Artificial intelligence is a major forefront of creative technology. We've reported in the past about AI's usage to remix rhythms, play virtual drums, and even write whole tracks. Yamaha, though, have taken the idea one step further with their new project, Dear Glenn.
The project is an attempt to recreate the performance style of pianist Glenn Gould using deep learning. As described on Yamaha's website,
As shown in Yamaha's graphic above, the process involved the analysis of many of Gould's recordings (no small feat, since the performer's idiosyncratic humming would often make it onto tape) which in turn were used alongside experts on the man himself to train a neural network.
The result is, supposedly, an AI system that can be given "any music score" and play it in the style of Glenn Gould. Indeed, it can even play along with a human, although it helps if said human is renowned classical pianist and electronic music connoisseur Francesco Tristano, who joined 'Gould' for a duet last month at the Ars Electronica Festival in Austria last month.
Look mum, no hands!
The final missing piece of the puzzle, then, is how the performance was realised for piano on stage. The likely answer is that the feat was achieved using Yamaha's Disklavier self-playing piano, whose hammers can be controlled by an external input.
Perhaps the fact that Yamaha didn't labour the impressive finer details of the Disklavier system goes to show exactly how much of an achievement their AI technology really is.
"When it comes to performance and playability, Casio goes to great lengths to recreate the grand piano experience, achieving impressive results": Casio AP-S200 digital piano review
“It feels incredibly authentic at the base-sample level, and provides vast potential which extends way beyond its original use by rock bands”: Cherry Audio Wurlybird 140B review
