How AI completed Schubert’s Unfinished Symphony No 8

22 Feb 2019576 Views

Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+Pin on PinterestShare on RedditEmail this to someone

The English Session Orchestra. Image: Huawei

Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+Pin on PinterestShare on RedditEmail this to someone

Did neural processing hit the right notes for a lost masterpiece?

Huawei recently proved how artificial intelligence (AI) combined with human expertise was able to compose the final two movements of Schubert’s famous Symphony No 8.

Commonly known as the ‘Unfinished Symphony’, it has remained incomplete for 197 years and it remains one of the most intriguing pieces of unfinished symphonic music of all time.

Musicologists are still in disagreement as to why Schubert failed to complete the piece – some cite his ill health, others claim that he was distracted by his follow-up piece of work – but what is agreed upon is that he was charting new musical terrain with the piece.

Huawei’s completed version of Schubert’s Symphony No 8 was created by running an AI model using the processing power of the dual neural processing unit (NPU) in the Huawei Mate 20 Pro smartphone, designed specifically with AI-based tasks in mind.

We spoke to Huawei’s chief marketing officer for consumer business in western Europe, Andrew Garrihy, about the project and how it all came together.

How was AI/machine learning used to predict and finish composing the symphony?

Using a powerful combination of human expertise and AI, we taught our smartphone to compose the remainder of the third, and a completely new fourth movement of Schubert’s Symphony No 8.

This was done by running an artificial intelligence model that benefited directly from the processing power of the dual NPU in the Huawei Mate 20 Pro smartphone. It analysed the timbre, pitch and meter of the existing first and second movements of the symphony, and then generated the melody for the missing music.

Similar to how we trained the Master AI image recognition capability in our smartphones, we created a machine-learning AI model to run on the Huawei Mate 20 Pro, which was trained to identify the features of Schubert’s music. The smartphone then predicted similar melodies based upon what it had learnt.

We then combined the power of AI with human expertise as Emmy award-winning composer Lucas Cantor arranged an orchestral score from the melody that stayed true to the style.

How did the project come about in the first place and how many people were involved?

Finding ways in which we can demonstrate how technology can have a positive impact in the world we live in is something we’re passionate about at Huawei.

Last year we taught our AI to drive a car, translate humpback whales’ ‘love songs’ into a piece of music, and used it to open up the world of deaf children through our StorySign project.

With the Unfinished Symphony, we saw another opportunity in which we could use the power of AI and human expertise to highlight the positive impact technology can have in culture today. This wouldn’t have been possible without pairing the technological innovation of Huawei’s AI with human expertise.

Lucas Cantor brought a huge wealth of practical and creative musical knowledge to the project. While our Huawei Mate 20 Pro smartphone then generated the melody for the final, missing movements, Lucas’s role was to arrange a score around the melody that stayed true to the style of Schubert’s Symphony No 8, which an orchestra could then play.

The English Session Orchestra, comprising some of the best session musicians in the industry and resident session orchestra for Abbey Road Studios, then performed the symphony for the first time.

This is a perfect example of how amazing things happen when you combine technology with human expertise to push the boundaries of what is humanly possible.

How was the finished symphony received by the experts? Do you see the AI process being used again to complete other unfinished works of art and culture?

We hope the project was enthusiastically received. Our technology is constantly learning and adapting to ensure our customers are getting the best experience when using their device.

This project proves the positive result of collaborative intelligence, demonstrating that when humans and AI join forces, we can achieve more. This project is just the start and has continued our determination to explore just what else we can make possible to have a positive impact particularly in our culture today.

We don’t have any current plans to finish other pieces of music but if our smartphone is intelligent enough to do this, what else is possible?

John Kennedy is an award-winning technology journalist who served as editor of Siliconrepublic.com for 17 years.

editorial@siliconrepublic.com