Last week, recently released sci-fi movie Morgan unveiled a new trailer.
But there was a twist: the trailer was created by an Artificial Intelligence.
Producers enlisted the help of IBM’s Watson – an advanced data processing AI – to come up with the clip.
You can watch it here:
How did this work?
Watson poured over hundreds of sample trailers, analysed them all, and used machine learning to come up with an idea of how the perfect trailer should look and sound.
Watson was allegedly able to determine what makes a moment eerie, how the music and actors’ tone of voice changed the mood, and how technical aspects came together to make a complete trailer.
It was then fed the full film, from which it chose scenes for the trailer.
Interesting stuff. And it’s only one of several recent examples of AIs being enlisted to create or edit entertainment media.
Which begs the question: could we soon see the entertainment world taken over by artificial intelligence?
A growing trend
Morgan‘s trailer isn’t the first use of an AI in the movie-making process. Back in June, an AI algorithm was used to create the entire screenplay for Sunspring, a sci-fi short set in the near future.
The result, admittedly, was a film that was at times incoherent – and unintentionally hilarious.
The screenplay featured directions like “He is standing in the stars and sitting on the floor”. and it was clear to most viewers that the 9-minute effort had been created by a machine, not a human.
But it can be hard to tell the difference
Morgan‘s trailer fared better though.
Perhaps due to its shorter running time, the promotional preview was similar to those created for the film by human editors – even to experts who know these kinds of trailers inside and out.
David Hughes has been creating movie trailers for a quarter of a century and runs his own ‘trailer house’, Synchronicity.
“I wouldn’t know if I hadn’t been told,” he says, noting that Morgan’s trailer passes his own personal Turing test.
But he was quick to point out that, while impressive, Watson could simply have been working with some good material in the first place:
“The film looks good, but it’s a lot more difficult to make a bad trailer from a good film than it is to make a good trailer from a bad film.
“‘Does it make you want to see the film?’ is the only criterion by which trailers should be judged. I want to see Morgan. So for me, job done!”
AIs in other media
Outside of the movie world, AIs have been put to use in other areas.
One such area is that of social media. Back in May, Microsoft launched Tay, a teenspeak chatbot that could communicate to Twitter users using internet slang.
The results were disastrous though, with nonsensical Tweets the norm. Things also soon got worrying when Tay ‘learned’ how to be racist and sexist. The bot even called for mass genocides.
It was a PR nightmare and Tay was quickly deactivated by Microsoft. But does this faux pas make it less likely for AIs to ever be truly accepted in the entertainment world?
It could become the norm
Not if creators can get their AIs right.
In the world of creative media, Hughes notes that even human-created content tends to rely on tried and tested techniques that could easily be replicated by an AI:
“Someone recently analysed the ‘peaks and troughs’ of recent film trailers, and found that a lot of them followed a kind of pre-set formula.
“Trailers are generally going to open with a film distributor’s logo and end (more or less) with the film title.”
The same could be said for other types of content. Pop music can easily be broken down into slight variations on the same song structures, so if an AI was able to work that out, it’s possible we could see machine-created music filling up the charts.
In fact, sites like JukeDeck already allow users to generate new songs in seconds.
It’s a brave new world. As to whether the current AI experiments will lead to an entertainment revolution however, only time will tell.