​ ​
happy mag subscription

New Yamaha AI piano system blurs line between human and machine

If you ever needed evidence of artificial intelligence taking over, this video of a Yamaha AI piano system performing live is such proof.

At the recent 2019 Ars Electronica Festival in Austria, Yamaha premiered the Dear Glenn Project AI System. In a dimly lit church, attendees watched as the piano played itself, in the style of legendary pianist, Glenn Gould.

Yamaha takes a step towards artificial intelligence with its AI piano system, capable of playing in the style of Glenn Gould – and raising all sorts of questions.

Like something out of HBO-produced series Westworld, the keys of the piano played of their own accord, the instrument standing solitary on the stage. Later, the piano was joined by another – this one played by contemporary pianist Francesco Tristano –  as well as a wind trio with members from the Bruckner Orchestra Linz. The result was an ensemble performance, combining humans and AI.

Gould was one of the most celebrated classical pianists of the 20th century, renowned for his technical proficiency and personal eccentricities. He was a mysterious figure who maintained a highly private life but was nevertheless loved by many.

A short accompanying documentary video delves into the process of how the whole thing came about. In order to see how Gould interpreted any given piece of music, over 100 hours of his audio recordings were analysed. This was combined with human input from Tristano and other pianists familiar with Gould’s style.

The documentary follows A.I. engineer Akira Maezawa as he interacts with the technology and then in turn with the performers.

“What are you doing?” A voice off-camera asks.

Maezawa answers: “I’m bringing Glenn Gould back to life.”

He notes that A.I. tends to create a very “bland” performance and that the system learns better by having humans teach it.

“So this isn’t just about analysing archival recordings, this is about using human love for Glenn Gould to extract parts that make his performance special.” 

The AI system was then able to use its machine learning technology to play in the style of Gould, recreating defining elements like his unique touch and pacing. The system is capable of playing pieces that Gould never performed himself.

However, the other incredible thing about the AI system is its ability to play along with other real performers – on the fly. Yamaha’s AI Music Ensemble Technology is able to analyse the performance of human musicians almost instantly and play along predictively. So even though Gould is no longer alive, these performers get to experience a sense of what it would have been like to play with him, and the audience gets to hear it.

“To bring artificial intelligence into connection with music should not end as an objective in a competition or an exhibition of achievements of two different forms of intelligence,” described Martin Honzik, Senior Director of Ars Electronica. “[It] should be the beginning of a discussion which searches for forms of application in order to improve us in our being human and to expand and improve our virtuoso actions.”

The whole endeavour was made possible with the support of Glenn Gould Foundation. Speaking on the project, executive director, Brian M Levine, described that the system should be “taken into the music mainstream” due to the “spirited debate” it will generate.

The system brings up interesting questions surrounding the intersection of humans and robots, but also artificial intelligence and art. There is something undeniably uncanny about watching a machine interact with art – a creation that is so achingly human.

In some ways, it feels completely wrong, but perhaps the room for experimentation, collaboration, and even a kind of resurrection, will lead to a whole new era of how we interact with music.

To learn more about the Yamaha Dear Glenn Project AI System, head over to the Yamaha website. And if you want to witness Gould performing in the flesh, check out the video below.

FIND OUT MORE

Leave a Reply

October 29, 2019