Beethoven did not complete his 10th symphony. Computers using AI now have.

[thewayne]

Now this is an interesting use of AI.

Beethoven was commissioned for the 9th and 10th symphonies around 1817, and in 1824 we got the 9th, which is an amazing work. But he died before he completed the 10th, and all that exists of it are some sketches, some fragments of music that assuredly do not add up to a full symphony.

Until now.

A team was assembled of music historians, musicologists, computer scientists, and composers, with the goal of bringing this symphony to completion.

They trained an AI by taking phrases from previous symphonies and teaching it how that phrase grew into larger movements, then eventually the process matured and grew. When they were satisfied with the results, they fed the system the phrases from Beethoven's notes for the 10th.

The world premiere performance of the entire happened this last Saturday. I assume a CD will be available perhaps in time for the holiday season? If you search Amazon for Beethoven 10th, there is lots of material about the process that created this. There's one music CD, but I'm pretty sure it's not this material.

EDIT: There IS a CD of this, but it's poorly reviewed. And most of the reviews are in German, conveniently Amazon has a built-in translation service. It seems like the CD quality itself is not very good, so maybe a better performance and recording will be forthcoming as time passes. The trick is to search for Beethoven X, not Beethoven 10th.

https://smile.amazon.com/Beethoven-Orchestra-Kaftan-Walter-Werzowa/dp/B09CRM3HYD/

This is a 3:38 scherzo, and it's damn good. Posted September 10, just a month ago, and performed by the Beethoven Orchestra Bonn.

Watch on YouTube

This is a 22 minute recording, posted October 9:

Watch on YouTube

Very cool article:
https://thenextweb.com/news/computer-scientists-completed-beethoven-10th-symphony-syndication

And the Slashdot thread:
https://entertainment.slashdot.org/story/21/09/29/2345224/ai-completed-beethovens-unfinished-tenth-symphony#comments

Comments

[silveradept]

It was certainly interesting to listen to. I also note the large presence of people claiming that they can tell that it's not truly Beethoven for this reason or another. I guess they're waiting for something that sounds more like a deepfake?

[thewayne]

I don't know, maybe it's like everybody doing their own "vaccine research": everybody's gotta put their own opinion out there.  I listened to the long piece, and it sounded really good to me!  To me, it sounded distinctly Beethoven-ish, and it was quite pleasing to my ears and I liked it.  I am not an expert in classical music, much less Beethoven, though I've listened to a lot of both, and I heard phrases in the piece that sounded - to me - distinctly like Beethoven to me. 

So for me, it passed the test.  Sounds like they're trying to prove a negative, when, in fact, what we have is a true "What If?"  Take the same software, a Group B of computer scientists, composers, etc. and let them loose on the same source material, and it would probably sound slightly different.  Take the same Group A and let them loose on IBM Watson, assuming it could be modified to extrapolate and compose music, and you'd have another variation.  It is a What Could Have Been, and I like it.

[silveradept]

I think that's the point both supporters and detractors are settling on, is that we don't actually know because the composer died before he finished the work. I feel like you could make an interesting album or playlist of exactly what you've described here: starting with the same sources, train a composing neutral network to finish the piece and have the same professional orchestra record the results. (After making sanity checks about whether or not the music produced by the network can actually be performed.)

I'll bet all of them would sound Beethoven-ish enough to be enjoyable.