hi INDiA Copyright 2020
Google honours Bach with first AI-powered Doodle
New Delhi, March 22 (IANS) Want to create your own music? Try Google’s first ever Artificial Intelligence (AI)-powered Doodle, dedicated to celebrate the birth anniversary of renowned German musician and composer Johann Sebastian Bach.
Bach, who lived during the Baroque period of the 18th century, produced hundreds of compositions including cantatas, concertos, suites and chorales.
The Doodle, developed in partnership with Magenta and PAIR teams at Google, is an interactive experience where the composer is playing an organ in celebration of his March 21, 1685, birthday under the old Julian calendar, encouraging musicians to compose their own two-measure melody.
With the press of a button, the Doodle uses Machine Learning (ML) to harmonise the custom melody into Bach’s signature music style.
With the interactive Doodle, one "can create own melody, and through the magic of ML, the Doodle will harmonise your melody in Bach’s style", Lauren Hannah-Murphy, Programme Manager at Google AI said in a statement on Friday.
"You can also explore inside the Doodle to see how the model Bach-ifys familiar tunes, or how your new collaboration might sound in a more modern rock style," Hannah-Murphy added.
To create the AI-powered Doodle, the teams first build a ML model called Coconet and then customised it to match Bach’s musical style.
Coconet can be used in a wide range of musical tasks-such as harmonising melodies, creating smooth transitions between disconnected fragments of music and composing from scratch.
The team then trained Coconet on 306 of Bach’s chorale harmonisations. His chorales always have four voices: each carries their own melodic line, creating a rich harmonic progression when played together.
Further, to ensure that people around the world could use the Doodle, the teams used PAIR’s TensorFlow.js, which allows ML to happen entirely within an internet browser.
The ML model can also be worked on Google’s new Tensor Processing Units (TPUs), for cases where someone’s computer or device might not support TensorFlow.js.
"The Doodle is the first one ever to use TPUs in this way," Hannah-Murphy said.