Neural networks, fuzzy logic and Netflix
October 18th, 2006
Once upon a time, the Infonaut tried to imagine what the space between all the ones and zeroes was about. Was there any truth to fuzzy logic, a possibility that binary theory allows for some wiggle room? Is there a space between “on” and “off”? The problem proved to be unhealthily dizzying and the puzzle was shelved. Until this month’s New Yorker arrived. Malcolm Gladwell’s piece on Epagogix, “the Formula,” reads vaguely like a Bruce Sterling short story, except apparently it’s non fiction. In an effort to successfully predict the box office outcome of studio fodder, a group of entertainment attorneys, headed up by Dick Copaken, have tapped the deep depths of risk assessment modeling to create the unthinkable - a formulaic, scientific model for predicting a script’s potential as a blockbuster. Factoring in bizarre, yet apparently salient details such as wardrobe, opening scene visuals and semantics, “the formula” is apparently hitting the mark with a suprising level of accuracy. Part marketing, part digital voodoo, and part outright absurdity, the company seems to be making its mark by running scripts through a neural network with a high degree of success. Does this help create better content? Decidedly not. It simply predicts, through the collective history of millions of statistics, the probability that one script is a better bet than another. Deal or no deal?













