 |
|
|
 |
|
|
|
|
Charles Wright |
| | | | | | posted 2/9/2002 19:24 |   | | | | | | | | | I have read in several different posts ideas about AI and creativity. Many people seem to think that an AI must in some way exhibit creativity in order to be considered concious or alive. The first definition of CREATIVE in Websters is "1 : marked by the ability or power to create : given to creating <the creative impulse>". (Don't you love those circular references.) the second definition is much more interesting. It is "2 : having the quality of something created rather than imitated : IMAGINATIVE <the creative arts>". This is what I think people mean when they want an AI to be creative. They want more than just imitation. The modern concept is "thinking outside the box". (Kind of hard for a computer, huh? ;) ) But my notion is this: that once we have a computer with a close to human-like intelligence, it would be next to impossible to think that it could think EXACTLY like a human. Therefore it is likely to put concepts together in ways that a human would not. Therefore, these early AIs could be considered MORE creative than humans. Of course we would be uncomfortable interacting with these AIs, because their thoughts would seem to jump around and we would have trouble following them because they were too creative. Therefore our follow up would be to limit their creativity so that the followed more closely to our ideal.
| | |
| | |
pa |
| | | | | | posted 2/10/2002 17:36 | | | | | | | | | | | |
Charles Wright wrote @ 2/9/2002 7:24:00 PM:
1) ...once we have a computer with a close to human-like intelligence, it would be next to impossible to think that it could think EXACTLY like a human.
2) Therefore it is likely to put concepts together in ways that a human would not.
3) Therefore, these early AIs could be considered MORE creative than humans.
4) Of course we would be uncomfortable interacting with these AIs, because their thoughts would seem to jump around and we would have trouble following them because they were too creative.
5) Therefore our follow up would be to limit their creativity so that the followed more closely to our ideal.
| |
Hi Charles, I faced the same things !
Due to poor in English, can u rewrite 2) & 5) in different words ? make it easier for me to understand.
1) I faced the problem before. If we are not carefully 'craft' the AI, finally, I have to follow God, to gave the 10 Commands ! I was tempeted to made such AI behaves like human in order to solve the problem ! Finally, I realised that AI is digital based, so work with it, not against it, management concept ! the final result is similar or same even the way it works differ from human !
2) I guess I MIS-understood this sentence, but try ! Do u mean AI researcers shall distinguis "the way/method of HUMAN intelligence work" from "the one of AI", despite the final result should satisfy human ?!
3) My final opinion is DEPENT design of AI ! If u think that AI utilize traditional Searchs (tree, heuristic, statistic/stocastic, etc) to create, so finally AI will be more creative than human, Yes ! the core issue is processing power ! this problem can be overcome !
4) It ONLY happen on EARLY stage of development ! Comparable to human like me. Because English is a foreign language for me, so it is NOT comfortable to communicate w/ me for native ones, unless after several years live in English spoken country !
5) "...to limit their creativity..." You mean we shall limit the creativity part of AI so it will perform closer to human style ?! If it is, the AI will SUFFER for UNDER-UTILIZE becaused it must be 'hamble' it self as LOW as the teacher ! It is TERRIBLE. The 1st 'manual solution' is to put 1000 teachers from diverse diciplines to teach. It is what I am doing now, it allows thousands of teachers using diverse languages to answer questions, PARALEL ! it is not practical, but it is very powerfull ! 2nd one is plug it into multi-media, so it can learn from education on TV, or e-Learning, etc ! I like the idea from John Travolta movie "The Battle Field on Earth ???" Learn by seeing from 'highspeed lights???'.
Rest of 5) can't understand, pls re-explain...
I expect your reply ! Thank's !
| | |
| | |
yaki |
| | | | | | posted 2/10/2002 17:59 | | | | | | | | | | | |
Charles Wright wrote @ 2/9/2002 7:24:00 PM:
2 : having the quality of something created rather than imitated : IMAGINATIVE <the creative arts>". This is what I think people mean when they want an AI to be creative.
| |
Good point, Charles. But think about this: Creativity (like intelligence and beauty) are in the eye of the beholder. Creativity is always ATTRIBUTED to someone (something??) by an observer; - it is a judgement call. More than anything, it measures the SURPRISE the observer experiences when he observes the action in question.
It is possible, of course, to try to "objectify" this attribute (view it as an objective feature of the action in question). If one insists on that approach (and I don't see why), he could define creativity as "the ability to produce (or the production of) a new pattern, not previously observed in the input."
| | |
| | |
Charles Wright |
| | | | | | posted 2/10/2002 21:04 | | | | | | | | | | PA –
2) An AI would connect thoughts differently than a human. Therefore it’s thought process would lead to a different conclusion that an average human.
