Can Computers Think Creatively?

Asimo - a humanoid robot created by Honda

Photo by AZAdam

Computers are very smart, but they have no imagination, right?

They can only slavishly follow instructions and rules – which means they can’t think for themselves and generate original thoughts.

This is why a computer can beat a Grand Master at chess, but it will never compose music to rival Mozart.

It’s why Lou‘s job is under threat from automation, but Jack has a bright future, safe in the knowledge that a computer will never replicate his ability to think creatively.

It’s why Asimov was a creator, but Asimo is a mere curiosity.

Or is it?

Consider the findings of an experiment reported in the New York Times, in which humans were pitted against computers to see who could come up with the best ideas for advertisements.

The humans were non-advertising professionals, given a brief and asked to come up with creative ideas for adverts. The computers were programmed with an algorithm for devising advertising ideas and given the same briefs.

Here’s a sample of the results:

  • COMPUTER IDEA: An Apple computer offers flowers (for advertising Apple Computers’ friendliness).
  • HUMAN IDEA: An Apple computer placed next to a PC with the claim: “This is the friendliest computer.”

  • COMPUTER IDEA: Two Jeeps communicating in sign language (for advertising a silent car engine).
  • HUMAN IDEA: A car driving alone in the country.

  • COMPUTER IDEA: A domed mosque with tennis ball texture (for World Cup Tennis tournament in Jerusalem).
  • HUMAN IDEA: A picture of ancient walls of Jerusalem with a tennis poster on them.

(‘Route to Creativity: Following Bliss or Dots?’ by Natalie Angier)

I think most of us would call that 3 – 0 to the computers.

The research panel agreed – they judged the computer ads to be consistently more original and creative than those devised by the human group.

What is going on here?

Does this mean the beginning of the end for human creative superiority? Not necessarily.

The researchers were Dr. Jacob Goldenberg, Dr. David Mazursky and Dr. Sorin Solomon of Hebrew University in Jerusalem. They weren’t actually trying to find out whether computers or humans are more creative – but to mimic the thought patterns of effective human creators. Note that the human group were untrained, with no previous experience of creating adverts. The computers, on the other hand, were programmed using formulas derived from successful adverts.

So the computers had an unfair advantage. It’s as if two groups of people were pitted against each other in a game of chess: the first group composed of people who had only ever seen chess matches played on television; the second group given a thorough grounding in the rules of chess.

No prizes for guessing who would win that one.

In fact, when the reseachers repeated the experiment and taught the formulas to the human group, they were able to beat the computers.

So what was in the magic formulas?

Creativity by Numbers?

The researchers were trying to disprove the popular idea that “the most original ideas are born of utter freedom, a shifting of paradigms, a circling of the square, a streaming of consciousness, a squelching of the internal editor”. To do this, they programmed their computers using a series of ‘thought templates’ to limit options and generate ideas according to specific rules.

One of the commonest templates they found is the so-called replacement template.

For example, they considered a Nike ad, in which a group of firemen are standing around in a rescue pose, looking up as though someone was about to jump from a burning building into their net.

In lieu of a net is a giant Nike sneaker, with copy boasting of how the new Nike walking shoes are “very safe places to land.”

In this advertisement, the sneaker replaces an object whose most salient characteristic is “cushioning.” Indeed, the life net cushions a person from death itself.

(‘Route to Creativity: Following Bliss or Dots?’ by Natalie Angier)

Rules, constraints and formulas. It’s hardly the stuff of Romantic imagination, or even the popular idea of thinking outside the box. But as we’ve seen before on Lateral Action, using constraints and thinking inside the box can be surprisingly liberating.

If you’re tempted to dismiss advertising as a fairly menial form of creativity, far removed from the lofty realms of the fine arts, you may be interested to hear of the work of composer and music professor David Cope. His EMI (Experiments in Musical Intelligence) software can create original music in the style of over a hundred different composers, which has even been mistaken for and original Bach. Have a listen to an interview with WNYC (at the foot of this page, or download here) – featuring some breathtaking excerpts of the computer-music – and judge for yourself. You can find more of Cope’s EMI’s compositions here. Personally I find some of them disturbingly good.

