and approximately horizontal surface is almost always at the average knee height for an adult.) And this has absolutely nothing to do with the possible uses of a chair that the chair was NEVER intended to have, i.e. be used as a step to get to things too high to reach while not on the chair, or as a rest for clothes or other objects...
How do you implement creativity as regards to interactivity in a program? It is always possible to code an inclination for the program to repeat some procedures which give out some results that the program will dubb "positive", due to the code. But this has nothing to do with creativity or curiosity. This is only a way to treat the lack of information and a possibility to fill it automatically. But I can't see a single way to have the AI understand the information it just gathered.

Mostly because, in the end, an AI is just a very complicated program running on a hardware that grinds everything down to 0s and 1s in order to make it simple. When machines will not operate with zeros and ones only, but with all the tiny fractions that separate them not only in their output, but also in the way they apprehend the world, then, there will be a tiny chance of getting a machine to fake understanding of the world that it feeds its system disc upon.

#124 GMScribe   Members   -  Reputation: 122

Like
0Likes
Like

Posted 06 March 2005 - 07:05 AM

I've never been fond of fussy logic. Our ability to understand irony/sarcasm is partialy due to 'tone'. The difference in interface makes it hard. But you can be ironic/sarcastic over the internet, and that's context and perception. Do you perceive the person ur speakin 2 2 use irony wen they do (based on personality or first time assumptions)? also irony is often more plausable than taking a sentence as none irony i.e. "an old man of 98 won the lottery then died the next day." it's more likely ironic than true, also it's normally said after a word such as 'like'. Sarcasm again can be measured like this. "did he really jump off the curb and die?" "of course he did ;)" well, it's more likely to be sarcasm than dieing from a 3cm jump lol, also, detection of a sign of humour in the sentence where english language states it wouldn't normally be denotes it to be funny or a joke.

Regards to the chair thing. We have built in functions that work WITHOUT a comparison needed, because we have that in common with all other humans. We know that height is the physical Z dimention of an object. If we had no language in common, we could denote size using our hands and act asif we were measuring it. Because our brains work similar and have similar abilities of comprehention, we would all probably denote that it's a show of height. If not, we would eventually realise we are wrong and correct our meaning of the guessture (i.e. the hand sign doesnt work in some circumstances, so it must mean something different). This is why we can't communicate fully to animals. However, we can ensure what procedures etc we build into a program and wouldnt have this issue, as we will know exactly how to talk to an AI.

BTW, our brains do sorta work on 1ns n 0's well, if nythin, real numbers, but they hav an analouge (still limited) degree of accuracy, which is the same as using a binary range of integers like 0 to 1000 to define something.

[Edited by - GMScribe on March 6, 2005 1:05:05 PM]

#125 Anonymous Poster_Anonymous Poster_*   Guests   -  Reputation:

0Likes

Posted 06 March 2005 - 06:04 PM



#126 Anonymous Poster_Anonymous Poster_*   Guests   -  Reputation:

0Likes

Posted 06 March 2005 - 06:08 PM



#127 Anonymous Poster_Anonymous Poster_*   Guests   -  Reputation:

0Likes

Posted 06 March 2005 - 06:09 PM

OK, so humans DO share the same universal knowledge of the world that surrounds them, and can palliate (hope that is right) lack of meaning of a word with gesture and correlate this to meaning in words. As far as I am concerned, it would require your basic AI sensory entries to be able to do that and understand sarcasm or irony. It is all very well that one is able to understand from the tone ones use where the sentence is leading and what aim it has. But without that tone identification, voice recognition (which is far from perfect right now with just one user), pattern recognition, the lot would be lost to use.
Imagine trying to figure out a way to explain what is a "schmorgaborst" or a "rudder". find a way around words with your hands in order to get understood by someone who does not speak your tongue. now you've got an insight of the level of difficulty to get understood by someone. And NOW, think of a way of doing that WITHOUT YOUR HANDS, and without using words you would need to define later in order to be understood. Are you finding this difficult yet? Wait till I ask you to do this in the language of the person you are standing in front. I am French. Try to EXPLAIN in French what a "barter" is without resorting to translation nor to gesture, since the word would probably be alien to me in french. Getting a grasp on it yet? OK! here comes the best part. Now do it in C#. If this one does NOT suit your taste, (and I can understand why) then do it in HTML, in Basic, in Java, in anything as long as your machine does not only stocks the information but also recognizes it as not being the primary input. Tell it what is a Hot-Dog, what is a seller, and show it a picture of Wall Street at lunch time. Can it recognize and point to a Hot-Dog seller? If it can, you got me there. Try showing it different pictures of different races of dogs always telling it they are dogs in the first place, and giving it their races names as subclasses. Try to have it tell you a definition of dog. Try to have it correlate that answer with our previous Hot-Dog without being utterly wrong. Once again, if you can do that, the rest is probably a walk in the park...
Maybe not at night, though...

