Journalists forced the neural network to write scripts Christmas movies
It turned out bad, and scary.
The experts of MIT Technology Review’s will knight and Karen Hao decided to see if a neural network to create the perfect Christmas movie and come up with a name for it. For this purpose, the program textgenrnn they loaded scenario 360 Christmas movies, and waited for the result.
It turned out not as expected. The program coped with the task of openly bad — the neural network gave a bunch of rambling sentences, reminiscent of just sets of words.
According to experts, it could be due to a small amount of program data and its imperfection.
Let the main task of the neural network failed, but most of the proposed names for the Christmas films have proved suitable for fighters or horror films.
Knight and Hao chose the most interesting ones and even illustrated a couple of suitable pictures.
And this. https://t.co/FkESFmrJlG pic.twitter.com/LPONj0qmIJ
— Ray 호 Ho ?? ?? φ (@bubb13_b0y) 21 Dec 2018
And how could we forget this. Ho, ho, ho. https://t.co/vTtUTyoP7s pic.twitter.com/I9LYRWizL5
— MIT Technology Review (@techreview) 21 Dec 2018
So, a neural network is proposed to make a film about a family Christmas terrorist, who fights for the new city and the film’s about the gardener, a detective who is trying to find a rural killer magically ruined Christmas, about the Princess of the cannibals, But the most interesting names of “Living in solitude from Christmas, Night, Santa Claus”, “Woman of chaos took home the faith” and “the Christmas battle”. That sounds really scary.
- Artificial intelligence has learned to understand memes
- The neural network was taught to “translate” the complaints of people in social networks into the language of medicine
- Picture created by neural network, graced the magazine cover (photo)