I’m no musical expert, but as a poet and poetry editor I pride myself on my literary judgment. So I was intrigued to come across The Policeman’s Beard is Half-Constructed, a collection of poetry and prose written by a computer program called Racter. Some of it’s not bad:

Slowly I dream of flying. I observe turnpikes and streets
studded with bushes. Coldly my soaring widens my awareness.
To guide myself I determinedly start to kill my pleasure during
the time that hours and milliseconds pass away. Aid me in this
and soaring is formidable, do not and winging is unhinged.

I like to think I wouldn’t be fooled by this kind of thing if it landed in my editor’s in-tray. On the other hand, I’ve seen plenty of worse poetry produced by humans.

Critical Thinking – the Critical Difference?

These examples are impressive – but are the computers really being creative?

Maybe the answer lies in the question. As far as I know, no computer would stop to think whether it was really being creative. The question wouldn’t enter its central processing unit. It’s very human to ask such a question – to desire authenticity and then evaluate the work to see whether it meets the criterion.

“To suspend criticism and think any idea is possible or good may ultimately be destructive to creativity,” said Dr. Goldenberg. As we saw last week, critical thinking is central to the creative process. And according to Goldenberg’s fellow researcher Dr Mazursky, “Humans can criticize themselves, and computers can’t”.

The advertising algorithms show it’s possible to program a computer to generate whacky ideas – but it took a human panel to judge their effectiveness.

A machine can write poetry or music – but only humans can decide whether the finished work is any good.

It sounds counterintuitive, but perhaps the human creative advantage comes not from our ability to generate unusual ideas, but to use our critical faculty to evaluate them.

I’m not saying computers will never achieve critical discernment on a level with human beings – but training them to do it will be a lot harder than teaching them a few lateral thinking techniques.

What Do You Think?

Do you think computers will ever truly think creatively?

Do you agree that critical thinking is what gives us a creative edge over our digital cousins?

How would you feel if you were moved by a story or a piece of music – only to discover it had been written by a robot?

About the Author: Mark McGuinness is a Coach for Artists, Creatives and Entrepreneurs. For a free 25-week guide to success as a creative professional, sign up for Mark’s course The Creative Pathfinder.

How to get creative work done in an "always on" world

Productivity for Creative People

Mark McGuinness' latest book Productivity for Creative People is a is a collection of insights, tips, and techniques to help you carve out time for your most important work – amid the demands and distractions of 21st century life.

β€œOf all the writers I know, I have learned the most about how to be a productive creative person from Mark. His tips are always realistic, accessible, and sticky. It’s not just talk, this is productivity advice that will change your life.”

Jocelyn Glei, author and Founding Editor, 99U

More about Productivity for Creative People. >>

Responses to this Post

Comments

  1. Sarah Cheverton says:

    Wouldn’t bother me to be moved by a piece of art, music or poetry generated by a computer – like you I enjoyed the excerpt you posted. Computers, after all, and computer programming does not self-generate, they are designed by people and the way they are designed inherently affects the way they work. Whether it’s in the inception or the implementation, there’s always a human in the machine somewhere…

    Besides, as a lifelong Star Trek fan, I’ve been flirting with the idea of marrying an android for years. I’ve got a crush on Data an intergalatic stellar system wide.

    Er, that’s not weird, is it?

  2. When the nodes and branches of a computer’s thinking are on the same level of complexity as a living organism (“bushy” instead of hierarchical) then we’ll have emergent (whole is greater than sum of parts) phenomena.

    That’s when computers will have personality and that’s when they’ll be truly creative.

  3. Well, what we silly humans fail to realize is that we slavishly follow instructions and rules as well–it’s just that when it comes to our own minds and bodies, we didn’t write the programming, so we’re not really sure how all the instructions and rules go together.

    But yeah, before I go into an over-simplified discussion about free will…what’s most interesting about this is the fact that the algorithm is based on conclusions of professional creative marketing minds.

    This kinda gears the man vs. machine idea as a false dichotomy since we’re the ones “teaching” the machines in the first place. The hope is that our organic creativity will eventually merge perfectly with our technology through open digital collaboration.

    Lets get the point where, say, a group of 100 or 500 minds seamlessly works together on a creative marketing campaign with the calculation component serving as an arbiter between all those thoughts and ideas.

    When we can do that, we’ll REALLY be cooking.