5) Because of point #2, it would be difficult for humans and the AI to communicate. We would want the AI to draw the same conclusions as us, not different ones. So we would change the AI to make it more like us.
yaki –
Yes, I agree with you. I like the work Douglas Hofstadter has done in this area.
| | |
| | |
Dr. Simon Ronald
[Guest] |
| | | | | | posted 2/11/2002 10:35 | | | | | | | | | | yaki writes
"It is possible, of course, to try to "objectify" this attribute (view it as an objective feature of the action in question). If one insists on that approach (and I don't see why), he could define creativity as "the ability to produce (or the production of) a new pattern, not previously observed in the input."
Yes, Yaki you are right :- in fact Bart Kosko provides a mathematical definition of creativity in his book on control systems, neural nets and genetic algorithms.
| | |
| | |
pa |
| | | | | | posted 2/11/2002 18:24 | | | | | | | | | | Hi Charles !
You are right, that is the problem I met, can't comment now, need a long run testing !
What is the theory of Douglas Hofstadter & Bart Kosko on creativity ? FYI, I learn AI not from class so 'blind' about theories !
Yaki !
the "new pattern" of the definition SCAREs me ! It is the part of problem of 2) & 5) !
can elaborate the world 'pattern' ? I worry mis-understanding your definition !
| | |
| | |
yaki |
| | | | | | posted 2/12/2002 06:38 | | | | | | | | | | Charles - can you provide a more specific reference to Hofstadter on creativity?
Simon - Kosko is cool, but too "fuzzy" for me.
pa - I must confess I don't understand your question.
| | |
| | |
pa
[Guest] |
| | | | | | posted 2/12/2002 20:53 | | | | | | | | | | Hi Yaki !
I think I issue a wrong question. This is the reason I scare when cross to "producre new pattern".
When I saw your definition on creativity as an ability to produce "new pattern", I recall 2 things:
1. Assume ! that creativity utilize Searchs, so "new pattern" means new pattern of searching ! In fact, only limited searching pattern can be produced from database, so it make me think it is impossible to produce new pattern.
2. When we go to human level, "new pattern" means the ability to react in new different ways that never done before ! For example, response toward happiness ! Ussualy, we enjoy it immediately, but now, delay for awhile, or forget it ! The ability to generate a new reasonable pattern is also limited, due to many factors !
My point is, either on database level, or human level, the PROBABILTY to produce NEW PATTERN is limited ! Only by discovering new idea on physic, biology, algorithm, etc, we are enabled to produce new pattern !
3. The 3rd, it crossed my mind that Alan may utilize a kind of pattern matching to produce answer from existing conversation log ? IF I'm right, so Alan is the 3rd case !
So back to my question, what KIND of pattern as mentioned in the definition ? 1st, 2nd, 3rd, or just a general definition w/o relation to AI ?
| | |
| | |
yaki |
| | | | | | posted 2/13/2002 10:00 | | | | | | | | | | My definition of Creativity is not restricted to any particular domain. It can be applied to art (as in "creative artists"), to business ("creative management"), to science (new research approach, or creative proof of a theorem) etc. etc.
As for the realm of "Intelligence" (artificial or not) - since I view intelligence as the ability to engage in lingual communication (verbal, written or any other), "to be creative" in this context means to produce language patterns which are NEW (do not constitute mere imitation of previously observed input).
As for Alan: Subjectively (from the point of view of a particular user), Alan is creative when he APPEARS creative - when he "surprises" the user. Objectively, Alan does have creative capabilities, because he produces complex responses using variables and concatenating sentence parts in ways not intended (or forseen) by his authors.
| | |
| | |
Lior Fainshil |
| | | | | | posted 2/13/2002 13:24 | | | | | | | | | | Ok. Let me try to explain how I see creativity.
Creativity of a computer program is how good it is at finding unexpected solutions to problems. There are 2 kew words here:
1)Solutions. It is important that the creative solutions actually solve something. Otherwise the most creative entity in the universe would be a random number generator.
2)Unexpected. Unexpected by whom? Here is where the subjective definition lies. Chess grandmaster that plays against a good chess program can say that it is very creative, but the programmer behind the software knows better. The program just checks millions of possible moves, and there is nothing creative about it.
So let me propose a slightly more objective definition of creativity. I see the creativity of a program as seen by the person who wrote the program. A creative program is one that can arrive to solutions, which its programmer, or another person who understands how the program works, finds creative.
I know that this definition is a bit problematic too. I tried writing here a more exact definition, but I got into trouble.
In any case, I think that creativity is something very important for computer programs. When I described my player for the LMW site, I wrote that it is not creative. I am however far more interested in creative programs and my original intention for future development of my player was to add to it creativity.
| | |
| | |
bill
[Guest] |
| | | | | | posted 2/13/2002 16:47 | | | | | | | | | | | |
Lior Fainshil wrote @ 2/13/2002 1:24:00 PM:
... Chess grandmaster that plays against a good chess program can say that it is very creative, but the programmer behind the software knows better. The program just checks millions of possible moves, and there is nothing creative about it.
| |
I have to disagree with you here Lior. You and say the programmers of Deep Blue, might well know intimately how the program works, you might think that it is 'merely' a brute force mini-max search, but you dont know nor can you predict the programs output, this was pretty much proved by the program beating the best human chess player.
When Deep Blue beat Kasparov it had to be creative, how else could it have won, if it was some how hardwired for every chess position, who hard wired it? were its programming team secretly harbouring a superior human player than Kasparov?
Incidently I think you are wrong about your own entry in the LMC, obviously you know how it works, but it can still display some degree of creativety, along the lines of the Deep Blue example. Infact since it is a general learner, it could probably display MORE creativety than Deep Blue in certain circumstances.
| | |
| | |
tony eadington
[Guest] |
| | | | | | posted 2/13/2002 17:26 | | | | | | | | | | | |
Lior Fainshil wrote @ 2/13/2002 1:24:00 PM:
...Otherwise the most creative entity in the universe would be a random number generator.
| |
Lior,
Don't underestimate 'random'.Imagine an LMC type contest that was to navigate out of an unknown maze, without leaving a trail or marking junctions. For just about every algorithm there will be a maze that will thwart you, but this algorithm, i think, will eventualy escape from any maze, the algorithm is: At each junction pick a random direction.
AWE.
| | |
| | |
Lior Fainshil |
| | | | | | posted 2/14/2002 00:54 | | | | | | | | | | In a program that isn't creative, the programmer can understand exactly how the program arrived to a particular solution. A creative program arrives to solutions, that are both unexpected by people and the author can't understand the exact of way of how the program manged to obtain the solution. Although he understands the principle of how the program works, he can't mentally retrace the route taken by the program to arrive to a particular solution.
This definitition is also not perfect, but it is definitely much more objective and seems to pretty much show what I mean by a creative program.
| | | Last edited by Lior Fainshil @ 2/14/2002 12:55:00 AM | | |
| | |
Charles Wright |
| | | | | | posted 2/14/2002 02:09 | | | | | | | | | | Yaki -
Douglas R. Hofstadter -
Fluid Concepts and Creative Analogies: Computer Models of the Fundamental Mechanisms of Thought - ISBN: 0465051545
Publisher: Basic Books
Pub. Date: February 1994
An example of computer creativity from the book - The computer program 'Letter Spirit'. This program takes the symbols from an alphabet and modifies them graphically to create new fonts. The characters in the new font are recognizable as the letters they represent, and at the same time when looking at them, we humans have a sense that they somehow belong together in the same font family, but they are all different from the original font. There is also another computer program exibiting the idea of 'conceptual slippage' in language.
| | |
| | |
Charles Wright |
| | | | | | posted 2/14/2002 02:12 | | | | | | | | | | | |
Lior Fainshil wrote @ 2/14/2002 12:54:00 AM:
In a program that isn't creative, the programmer can understand exactly how the program arrived to a particular solution. A creative program arrives to solutions, that are both unexpected by people and the author can't understand the exact of way of how the program manged to obtain the solution. Although he understands the principle of how the program works, he can't mentally retrace the route taken by the program to arrive to a particular solution.
| |
Lior -
Your definition reminds me of the argument for free will vs. determinism.
| | |
| | |
yaki |
| | | | | | posted 2/14/2002 07:14 | | | | | | | | | | Thanks, Charles.
IMHO, the archetypical example for creativity is the capacity to generate random numbers. Random number generation programs aren't really "creative", in the sense that the person who wrote the program knows precisely what is going on. In this strict sense, God (or 'Nature') is the only creative entity (through its agents: radioactive substances, global weather and such...)
| | |
| | |
lodsw |
| | | | | | posted 2/14/2002 10:22 | | | | | | | | | | I don't really follow this. A program is "creative" if you don't understand how it works?
If someone explained how the program produced it's output would it stop being creative?
Does a description of a poet's brain states while composing (which explains how they produced the poem) eliminate creativity?
Hmm actually this does remind me of how when you get a new technology in AI everyone raves about it but when it is well-understood no one thinks it is intelligence (eg chess-playing AIs).
I think "creativity" is a quality human's find when they seek but which doesn't necessarily exist outside human minds. We want to believe that there is something mystical about creativity/intelligence which a machine cannot reproduce.
With intelligence you can at least offer the more general (perhaps a bit alien) definition "survival".
| | |
| | |
Lior Fainshil |
| | | | | | posted 2/14/2002 11:17 | | | | | | | | | | | |
lodsw wrote @ 2/14/2002 10:22:00 AM:
If someone explained how the program produced it's output would it stop being creative?
| |
You already know everything there is to know about the program. No one can explain it better.
| | |
| | |
yaki |
| | | | | | posted 2/14/2002 15:31 | | | | | | | | | | | |
lodsw wrote @ 2/14/2002 10:22:00 AM:
I think "creativity" is a quality human's find when they seek but which doesn't necessarily exist outside human minds.
| |
I concur wholeheartedly. This is what I meant by the "subjective" meaning of Creativity.
| | |
| | |
yaki |
| | |
|
|
|
|
