November 8, 2016
Machine built to judge books by their covers reveals pretty much nothing
by Taylor Sperry
Since time immemorial, jacket designers have lamented the failure of publishers and editors and marketing directors to understand the exquisite genius of their cover art — and their consequent insistence that a design be rejiggered to look “more like a work of serious non-fiction” or “more like a beach read” or “more like a definitive biography.”
Publishers and editors and marketing directors can say these infuriating things because there exist in the market very clear ideas of what certain kinds of books look like.
Nonetheless, scientists Brian Kenji Iwana and Seiichi Uchida of Japan’s Kyushu University have developed an algorithm “to study book covers and determine the category of book they come from,” MIT Technology Review reports. “The results make for interesting reading,” apparently, because the machine identified the correct genre “more than 20 percent of the time. That’s significantly better than chance.”
For the time being, we might look elsewhere to appreciate technology’s triumph over the human brain. For more on the homogeneity of cover art see Chad’s great piece on “why all books are yellow now.”
Taylor Sperry is a former Melville House editor.