'Bot Dylan' AI writes its own catchy folk songs after studying 23,000 tunes
The future of music: 'Bot Dylan' AI writes its own catchy folk songs after studying 23,000 tunes
Computer composes new tunes after being trained on 23,000 Irish folk songs
This allowed AI to learn the patterns and structures that make for a catchy tune
So far it has created over 100,000 new machine 'folk tunes', researchers say
It marks a significant step forward for the capabilities of artificial intelligence
By RICHARD GRAY FOR MAILONLINE PUBLISHED: 05:01 EDT, 26 May 2017 | UPDATED: 07:02 EDT, 26 May 2017
Researchers have created a 'Bot Dylan' computer that is capable of writing its own folk music.
The system uses artificial intelligence to compose new tunes after it was trained using 23,000 pieces of Irish folk music.
This allowed the machine to learn the patterns and structures that make for a catchy tune before it created its own pieces of music that we showcased at a concert in London this week.
It marks a significant step forward for the capabilities of artificial intelligence.
HOW DOES IT WORK?
The computer system has been trained using 23,000 pieces of Irish folk music.
Dr Ben-Tal and his colleagues chose to use Celtic folk music to teach their AI composer as it has a relatively well-defined structure and plenty of material to train it from.
They used transcriptions of folk tunes that use a reduced form of music notation known as ABC.
Their machine is able to take one ABC symbol and suggest the next by drawing on the patterns it learned from the tunes it was initially fed.
Dr Ben-Tal said so far it has created more than 100,000 new machine 'folk tunes'.
Creative and artistic expression has long thought to be beyond the capability of AI, and many have insisted it will be one of the few areas where humans will have an edge over machines.
But the new computerised composer, developed by scientists at Kingston University in London and Queen Mary University of London, suggests the line between man and machine may not be so clear.
Dr Oded Ben-Tal, a senior lecturer in music technology at Kingston University in London, said: 'We didn't expect any of the machine-generated melodies to be very good.
'But we, and several other musicians we worked with, were really surprised at the quality of the music the system created.
'People are reluctant to believe machines can be creative – it's seen as a very human trait.
'However, the fact of the matter is, technology and creativity have been interconnected for a long time and this is just another step in that direction.'
It is the latest in a number of AI technologies now capable of producing artistic work.
Google's DeepMind machine learning algorithm has been churning out psychedelic works of art while AI has also been used to write - albeit bad- movie scripts and short novels.
There have been other attempts to get computers to write music such as Google's Magenta project, Sony's Flow Machines and British start-up Jukedeck.
But he adds that the machine does not appear to be able to generalise what it learned about music to other contexts.
For example, it still has no concept of how rhythm works, but replicates it from the songs it has learned.
Dr Ben-Tal believes the AI system may never be able to completely replace human composers but it could provide a tool to help amateur musicians write their own music.
He said: 'For beginners, a system like this would help get you started and avoid the intimidating aspect of composing your own tune as you could work interactively together.
'An experienced composer could work with the system to generate new ideas by using their own musical concepts as a starting point.
'This opens up a whole new world of possibilities for music making.'