You may have to turn your volume up to hear this video properly. Researchers at the University of Toronto are using a project called neural karaoke to teach computer programs to write songs. PhD candidate Hang Chu fed the program tons of existing tunes to study, then dance, then lyrics.
For the final step of the latest work, the program trained on a collection of pictures and their captions to learn how specific words can be linked to visual patterns and objects. When fed a fresh image, the program can compile some relevant lyrics and sing them using phonemes, or units of sound, linked to the words in its vocabulary. The system builds on previous work that could take a picture and generate lyrics in the style of Taylor Swift. Should the program choose words that it cannot say, it replaces them with an “oooh” sound.The result might be called impressive for a machine, but if a human wrote this, you would suspect possible drug abuse or maybe brain damage. (via the A.V. Club)