  4. Critical thinking and creativity both require practice. We can learn skills, like playing chess, but we also need to keep them sharp. Computers don’t lose their chops like we do.

    There’s something else, too: intuition. When programmers can translate intuition into ones and zeroes, computers might have a creative edge. Until then, all the analysis in the world can’t replace a gut feeling.

  5. @ Sarah – It’s OK, your secret’s safe with us. πŸ˜‰

    @ Michael – So personality is a prerequisite for creativity? I could go with you on that.

    @ Joe – “The hope is that our organic creativity will eventually merge perfectly with our technology through open digital collaboration.” I guess we’re starting to get there in some fields, if you cut the ‘perfectly’.

    @ Stacey – “When programmers can translate intuition into ones and zeroes, computers might have a creative edge.” I’m sure someone, somewhere, is working on it…

  6. Concerning Racter and its purportedly all-computer written novel “The Policeman’s Beard Is Half-Constructed”, I’m afraid you’ve fallen for a hoax – of a kind.
    There’s no way the commercially released version of Racter (which I licensed at the time, as I did the Inrac programming language it was based on) could have generated that book.
    So the developers were either deploying a much more powerful, non-released and never disclosed version (in which case their claims were blatantly untrue), or the “novel’s” text was manipulated (in which case: ditto).

  7. @ Fantomaster – Yes I did read that the commercial release of Racter couldn’t have generated the text. The piece I read suggested that the text must have been generated by a purported ‘full version’ of the program, which hadn’t been released. It didn’t go so far as to suggest that the text had been manipulated by a human editor.

    If that is the case, then it’s still an interesting experiment, but as you say, not as radical as the introduction claims. More like an automated version of Burrough’s cut up technique – i.e. a stimulus to human creativity rather than a replacement of it.

    Would you recommend the Racter program, even in this limited capacity?

  8. Seeing that it’s available for free online (as are some of its clones), I actually would recommend having a bash at it: it can be quite entertaining for appr. 20-30mins after which it gets pretty repetitive, just like most all chatterbots. The best thing about it IMV is that it actually IS quite literary and ambitious on that score, i.e. not the usual techno or psycho drivel other bots will hit you with.

    E.g. I always fondly remembered one of its iterations: “As Nietzsche said, the soul is but an amusement of the body.” (Bogus quote, of course, but rather witty.)

    Yes, I agree it’s a fascinating concept and the INRAC language is actually quite powerful, holding a lot more potential than Racter (in its commercial version) would let you believe on first sight.

    Of course, this was all based on DOS and 80s technology when memory and harddisk space was still a huge issue. It’s a pity that they resorted to that marketing gimmick (or, arguably, trickery) by making claims that were so blatantly untrue when it was actually quite a sophisticated, promising venture in its own right.

  9. Thanks, think I’ll check it out. Could be a fun way to get started writing a poem, even if I have to roll my sleeves up and work on it myself at some point. πŸ™‚

    That bogus Nietsche quote is great, just Tweeted it (with credit of course). http://twitter.com/markmcguinness/status/4634259716

  10. Hehe, nice – just tweeted about your “Why Critical Thinking Is Not a Creativity Killer” myself.

    You’ll probably get better results (and easier to achieve) with text spinning tools such as Power Article Rewriter – no implemented English language grammatical syntax or anything but more versatile in terms of structuring things etc.

    (I know this from experience: I’ve actually generated a few thousand “poems” that way once as text filler content for an SEO support blog a while back, heh…)

  11. I think the critical piece to the discussion is something you touched on in one of your questions, Mark: emotions.

    To be honest, I’m not surprised that a computer can come up with more “wacky” ideas than a human, and that we may judge them to be more creative than those that the humans came up with, but the process was entirely algorithmic.

    Call me a romantic, but the best art is usually infused with some sort of passion or emotion, and it’s that connection with the artist on an emotional level that makes it great. I might hear a song written by a computer and think it’s catchy and snap my fingers to the beat, but I doubt it’ll ever move me emotionally.

    It’s creativity by numbers, and that will get you to a certain point, but it will never get you past the threshold of exceptional.

    My 2 cents (CAD).

