Computer says: Rachmaninov was more original than Beethoven

A South Korean team created a computer model to examine 900 piano works by 19 composers, written between 1700 and 1900.

Rachmaninov came out tops with the ‘highest combined novelty score’ – the extent to which each work differed from its predecessors.

Juyong Park of the Korea Advanced Institute of Science and Technology, said: ‘Our model allows us to calculate the degree of shared melodies and harmonies between past and future works and to observe the evolution of western musical styles by demonstrating how prominent composers may have influenced each other.’

Beethoven didn’t even make the top ten.

Read here.

share this

Share on facebook
Share on twitter
Share on linkedin
Share on google
    • AI (artificial intelligence) is showing great promise in a number of industries, including financial services, healthcare and automotive. The general weaknesses in the category include poor data quality and bias. Attempts to bring AI to the arts have produced some interesting results. For example, there is a recent AI-powered study that sorted through some late works by Shakespeare showing the passages written by collaborators. Attempts to finish Schubert’s 8th symphony using AI were silly and totally unconvincing. Same with Beethoven’s 10th symphony. No doubt this South Korean use of AI to determine originality is filled with all sorts of landmines (bias, data inconsistency and subjectivity) that make it invalid.

      • Nonetheless, it seems that AI is quite useful for sonic art, since distribution of sound splashes and random pitches is exactly what algorhythms like to do.

  • One should not forget that AI is called ‘Artificial Intelligence’ for a reason.

    ‘Originality’ is a psychological category which cannot be measured by material means, like counting deviations from standards – which evolve themselves also, anyway. This ‘research’ looks like squaring the waves of the sea.

  • Who cares about what the computer says?
    The programme was created by a human…who apparently did not favour Beethoven:end of story.

  • Although I believe this is absolute nonsense, it does raise some interesting questions . When one thinks about it, the notes and harmonies of some early Beethoven works are not particularly original ( Op 2 Sonatas for example ) but then one comes to the E flat Sonata Op7 which both harmonically and in its sheer scale broke the mould : nothing like it had ever been done before as far as I am aware ). For a while he then went back to convention, and then works appeared in ever more original forms e.g. Op 26 , 27 1/2 , later piano works such as the last 5 Sonatas , Synphony 9 etc . Thus thus the Algorithm seems to have been incapable of recognising this.

    The originality of Rach’s piano music intrigues me : as a long term student of the Preludes and of the 48 I have come to realise that Bach was a considerable influence on Rach, because both have the habit of altering a repeated figure one note at a time. Nonetheless, so many of the Preludes are immediately recognisable as Rachmaninov’s work . Can this be said of Mendelssohn ? Doubtful at best in my view

  • Reminds me of the great David Lodge (in “Changing Places”, if I remember correctly), who already in 1975 made fun of a computer based analysis of literature. You can count the words, the notes, the motives, but that’s about it.

    • Yes, but often the busy surface is there to hide the lack of musical substance, and then the music turns into Great Stuff for pianists but less so for the audience.

  • There’s a good summary in the (London) Times today. If you just focus on piano music (bye, Stravinsky) and you consider primarily harmony, not rhythm, then this is where you land…

  • Beethoven was notable for taking a kernel of an idea and working it to death rather than come up with a brand-new melody whenever it was time for a new idea. (I actually read an essay by somebody-or-other who basically accused Mozart of laziness because he didn’t rework his material the way Beethoven did — he just tossed out another melody.)

    If that’s a lack of “originality,” then we should all be so unoriginal.

  • There’s an old saying among computer programmers that “to err is human; to really f*** something up requires a computer.”

  • This is why AI is fundamentally stupid. It can only compute, not feel or truly hear. It’s the difference between factual knowledge and deep wisdom.

    (Nothing against Rachmaninov, btw, whom I mostly enjoy – especially the All-Night Vigil and his 2nd Symphony).

  • Hilarious! When are these ‘experts’ going to learn that computer programs are not neutral things but have all the biases of the people who wrote them. This piece of crap research indicates how worthless they are. Rachmaninov concertos are among my favourites and highly enjoyable. But original as compared with L van B? Not too much. Where’s the trash can?

  • A lot of commenters here deride what the computer (actually, the programmers) did in the analysis. Comparing and correlating melodic and harmonic fragments to glean insights into creative process is sort of what musicologists do on a smaller scale, but the analysis here did it in a Big Data way.

    The point of any such metric is to recognize what’s being measured and what we can or can’t infer from it. Does anyone dispute that Beethoven reused or recycled musical ideas? (I’m thinking just now of his signature ascending trills). But Beethoven’s innovations had more to do with musical structures and development than with the musical kernels a pianist plays at any given moment, and his influence is something else altogether. As the authors themselves state in their paper:

    “Novelty measures how different a work is from the past, representing originality and unpredictability of generation. Influence measures how much a work has been referenced in the future, representing its success and impact as an inspiration for future creations.

