Art is no longer in the eye of the beholder - a robotic art critic has been designed which rates paintings based on their creativity, uncovering the influence of others.
A team of AI researchers began with software that attempts to classify objects in an image into different categories using what are known as 'classemes'. Then, they fed in Wikiart - a database of about 62,000 paintings and used network algorithms to figure out which paintings influenced others.
The computer was then about to output a list of paintings ranked according to creativity - how much influence they took from their predecessors, and how much influence they went on to have. The less of the former and the more of the latter, the higher the score.
No fan of Rodin
The computer rated some artists including Michelangelo, Vermeer, Monet, Goya, Munch and Lichtenstein, very high - while dismissing the creativity of artists like Rodin, Ingres and Durer. Its creators, Ahmed Elgammal and Babak Saleh say that the results match quite closely in most cases with the assessments of art experts.
Elgammal and Saleh add that they believe the algorithm could be used in other contexts - like sculpture, or literature - too. But it's likely to also be useful in image processing systems too. The full details can be found on arXiv.