  12. Thanks Goodness they are still machines now and can’t think creatively otherwise they dominate us someday (just like some movies show this)

  13. First of all , I would argue that mimicking the style of a composer is not the same as creating a style of composition.

    The fundamental question for me is one of nature vs. nurture. Are we simply the sum of our experiences? Is so, a complex computer should be able to replicate that. If, on the other hand, there is something inherently individual in our DNA, that will be more difficult to emulate. In the end, unless you believe that we have a soul that imbues us with certain personality traits, the human body is essential a machine, albeit a highly complex one.

    As for enjoying art, music or literature from a computer, it doesn’t matter to me. Then again, I’m not that discriminating with my human-generated content either. If I’m entertained, I’m entertained.

  14. The answer to the question “can computers think creatively” is not really addressed here. Without a better definition of creativity, we are only discussing parlor games. You might as well ask “can computers poop?” and invoke Vaucanson’s defecating duck automaton.( http://dugnorth.com/blog/2008/02/vaucansons-defecating-duck-automaton.html)

    The real interest, for me, lurks in the essay in the recognition that computers cannot critique their own output.

    Gizmos for “creating” everything from philosophical propositions (Raymond Lull) to classical music (Mozart) have been around for centuries.

    The ability to recognize, or create, value in context, is uniquely human. The pivotal concept is not creativity as permutation, which can be at many different levels of order (See Pierce on art and information theory). It is appreciation. To appreciate means to find value in, but it also means to add value. The mechanisms of permutation cannot appreciate anything. They can make us think, and provoke us to new creation, but so can a stain on a garden wall…as Leonardo Da Vinci noted long before computers.

    Thanks for an interesting and provocative essay.

  15. @ Fantomaster – That Power Article Rewriter thing may have incidental value as a tool for literary experiments, but it’s primary application is pretty depressing.

    @ Adam – Yes! It’s not just about how ‘good’ something is, but whether it matters.

    @ Phaoloo – To judge by the amount of time some of us spend hooked up to computers, that day may already have arrived. πŸ˜‰

    @ Christy –

    The fundamental question for me is one of nature vs. nurture. Are we simply the sum of our experiences? Is so, a complex computer should be able to replicate that.

    “this is an art
    Which does mend nature, change it rather, but
    The art itself is nature.”

    Shakespeare, The Winter’s Tale

    @ Zeitguy – This is probably the closest I’ve come to a definition of creativity on this site. I agree that permutation is nothing without appreciation – which is why I regard critical thinking as part of the creative process.

    Thanks for the pooping duck – great to know the science of robotics had such an auspicious start! πŸ™‚

  16. Power Article Rewriter: sure it’s a “content spam” tool, albeit a fairly sophisticated one, essentially used for search engine optimization.
    Or, as Lawrence Durrell wrote so nicely: “When a civilisation has decided to bury its head in the sand what can we do but tickle its arse with a feather?”

  17. What indeed. Durrell would have been worth following on Twitter…

  18. I’ve spoken with programmers who believe that everything in life can eventually be brought to a series of 1’s and 0’s – I think it is interesting, but only as a lens in which to view our own creativity. By breaking apart the creative process into a series of rules, I think humans can learn – and create more effectively.

  19. Thanks Jessica, yes it’s interesting to break things down into their component parts [1’s and 0’s, rules, lego bricks, whatever], but I’m more interested in what you can build with the pieces afterwards…

  20. Well, they actually create professional classical music:

    http://www.guardian.co.uk/music/2012/jul/01/iamus-computer-composes-classical-music

    http://www.nature.com/nature/journal/v488/n7412/full/488458a.html

    Iamus’ debut CD, recorded by London Symphony Orchestra and first-shelf musicians can be listened here:

    melomics.com/@iamus/cd-iamus

  21. abd saleh says:

    okay the thing is we have to ask the right questions
    what Idea means ?
    what thinking means ?
    what is the process of thinking ?

    well , i thought about it and i came up with this , there is no original idea , every thing we make and we think its new , its only an upgrade for something we used to know , like the car , it was a wagon before with horses and then we replaces the horses with engines and so on .

    so in brief i can say its what you know , your background of knowledge , we have information and an algorithm to process these information , if we could give the computer the knowledge base we have in our minds and give the computer the way to process these information like we do , then a computer can be smarter than we are .