    “While originality and success are both important characteristics of meaningful creative works, they do not correlate perfectly. That Handel was less novel than Bach and many others but had more influence on Classical and Romantic composers is a good example.

    “Similarly, Beethoven, Schubert, and Liszt were less novel than Mendelssohn and Schumann, but eventually came to exert more influence and inspire more piano music to follow.”

    For the record, they also found Bach, Brahms and Mendelssohn at the top of their ranking for ‘musical novelty.’

  • Hahaha. How about an informed discussion?

    Here’s the actual paper: https://arxiv.org/pdf/1905.08665.pdf

    Quotes:

    – “we are leaving out the tempo, rhythm, instrumentation, etc…” – basically everything was “chordified” – i.e. converted into a series of simultaneous notes. And only piano.

    It considers only “novelty”, as in appearance of combinations that have not appeared before historically (within the solo piano set of 900 works), so:

    – “it is unlikely that novelty alone is a sign that the work is of any value; if it were, one could simply assemble elements not found in the older works, and claim to have created the most valuable work.”

    – “..our results above suggest that novelty alone would not cause one to be considered ‘the greatest’; Beethoven, for instance, stand among the lower half in computed novelty. This is in line with many recent research findings that a creative work’s impact on its posterity does not
    depend solely on the degree of its novelty..” – DUH!

    – “..an interesting finding was that being more novel, i.e. more willing to break from convention, did not necessarily translate to being influential on the posterity. This means while novelty is still necessary in a creative endeavor [..] it cannot account for all the creative, artistic qualities..” – You think??

    – “This suggests a future research direction in which a more elaborate modeling of codeword transitions and other elements of music are considered. Possibilities in the former category include the change of the number of notes, the tonality, melody and the chord progression, to name a few. Possibilities in the latter category include the rhythmic structure of music that is recently gaining increased attention and the global structure of a composition, given the common assertion that the most significant innovation in the piano music during the Classical period was the establishment of the sonata form which may have little to do with the codeword transitions. Extension beyond the piano is also an obvious possibility.”

    Yes, it is a good idea to incorporate melody, tonality, chord progressions, rhythmic structure and musical form in future studies!

    • Isn’t that what good musicologists already do?

      I suppose a better conclusion (if there is one) would be that a proper assessment of music should depend equally on its novelty and it’s influence. And it would have been more fruitful to examine the connections between novelty and influence (perhaps that was too complicated)

      In any case, it looks like they only used piano music, which will naturally produce ridiculous results.

  • Debussy and Stravinsky alone were streets ahead in revolutionising the musical language when compared with Rachmaninovs 19th century nostalgia. To compare him with Beethoven, whose late works ( take alone Opus 106 and 131 ) are among the most radical and trailblazing experiments in Western music is not revealing anything other than the idiocy of AI when it comes to evaluating creative work.

    • It may be informative to be reminded of the difference between sound and tone. The only time that modernist, original sound artist Stockhausen engaged with music, was in his student days, when he made a statistic analysis of a movement from a Mozart piano sonata. The idea was, to determine how often certain interval combinations occurred in relation to the overall structure, to discover the meaning of the structure through the variety of numbers of interval combinations. The outcome was nil, which – from a scientific point of view – was a valuable outcome. Science is not the best instrument to understand music which is a psychological, not a scientific art form.

  • I’m a little surprised Rachmaninov would be #1, but I’m not surprised he outranks Beethoven, who was great, but was he original? Was Mozart? They both built better versions on existing platforms- Haydn mostly. I can’t see the list, but I would think Wagner and Berlioz would rank higher. Then Schonberg and Stravinsky. Debussy?

  • Obviously this research was a waste of time, money & resources as many dubious scientific researches are. I personally am quite capable of enjoying Rachmaninov in small doses, his Symphonic Dances much more frequently than the rest of his output, but he is played far too often by pianists these days. Then the world did not turn out the way the modernists of the early 20th century imagined. The people want generous melodies and Rachmaninov gives it to them.

  • The best music by Rachmaninoff, where he equals the definitely-great, is the 1st mvt of the 3rd piano concerto (after which the rest is an uncalled-for bonus). That singular movement is a piece of genius, in the true sense of the word. Much of the other music – be it for orchestra or piano – has so many great episodes or moments, but is quite uneven. Obviously, only so when compared with the best of the repertoire. It is a good thing he is no longer being looked-down upon from an academic snob point of view, comparing him with Debussy, Stravinsky, Schoenberg, etc. which is unfair anyway.

  • >