The thin, grim line between music and necrophilia

The thin, grim line between music and necrophilia


norman lebrecht

January 27, 2022

From my monthly essay in the February issue of The Critic, out now:

The recent attempt to manufacture a tenth Beethoven symphony by means of Artificial Intelligence has proved about as intelligent as cloning Albert Einstein out of paper from his wastebasket. 

The outcome, 21 minutes long, is performed on YouTube by the Beethoven Orchestra of Bonn. It welds fragments of a discarded project onto bits of other symphonies in a manner so uninspiring that it reduces Beethoven to the level of a Hummel. The work’s opening, a student-essay paraphrase of the fifth symphony, is all you need to hear. 

The rest just gets worse. What possessed the brains of Bonn to think they could create a Beethoven symphony ab virtually nihilo? Probably the thought that fellow-necrophiles had done such things before….

Read on here.

You don’t have to play the video (but you probably will).


  • J Barcelo says:

    Well, I’ve heard worse! The CD catalog is filled with symphonies of (mercifully) forgotten composers who couldn’t write as well as AI. And AI probably spells “necophilia” correctly as “necrophilia”.

  • kk says:

    Tahir Nejat Özyılmazel, b. July 24, 1948 in Yozgat, Turkish pop singer and actor. Is that the one?

  • M McAlpine says:

    Well at least it is listenable, which is more than can be said for 95 percent of the simply awful modern music we hear foisted upon audiences

  • RW2013 says:

    The music of Hummel is better than it is given credit for.

  • Y says:

    The intensity of the sneering vitriol hurled at this worthwhile and interesting experiment is simply astounding to me. It makes me wonder if people actually feel threatened by this avenue of research and composition.

    Personally, I find this work and the possibilities it reveals quite thrilling. The resulting music is, in my opinion, both thought-provoking and enjoyable. I’d like to hear more.

    • La plus belle voix says:

      On the one hand the result is unintentionally risible, just listen to the suddenly vigorous da-di-da-da strings motif at 00:20 and the ensuing bizarre timpani flourish, interesting, but nothing like Beethoven; on the other hand AI is in its infancy, so who knows where it might lead? Composition students in their first semester called upon to produce pastiche works do better than this.

      • Y says:

        “Composition students in their first semester called upon to produce pastiche works do better than this.”

        Dear Lord! Not in my ample experience! Have you seen the state of composition departments these days? You’d be lucky to find a first-semester composition student who can harmonize a simple melody, let alone write a whole symphony.

        • John Borstlap says:

          In my study years at the conservatory, students got particularly trained to destroy whatever musicality they brought to the lessons.

    • Will Wilkin says:

      I would go insane if human composers were replaced with AI. I would stop listening to music because it would no longer exist.

  • Peter San Diego says:

    I’m not tempted to listen to the video (or to watch it, for that matter). However, regarding the project as an engineering experiment rather than an artistic endeavour, it may represent a fruitful step (but hardly the first one) in AI composition (computers were “taught” to “compose” fugues decades ago). And just as one still watches humans playing for the world chess championship even though computers can beat them, so AI is (a fortiori!) at no risk of displacing human composers and compositions.

    • John Borstlap says:

      The idea that AI programmes can compose music, stems from the idea that writing music is a purely intellectual thing. It is not, and intelligence of whatever kind only plays a part in the creation process. Imagination cannot be put into an AI program, because it is a natural given, and a realm beyond the purely intellectual.

      • Peter San Diego says:

        I actually agree with you. AI composition may yield increasingly interesting music, listened to as experiments, but I rather doubt that the time will come when an AI composition will be able to pass the musical equivalent of a Turing test.

  • Tomtom says:

    It hasn’t got the ‘quirk’. Without a ‘quirk’ it cannot be by a human being.

  • John Borstlap says:

    I like the first sentence, and I imagine a couple of excited scientists frantically pasting paper fragments together, with trembling hands, thinking they will find another relativity theory.

    I agree with most of the article, but the Mahler X as polished by the Matthews brothers and Bertold Goldschmidt is a good piece, very Mahlerian, and – maybe except for a couple of places – convincing.

    The Elgar reconstruction was a Payneful affair, as one commentator noticed after the première. I did not find the result very convincing, although the scoring is quite good.

    • christopher storey says:

      For once, I cannot agree with john Borstlap about the Payne/Elgar 3 . It is a most wonderfully nostalgic piece of construction, the most curious aspect of which is that the first minute or two , which music was completed and orchestrated by Elgar, are the one part of the work which does NOT sound like Elgar

  • Anthony Sayer says:

    This is what happens when contemporary classical music is incapable of adding to the canon.

  • christopher storey says:

    Gawdstrewth, spare us !!

  • Dante Santiago Anzolini says:

    it sounds as if the mechanism that created this pastiche is in real infancy, unaware of certain ‘taste’, and ambitiously driven to throw much into the mix. I just wonder what is the reason to insert data and look obliviously for the result -any result at all. If this is laughable for its formal randomness, unexpressive harmonies, and above all the fact that one hears modern turns and clishes that escaped into the mix and was used warmly and effusively, why bother?

    a mistery. It would be more humane to bring the computer to make verifiable analysis of classical music data, then to use it in a composition class today, under an experienced teacher – to see if kids learn harmony and counterpoint like in the great old days, where computers were still fed by pumched carton cards and did not dear to compose!…or should youngsters keep writing all types of clusters in all type of formal discredit to harmony, all justified by another propelled and fallacious myth about the use of math in music? well, anybody’s choice….

    the triplet motif that turns into the three eighth-notes is …..that ultimate really bad joke on how not to write the beginning of the 5th… not to hear it, how not to feel it….how not to use it….take “La Gioconda” and draw moustaches on it….

  • Jim says:

    The project is doomed. I’ve said it before and it bears repeating: Beethoven was very good at not writing in the style of Beethoven.

  • John Berry says:

    An extremely long infancy! About 60 years ago the ILLIAC string quartet was “composed” by (I think) a Univac computer, and the score was published in Scientific American. There has been no progress since then for inanimate “composers”. On the other hand, if AI stands for AUGMENTED INTELLIGENCE, the growth of digital tools to assist human composers has been stunning!