Machine creativity

A very stimulating meeting at the British Library last week, called ‘Digital Conversations’, asked whether computers can be creative.

There were several short talks:

– one about a machine that generated chord progressions for new jazz tunes. (The speaker, Tom Hedges, brought a jazz pianist who played a progression the machine had generated, though later on the pianist said it could have been as easily generated by the audience calling out four numbers at random, and then proceeded to demonstrate this…)

– one about the world’s first computer generated musical, which is opening in the West End shortly. (Not so much computer generated actually, as I understood it the computer made suggestions to a couple of composers. This one left me cold I’m afraid.)

– a couple of painting machines, including Paul, which I have blogged before. Paul is not ‘creative’, in the sense that it sketches a model in a particular style: it is told the style and given the model. (It’s a great programme though.)

The other was The Painting Fool whose website says “I have been built to exhibit behaviours that might be deemed as skilful, appreciative and imaginative.” According to the designer Simon Colton, the machine reads the Guardian on-line before painting, and its mood is then determined by a sentiment analysis programme. (If the news in the Guardian is very bad, it will refuse to paint at all.)

How it paints is then a mixture of a real response to the model subject, and its own ’emotions’. The model’s face is also analysed for emotions. For each emotion a set of artistic instructions is given, eg:
“For anger we specified a line rather than curved segmentation style; a colour mapping to shades of green, except for the eyes, where shades of red were used; and a pencil rendering style which quickly sketched the outlines of segments.”
“For disgust, we specified low saturation greyish colours (to indicate rottenness); a shape transform which distorted the face by stretching it; and a decreasing-circle acrylic painting style which further distorted the face and highlighted the facial features.”
(etc).
The software can also create collages, representing news stories: “The system worked via internet retrieval of a news article, text extraction of important keywords, internet image retrieval using the keywords, and nonphotorealistic rendering of the images.”

– a machine by Stephen McGregor that wrote ‘poetry’ – though to my mind the poetry was just lists of phrases, with little attempt to provide structure or coherence. For example:

“the repetitive attention of some traditional african chants
a heroic struggle, like the personality of a soldier
an unbearable symbolic timing, like a scream
blue overalls, each like a blueberry”

The poems were very well read by the British Library’s own poet-in-senior-management, Richard Price, but still didn’t sound quite right to me.

There’s a good article about computer generated poetry by Elizabeth Preston.

Speaking about copyright and the legal issues of computer creativity, Burkhardt Schaefer mentioned that there’s a company called Qentis which claims to have generated – and copyrighted – every possible combination of 400 words, so that anything written from now on will already be copyrighted and writers will have to pay them royalties. (Luckily this is a spoof project written by a digital artist, Michael Marcovici…. but it’s a worrying thought. Also the maths are impossible. If there were only 400 words in the English language, factorial 400 is 64034522846623895262347970319503005850702583026002959458684445942802397169186831436278478647463264676294350575035856810848298162883517435228961988646802997937341654150838162426461942352307046244325015114448670890662773914918117331955996440709549671345290477020322434911210797593280795101545372667251627877890009349763765710326350331533965349868386831339352024373788157786791506311858702618270169819740062983025308591298346162272304558339520759611505302236086810433297255194852674432232438669948422404232599805551610635942376961399231917134063858996537970147827206606320217379472010321356624613809077942304597360699567595836096158715129913822286578579549361617654480453222007825818400848436415591229454275384803558374518022675900061399560145595206127211192918105032491008000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000
and of course there are very many more English words – Google estimates 1,025,109.8! – which your 400 would have to be chosen from, making the maths even more horrendous – so I think we creatives are safe for a while yet!)

One of the most interesting statements of the whole evening, by Simon Colton, was that creativity wasn’t the same thing as a random number generator – ‘there are so much better ways to do it’.

See separate post for my own doubts about a common ‘creative’ methodology.

Leave a Reply

Your email address will not be published. Required fields are marked *