At the time of his death, Beethoven was working on another symphony. In fact, he had corresponded about it with the Royal Philharmonic Society of London. As was his style, he was jotting down melodies, working out solutions to problems, and generally starting the forging of another symphony. Unlike Mozart who composed as effortlessly as others breathe, composition for Beethoven was often a struggle, a process of heating, hammering and blacksmithing ideas into final products.
In any case, his work on this piece had not progressed very far. He only left behind fragments that are very incomplete and certainly not fully thought-out, genius-approved music. If only we had some way to get inside his brain – thinking the way he would have thought – and could finish his work. Well, while no computer brain will ever be able to rival Beethoven’s mind, with a bit of clever AI software we can at least paint a pale imitation of his intended canvas.
So, with a little help from artificial intelligence programming, assisted by a human composer, a very hypothetical score of the third and fourth movements was created from Beethoven’s fragments. In particular, Ahmed Elgammal helmed an AI team that input completed Beethoven scores with the aim of creating a set of Beethoven composition rules that would inform further composition. Austrian composer Walter Werzowa was commissioned to weave together the extant fragments with the AI-generated product. You can read about the project here.
After perhaps lowering our expectations a bit, it is interesting to listen to this very approximate estimate of what might have been. Here is a recording of the Scherzo and Rondo as premiered on October 9, 2021 and performed by the Beethoven Orchestra of Bonn.
There is certainly a Beethovenesque quality to the music and it even occasionally reaches a higher plane, but overall it doesn’t inspire visions of genius. In the same way that a painting “from the studio of Titian” was painted by an apprentice, this music sounds like it was composed by a student of Beethoven which, actually, is what the AI program is. That is not to detract from the accomplishment; creating Beethoven-quality music is probably the highest compositional bar one can imagine.
The organ part in the Rondo is a nice touch, a musical experiment that late Beethoven might have contemplated, much as he placed voices in the final movement of the 9th. But unfortunately the music seems tentative and immature especially compared to the very confident and “modern” nature of his late quartets.
The music is also very light compared to the gravitas of the 9th. It might be logical to assume that a 10th symphony would have surpassed the 9th. Of course, that’s not necessarily true. After writing the defiant, heroic 3rd symphony, Beethoven relaxed a bit with the 4th, a pause he probably needed before shaking the world again with the revolutionary 5th. And as wonderful as the 7th and 8th symphonies are, perhaps they represent him resting up before tackling the 9th. So, there is no reason to think that the 10th would have obviously eclipsed the 9th. Further, the contrast between his even and odd symphonies is often remarked upon. Numbers 3, 5, 7 and 9 are quite dynamic while 4, 6 and 8 are more mellow. Would even-number 10 have continued in that theme? Who knows.
In the end, computers are not humans. And among humans, there was only one Beethoven.
PS: Another data point: This is not the first time inquisitive musicologists have pondered the question of the 10th. Here is a stab at the symphony’s first movement, as assembled by Barry Cooper. Check out the hefty outburst at 4-1/2 minutes.