This automated script browses the internet for yesterday’s most interesting news article and fetches the attached images. It then applies a complicated algorithm to detect and cut out its most interesting part, posting them daily on the twitter account @yesterdaysnews. Although its method is closely modelled on the latest computer vision methods, the bot’s choices are often quite different from a newspaper reader’s – especially with typical, information-rich press images. Thus, it obscures the original context, rather produces nondescript, meaningless pixel clutter.
