In the latest news sure to make tin foil hat-heads flip out, two computer scientists Lior Shamir and Jane Tarakhovsky of Lawrence Technological University in Michigan have developed a computer program that apparently “understands” visual art. The program demonstrates a computer’s capacity to evaluate and dissect art in a manner similar to the way art historians scrutinize a work. The program is able to analyze paintings by detecting similar styles and placing them within a historical context. Eerily enough, the machine was even able to outperform un-trained humans; it’s supposedly able to tell the difference between “closely related schools of art such as Early and High Renaissance or Mannerism and Romanticism” (damn). In the initial experiment, the computer was given 1,000 paintings by 34 well-known artists. Then, free of human interference, the researchers allowed the algorithm to break down similarities between them based on visual cues. It succeed in linking post-impressionists Gauguin and Cézanne, as well as the founding fathers of surrealism Salvador Dali, Max Ernst, and Giorgio de Chirico. Guess that Art History degree has just become even more useless, huh?
(Image via Kurzweil)
I don't think a machine's ability to decode visual cues is all that threatening. Now, when the machine decides it knows what it likes and doesn't like and breaks down my front door to adjust what's hanging on my wall, well, that's a whole 'nuther cup of tea.
Let's just hope that Skynet will have good taste.