#128 Anonymous Poster_Anonymous Poster_*   Guests   -  Reputation:

0Likes

Posted 06 March 2005 - 06:10 PM

OK, so humans DO share the same universal knowledge of the world that surrounds them, and can palliate (hope that is right) lack of meaning of a word with gesture and correlate this to meaning in words. As far as I am concerned, it would require your basic AI sensory entries to be able to do that and understand sarcasm or irony. It is all very well that one is able to understand from the tone ones use where the sentence is leading and what aim it has. But without that tone identification, voice recognition (which is far from perfect right now with just one user), pattern recognition, the lot would be lost to use.
Imagine trying to figure out a way to explain what is a "schmorgaborst" or a "rudder". find a way around words with your hands in order to get understood by someone who does not speak your tongue. now you've got an insight of the level of difficulty to get understood by someone. And NOW, think of a way of doing that WITHOUT YOUR HANDS, and without using words you would need to define later in order to be understood. Are you finding this difficult yet? Wait till I ask you to do this in the language of the person you are standing in front. I am French. Try to EXPLAIN in French what a "barter" is without resorting to translation nor to gesture, since the word would probably be alien to me in french. Getting a grasp on it yet? OK! here comes the best part. Now do it in C#. If this one does NOT suit your taste, (and I can understand why) then do it in HTML, in Basic, in Java, in anything as long as your machine does not only stocks the information but also recognizes it as not being the primary input. Tell it what is a Hot-Dog, what is a seller, and show it a picture of Wall Street at lunch time. Can it recognize and point to a Hot-Dog seller? If it can, you got me there. Try showing it different pictures of different races of dogs always telling it they are dogs in the first place, and giving it their races names as subclasses. Try to have it tell you a definition of dog. Try to have it correlate that answer with our previous Hot-Dog without being utterly wrong. Once again, if you can do that, the rest is probably a walk in the park...
Maybe not at night, though...

#129 GMScribe   Members   -  Reputation: 122

Like
0Likes
Like

Posted 07 March 2005 - 09:10 AM

That's why i think the fact we ARE the programmers of AI helps us in that we can make a different, suitable interface to compensate?

#130 Nice Coder   Members   -  Reputation: 366

Like
0Likes
Like

Posted 07 March 2005 - 11:16 AM

What if anything, is stopping us from making 'axomic' consepts which the ai understands? For eg. x IS y, x is NEAR y, x LIKES y, x DISLIKES y, x ISMEMBER y, x HASPROPERTY y, X ISVALUABLE ....

From,
Nice coder

#131 Madster   Members   -  Reputation: 242

Like
0Likes
Like

Posted 10 March 2005 - 12:12 PM

Quote:
What if anything, is stopping us from making 'axomic' consepts which the ai understands? For eg. x IS y, x is NEAR y, x LIKES y, x DISLIKES y, x ISMEMBER y, x HASPROPERTY y, X ISVALUABLE ....

the fact that you can come up with as many elements that fit those relationships as you can come up with relationships themselves, thus (with general AI in mind) you'll always run short on relationships.

Btw someone mentioned fussy logic, and that just cracks me up XD
(fussy : Easily upset; given to bouts of ill temper)
No one likes a fussy AI =)

#132 Nice Coder   Members   -  Reputation: 366

Like
0Likes
Like

Posted 10 March 2005 - 06:01 PM

I was thinking of a handful of atomic relationships, from which we can build a full bredth of relationships.

From,
Nice coder




Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS