In the latest episode of #CoffeeWithAGoogler met with Christine Robson of Google’s Machine Learning teams. We spoke about what was new and exciting in the realm of Machine Learning, including Project Magenta. We see how Magenta came up with a machine-generated piece of music, and how some artists jammed along with it!
In this piece of music, we trained the model using the first four notes of ‘Twinkle Twinkle Little Star’, and using a recurrent neural network, trained with everything from the Beatles to Bach, it came up with the tune.
Listen to a great conversation with an expert on Machine Learning — where Christine steps us through some pretty advanced concepts, but makes it so clear that even I could understand it!
Magenta is open-sourced, online and part of the TensorFlow project.