So Predictable

October 19, 2006 9:49 PM

So Predictable - Malcolm Gladwell talks at the recent New Yorker Festival about success-predicting software for the music and film industries.
posted by forallmankind (18 comments total)

it's pretty long (with Q&A at the end), but there's a few highlight links to get the gist. Includes analysis of where The Interpreter went wrong, and suggestions for how it could have been improved.
posted by forallmankind at 9:50 PM on October 19, 2006

Related New Yorker article by Mr Gladwell.
posted by sien at 9:59 PM on October 19, 2006

Markov chain-based statistical studies, along with requisite computing power and storage increases, now make a peculiarly novel, if roughly conceived kind of “metacomposition” technique possible, adding yet another layer of nuance to the fickle relationship between popular music consumption and creation.

PolyphonicHMI offers a product called Hit Song Science, analyzing components like tempo, cadence, rhythm, etc. in Billboard chart hits, looking for commonalities. This product now figures into the creation of much of current popular music, to help maximize marketability and unit sales.

Network engineer Loren Wilson developed a project wittily named “Pitchformula”, using Pitchfork reviews of previously released music products as a basis for composing songs that would (in theory, at least) earn equally favorable reviews and sales.

Psychoacoustical research done at MIT Media Lab examined cognitive processes involved in listening to pop music, its results leaving many avenues open to the music industry for increased profitability.

Music evokes very strong responses in people. Figure out what people like and you can make billions.
posted by Blazecock Pileon at 10:30 PM on October 19, 2006 [1 favorite]

What a fascinating presentation. Not only is Gladwell an engaging speaker, the topic covers some really fascinating ground and not just in terms of "arts".

Towards the end, during the Q&A, there's some bit about "gaming the system(s)". And it goes into some brief discussion about the tensions between creative types and their "cottage industry" thinking and the big industry types and their desire for repeatable manufacturing of product. Its a string of concepts that isn't constrained just to the "arts" but to any creative, inventive endeavour vs. the desire to repeat the success indefinitely.

-1 point for a 1 link post to a video. +3 points for the quality of the content.

Thanks for giving my brain something good to chew on.
posted by C.Batt at 11:12 PM on October 19, 2006

I am, at the same time, fascinated and completely afraid.
posted by tev at 11:27 PM on October 19, 2006

That was extremely interesting, thanks. I was aware of mathematical analysis of music, and that makes sense to me, but analysing storiesI find staggering. I had lots of things to say, but I'm going to have to cogitate on this one for a while.
posted by MetaMonkey at 2:55 AM on October 20, 2006

Interesting. I wonder if anybody has done something equivalent for visual arts?
posted by rongorongo at 4:06 AM on October 20, 2006

great link, thanks.
posted by johnny novak at 4:12 AM on October 20, 2006

Cool, I just read Gladwell's NY'er article yesterday. One thing I was thinking about is some of the concepts from Gladwell's The Tipping Point. In The Tipping Point Gladwell illustrates how small groups of people can have a large impact on product sales. How does the tipping point concept jive with a seemingly mathematically certainty of predicting pop music or film success?
posted by sexymofo at 5:04 AM on October 20, 2006

This makes me wish I understood math better.

Its amusing to me that they go to all this trouble to create "hits", then compress the hell out of their "product" during mastering to make it annoyingly unlistenable.
posted by Enron Hubbard at 5:13 AM on October 20, 2006

Thank you - I enjoyed that very much.

Is it just me or is he Sideshow Bob's Good Twin?
posted by Grangousier at 5:30 AM on October 20, 2006

This makes me wish I understood math better.

Because Gladwell certainly doesn't. What a butchery of basic science his work is.
posted by fourcheesemac at 6:26 AM on October 20, 2006

fourcheesemac: could you expand on that a bit?
posted by leotrotsky at 6:51 AM on October 20, 2006

I'm missing something here. What is the raw input for this (for the film one)? How are these films described to a computer well enough to build a statistical model? It sounds like this is "unsupervised" learning, where the parameters are not pre-determined. If so, how is the data inputted in such a way that a model can be determined?
posted by rottytooth at 7:51 AM on October 20, 2006

David Poland, who writes about the movie business has this to say about the part of the Gladwell article that applies to the movie business. He thinks that the movie prediction is worse than utter crap. I tend to agree.
posted by bove at 10:06 AM on October 20, 2006

Thanks bove; that gives some counterweight to Gladwells article. This David Poland even speaks of a con game.

I gladwell selling, unwittingly perhaps, his credibility?
posted by jouke at 12:35 AM on October 21, 2006

Hah. I wonder if that could be applied to "spin" in politics?
posted by black8 at 2:27 AM on October 21, 2006

This video contains spoilers of Dear Frankie and The Interpreter.

A lot of what he talks about in terms of how to write a successful movie (e.g. don't cover the villain's entire face with a mask) is already well-covered territory in film schools and screenwriting texts (and, I think, pretty intuitive to people who spend their spare time thinking about movies). That said, it didn't stop Spider-Man from cleaning up.
posted by bingo at 5:33 AM on October 21, 2006