…and I Heart Huckabees, Lost in Translation, Fahrenheit 9/11, The Life Aquatic With Steve Zissou, Kill Bill: Volume 1 and Sideways, according to this well-written in-depth New York Times Magazine article Sunday. In one contender model, Napoleon Dynamite accounts for 15% of the total error, and a small group of mainly independent movies above totals more than half the remaining error. To paraphrase the article: when a previously unknown team introduced one significant improvement, singular value decomposition, and smashed into 4th place, other teams soon followed suit and now all top ten use the technique.
Writer Clive Thompson wondered: The question haunting Netflix as well as recommendation engines used by Amazon and iTunes: Just how predictable is human taste? If we can’t understand our own preferences, can computers be any better? Sometimes the way singular value decomposition groups movies makes sense, but other times no rationale seem apparent. The algorithms find connections so deep and subconscious, even customers themselves wouldn’t recognize why.
