JUST WHEN you thought it was safe to go back in the cinema, Fox has employed a company who are using AI to study movie trailers in search of what makes a good (and profitable, it is Fox) flick.
Researchers have published a paper called Convolutional Collaborative Filter Network for Video Based Recommendation System (or CCFNVBR for short, which sounds like someone stepping on a plug, so they call it ‘Merlin' instead) to explain how machine learning is being used to go frame-by-frame in a way humans simply couldn't do, labelling, learning and then suggesting things in other trailers that match.
It's not dissimilar to the way that the BBC supercomputer created its AI generated programme earlier in the year.
The idea is simple - the more objects match, the more likely the trailer is to appeal to similar people. So once they get that downpat, they'll actually start being able to commission the "right sort of movies" in the first place.
An example in the paper is recent Marvel spandex ripper Logan.
"a trailer with a long close-up shot of a character is more likely for a drama movie than for an action movie, whereas a trailer with quick but frequent shots is more likely for an action movie."
In a demonstration, 11 movies out of 20 predictions were matched by the old fashioned method of asking people what they like. Not every one is right - for example Terminator Genisys is probably not a great match for The Revenant, and we're not sure where the Ant Man / Spectre link comes from, but for early doors results, it's pretty impressive.
The important thing to remember here is that this isn't the end of the experiment - far from it, in fact. The point of machine learning is that, if it goes as it should, it will keep getting better and we'll all end up with an amorphous mass of virtually identical Hollywood movies. And won't that be fun. μ
Oh and it'll also help give aural pleasure
But it might still not be enough to make virtual reality super appealing
And a ridiculous competition
Now you can talk to your silly-looking earbuds too