Brian Kenji Iwana and Seiichi Uchida, two researchers at Kyushu University in Japan, have trained a ‘deep neural network’ to analyse book covers to see if a computer can identify the genre of a book by its cover art and design. The network proved accurate 20% of the time, with the system doing well with books about computers and technology due to their generally similar layouts used. Cookbooks were also among the most easily identified, but only if the cover featured photos of food. If the image was of a chef the neural network tended to give ‘entirely ambiguous’ results.