Just read the article by Babak Saleh and Ahmed Elgammal on Large-scale Classification of Fine-Art Paintings. You can find their academic article on data analysis of art works here: http://arxiv.org/pdf/1505.00855v1.pdf
Babak and Ahmed develop “a machine that is able to make aesthetic-related semantic-level judgments, such as predicting a painting’s style, genre, and artist.”
I find this quite interesting.
If through automated data analysis we find relations between artworks that we did not find before, do these new relations enrich our appreciation and understanding or did we not find these relations earlier because we cannot understand and appreciate them?
I think it will enrich our appreciation and understanding.
We, humans, are good in giving meaning and I expect we’ll build and add on new relations between art works that were discovered through machine learning.
In “Project Definition” people could send a SMS text message with a keyword or several keywords to us and based on the words in the message we would then do an image search and display the results on large screens.
Instead of defining a word with other words we wanted to let images define that word.
It was interesting to see how people would interact with the screens and each other. After seeing a definition people would think of something else they wanted to see. So based on the Liechtenstein definition in the image above perhaps the next person would send a message with Warhol or Monaco or anything else depending on their particular associations. Instead of the images only defining the word, the combinations of images would stimulate us to create something new.
Data Analysis of Art
When data analysis finds new relations between art works I expect something similar to happen. Instead of dismissing these new found relations, after seeing them we’ll probably either “understand” them immediately or connect the new relations with something we understand or have seen before.