BBC News has an extremely thought provoquing article entitled Can computers be creative?. The article discusses creating art with computers as a tool, which is nothing new (see some of Perlin's or Musgrave's work). It doesn't even class as AI, despite the randomness of some parameters, as the artists has overall control of the process and judges the quality of the final image (deciding when to stop working on it).

However, the debate can go further, thereby falling into our beloved field of artificial intelligence.

Contrarily to general opinion, computers are very creative. A discussion I started over on flipCode a while back (Randomness In Games) argues that even simple random number generators are sufficient as long as they are seeded with time -- thereby no longer being predictable to the average human.

Being able to create Art with this creativity is another matter. Since computers are good tools for assiting the creation of Art, this could be easily taken a step further to allow them to create Art on their own. Then it's just a matter of judging it... And that's where the skill comes in: being able to judge an image as a piece of art. Humans do that with complex emotions based on sensations and memories provided by the visual patterns. Computers cannot yet do that, yet! They can't even understand what's in the picture, let alone feel something about it. And even if they did, chances are humans wouldn't appriciate it in the same fashion.

However, I can see computers being able to produce main-stream Art; Art created purpously to please the masses. That's just a matter of detecting patterns in existing Art, duplicating and combining it with slight changes. Music is a good example for this: I can forsee computers being able to produce 'music' based on trends and fashion (granted that main-stream pop stuff hardly classes as music, but still, the concepts could be extended).

Anyway, I'll stop here. As I said, a very thought-provoquing article.

Alex,

I'm glad that you qualified your description of the relationship between randomness and creativity by recognising the important that judgement plays in the creative process. The ability to produce endless variety does not necessarily imply creativity on the part of a computer program, although the same program may be used by skilled artists to produce creative works.

The ability to judge the potential for creativity of one's own work is of utmost importance in the cerative process as this is the driving force that motivates the production of new works.

I have recently completed a PhD developing computational models of creativity. My research lead me to the conclusion that the most important element missing from models of creative processes, such as Cohen's AARON, is the ability to judge the potential creativity of a work in progress. I went on to discover that the aesthetic affects of novelty in a work are quite subtle and the search for the right type of novelty (curiosity) plays an important role in driving the creative process of individuals and groups. The end result was that I was able to create autonomous groups of curious design agents interacting and sharing artefacts (genetic artworks) without any external input. I've called this approach to studying creativity "artificial creativity" because of its similar methodology to "artificial life". If you'd like to know more about my research, visit my academic webpages at http://www.robsaunders.net

-RoB-

I completely agree. I just haven't really read much about this topic, hence the possible reason for diverging terminology.

I call creativity the ability to produce different and varied solutions, which for computers is just a matter of 1) modelling a representation and 2) randomness. Admittedly, this is unconcious creativity: the computer has no idea what the solution means or implies. Humans on the other hand, are creative in an aware fashion, prompted by heavy emotional baggage. This bias can also limit their 'creativity', but enhance their artistic potential.

Art would then be a consequence of creativity, namely a solution that appeals to the agent/computer in such a way that prompted its creation. As you mention, creativity... (I'd never thought of that factor ;)

It seems we agree deep down even though the wording is not identical.

Your work sounds fascinating... multi-agent systems applied to pretty much anything are very interesting to study: I've just discovered an emergent way of mapping huge complex terrain with multiple very simple reactive agents. Applying them to art creation sounds even better!

What department were you at in Edinburgh? I'll get your thesis out of the local library ;)

Terminology is a very tricky thing to agree upon when dealing with slippery concepts such as intelligence, creativity and consciousness. So it's just as well we have this Philosophy Institute to sort them out... :-)

The problem with defining creativity as the production of different and varied solutions is that it doesn't really capture what is generally meant when a person is described as creative. Admittedly, creativity is an emotive term and definitions that are commonly agreed upon are hard to come by.

A solution is called creative when it is both novel and appropriate. Novelty requires that the solution be unexpected to the judge of creativity and this depends upon their previous experiences of similar situations. This explains why two people judging the same solution can come to different conclusions as to its creativity. Similarly it explains why the creativity of solutions can be lost on the individual that produces it: to them it was the most obvious solution and was not unexpected.

A person is judged as being creative when they produce creative solutions. A person does not have to produce different and varied solutions to a problem to be considered creative. One solution will suffice as long as it novel and appropriate. Producing many varied solutions is a good method for finding a single creative solution but it is not the only way. Experiencing diverse experiences so that they can be applied to as yet unseen problems can be just as important.

It is not necessary to appeal to "emotional baggage" to explain the possible limitations of conscious creativity. Emotional responses to creative works are tied with the appropriateness of the work, especially works of art, but does not necessarily limit the creative process. What limits the creative process is that solutions are unexpected from previous experience and hence not considered. This argument can be applied to a completely rational learning system as much as an emotional human. A completely rational system that has learned that certain approaches are unlikely to yield an appropriate solution for a range of problems is unlikely to try those same approaches when faced with an apparently similar problem.

I did my undergraduate in Computer Science and Artificial Intelligence at Edinburgh University but I conducted my PhD research at The University of Sydney so you won't find my thesis in the library at Edinburgh. I will make my thesis available from my website when the review process is complete -- hopefully early in the new year.

Reading your answer I can see at least two different kinds of creativity: There's the "problem solving" creativity that solves something in a novel and appropriate way, and there's the "artistic" creativity, which doesn't really solves a physical problem but rather produces something (like a novel or painting) out of nothing that is novel and appealing in an aesthetic sense.

Does this sound like a valid distinction, or are the two types actually specializations of a more general concept? If it is valid, what is the possible benefit in recognizing this distinction?

Jeroen

Jeroen,

You are quite right to distinguish between what appear to be two different types of creativity, however, it can be argued that the aesthetic appeal of artistic works can be described as the appropriateness of those works. To have any value as an artwork it must appeal to its audience.

A certain amount of aesthetic appeal is necessary for new artworks to be considered good but to be considered creative these works must go beyond that which has gone before and confound expectations with some novelty. Hence, I think we can still describe artistic creativity as the production of novel and appropriate works that satisfy a need.

Conversely, creative problem solving often appeals to certain notions of aesthetic as well. There are many appropriate ways to solve ill-define problems but creative designs are often aesthetically pleasing as well, e.g. architecture shows how many ways there are to satisfy the basic human need for shelter but few are considered truly creative, often because they solve problems "elegantly".

Even in science, supposedly an entirely objective pursuit, we can find plenty of examples of aesthetics playing important roles in the production and (more importantly) the acceptance of new theories.

-RoB-

An indirect follow-up to the original post: an article on Wired entitled Corner of Art Place and Tech Lane, with a cool quote - if nothing else:

"When art and technology come together, magical things happen."

I've been meaning to think about the topic a bit more (and post a reply to your post), but I've been quite busy!

That's interesting. I've also been toying with notions of novelty and curiosity with my stereo vision robot Rodney. For the robot's purposes I've defined "novelty" as being the difference between what the machine expects to see and what it actually sees. For example, if the robot turns its head to look in a northward direction statistically it expects to see a TV and bookshelf, but if it sees me putting a book onto the shelf then that is a novel (statistically rare) situation and the robot pays closer attention and engages various learning algorithms.

To devise a system which can engage with the world in an interactive and possibly creative way I think you would need to try to maximise some function of novelty over time, such that the system becomes a sort of continuous thrill-seeker. Maybe play behavior could be defined as this sort of novelty-seeking activity.

- Bob

Bob,

(My apologies for my recent disappearance from this site -- I was unavoidably detained in the Great Barrier Reef for a conference!)

You might want to take a look at the work of Mark Peters at the University of New South Wales, here in Sydney. He has recently completed his PhD thesis on the role of novelty in vision-based robotics and I found his papers and thesis an interesting read.

http://www.cse.unsw.edu.au/~markpeters/

-RoB-