Tuesday, January 19, 2010

Computers and Art

Two articles about what computers have been getting up to. What struck me about both is that they describe how computers are moving beyond raw computational problems and are now helping with problems which require "taste". Computers are now better at distinguishing fake paintings from genuine ones, and can mimic the style of chess grandmasters so that humans can no longer tell whether they are playing a human or a computer pretending to be that particular human.

The first is about the use of computers to distinguish fake paintings from authentic ones. The approach is based on the idea of a visual vocabulary which is distinctive to each artist. This vocabulary consists of "words" termed "filters" which can be used to very compactly describe any work by that artist but would be much less efficient at describing a work by another person.
To carry this out computationally, the team obtained very high quality scans of all the Bruegel drawings, both authenticated ones and fakes. They broke the digital images of the authentic ones up into tiny patches, each just a few pixels wide, and then used a machine learning algorithm to identify a small set of those patches that could be used as filters, in imitation of the visual system. The algorithm picks the filters so that the smallest number possible is needed to generate every patch in the Bruegel. These formed the “words” of Bruegel’s own unique visual language.
The team was inspired by how the human visual system has evolved to very efficiently encode the world we find ourselves in; presumably, if we had evolved on another planet, or underwater, our visual system would have use a very different set of filters.
Graham, Rockmore and Hughes applied these ideas to art authentication by imagining an organism that had somehow managed to evolve a visual system while only ever viewing Bruegel drawings. The organism would be able to see Bruegel drawings using very few filters, but when it looked at anything else — including fake Bruegel drawings — it would have to use many more.
The second is an article by Kenneth Rogoff in which he describes what he thinks the impact of Artificial Intelligence on the Economy will be.
Many commercially available computer programs can be set to mimic the styles of top grandmasters to an extent that is almost uncanny. Indeed, chess programs now come very close to passing the late British mathematician Alan Turing’s ultimate test of artificial intelligence: can a human conversing with the machine tell it is not human?
Ironically, as computer-aided cheating increasingly pervades chess tournaments (with accusations reaching the highest levels), the main detection device requires using another computer. Only a machine can consistently tell what another computer would do in a given position. Perhaps if Turing were alive today, he would define artificial intelligence as the inability of a computer to tell whether another machine is human!

No comments: