Future-of-publishing experts believe these will come to be useful for purposes of selling books. So far I'm skeptical. The books Amazon predicts I'll like seem completely random, although that may be because such books as I've bought from Amazon are either very obscure and hard to find, or else gifts for overseas relatives. The last book Amazon recommended to me wasn't even in a language I recognized, and the one before that was about how to draw sports figures.
But it's true that a decade or so ago, I bought an Irish folk CD from Amazon and Amazon recommended to me a jazz piano CD I'd just bought elsewhere. So I guess a reasonable percentage of people who bought the one CD also bought the other – it would be fascinating to know why.
Pandora does a better job for music. Interestingly it sometimes tells you what logic it's following -- when I told it I liked Robyn Hitchcock, it told me that I like “music where the vocals are central to the mix.” A weird way of looking at it – I would say that I like songs where the lyrics are important. And if the lyrics are important, the recording engineer is more likely to make the vocals central to the mix. But who knows, maybe at some level I really do like the sound of vocals being central to the mix, and got interested in words as a side-effect of that... my daughter's attitude seems to be far more common – she told me, “I want songs to have words, but I don't want the words to be clear enough to understand.”
On what basis can we say that someone who likes Calvino might also like Cortázar? Is it possible, in principle, told my five favorite Calvino stories, to deduce my five favorite Cortázar stories and two favorite Philip K. Dick novels?
Incidentally, Google announced a while back that they've turned on an
"artificial intelligence" tasked-array system capable of designing its own webpage -- this does look like the webpage Wintermute from Neuromancer would design, if Wintermute was really into pandas.