DeepBach: The AI System Producing 21st-Century Classics

December 26, 2016

Germany composer Johann Sebastian Bach remains a staple in classical music. His illustrious works such as Brandenburg Concertos, the Goldberg variations and Mass B minor remain some of the most powerful pieces ever made. Bach spent over 50 years perfecting his ability to adapt rhythms, forms and textures from styles abroad. Many regard him as one of the greatest composers to have ever lived.

And one new artificial intelligence system can recreate Bach’s biggest achievements in just minutes.

Fast forward to the 21st century where music is created by statistical models and artificial intelligence in no time. DeepBach, an AI employing an agnostic machine-learning approach, produces music that would intrigue Bach himself. Gaetan Hadjeres and François Pachet developed DeepBach. The system has the ability to ‘compose’ Bach-style polyphonic choral music indistinguishable from its 18th-century counterpart.

[Image Source: Pixabay]

Flow Machines Deep-Learning algorithm is trained using 400 chorale sheets of Bach’s music. Flow Machines chose Bach’s music due to his prolific composition all in a singular nature. This allowed DeepBach to quickly learn the style and rhythm of the composer. The user can then impose notes, rhythm’s or cadences in order to re-harmonize the melodies. For those who want to try a discrimination test by putting Bach and DeepBach head to head should try this quiz.

The video below is a harmonization in the style of Bach generated using DeepBach.

Alternatively, If you are more into pop music and want to experience the first pop song production by an artificial intelligence machine, take a look here.

AI Machines are often automated to complete bounded tasks such as playing chess or classifying DNA sequences, but when it comes to creativity, the human reigns superior. It is said that machines do not have the high level of cognitive ability required to be creative. So much so Erica Wagner, an associate professor in machine learning at the University of Oxford says “Creativity is arguably the most difficult human faculty to automate”. But does DeepBach not show the creative ability described here?

It might appear as if the machine shows signs of creativity. However, it doesn’t know what being creative means. Therein lies what could be the defining factor. Would the algorithm pause for a minute and evaluate the authenticity of its own work? Perhaps that is what true creativity is, knowing that one is creating and what DeepBach is producing is simply a ‘creativity template’. Nonetheless, DeepBach is an impressive tool with unbounded potential and we hope to hear more 21st century renderings of our historical classics.

SEE ALSO: This Artificial Intelligence Program Knows What You Fear

Featured Image Source: Flow Machines

(Visited 24 times, 1 visits today)

Newsletter

Stay on top of the latest engineering news

We'll keep you posted!

Comments