Apparently, It Is Obvious

Although the results from Tuesday’s Adaptistration Classical Music Challenge are unlikely to be the final word on the matter, it appears that it’s still pretty easy to tell the difference between an orchestra of human beings and one of computer generated sounds…

The challenge consisted of participants listening to two different recordings of Edvard Greig’s Holberg Suite, Mvt. #1. One version was created using the Vienna Symphonic Library and the other one with a human orchestra. Afterward, participants identified which recording was which or if they couldn’t tell the difference.

27 readers had the time to compare each recording and 25/27 (93%) were able to accurately identify each sample, 1/27 (4%) was incorrect, and one 1/27 (4%) could not tell the difference, as the chart below illustrates.

(click to enlarge)

I think this was a worthwhile test because the Guardian Unlimited technology correspondent, David Smith, claimed that new Vienna Symphonic Library software package is capable of mimicking “human musicians in the performance of greats such as Bach, Beethoven and Mozart so convincingly that a casual listener to Classic FM would be unable to tell the difference”.

Apparently, the casual listener that also reads this blog would beg to differ.

If you missed the challenge and didn’t vote, no worries, you can still feel free to compare the two recordings:

Sample #1 – Edvard Greig, Holberg Suite, Mvt. #1, 4.8MB MP3 file
Sample #2 – Edvard Greig, Holberg Suite, Mvt. #1, 1.2MB MP3 file

About Drew McManus

"I hear that every time you show up to work with an orchestra, people get fired." Those were the first words out of an executive's mouth after her board chair introduced us. That executive is now a dear colleague and friend but the day that consulting contract began with her orchestra, she was convinced I was a hatchet-man brought in by the board to clean house.

I understand where the trepidation comes from as a great deal of my consulting and technology provider work for arts organizations involves due diligence, separating fact from fiction, interpreting spin, as well as performance review and oversight. So yes, sometimes that work results in one or two individuals "aggressively embracing career change" but far more often than not, it reinforces and clarifies exactly what works and why.

In short, it doesn't matter if you know where all the bodies are buried if you can't keep your own clients out of the ground, and I'm fortunate enough to say that for more than 15 years, I've done exactly that for groups of all budget size from Qatar to Kathmandu.

For fun, I write a daily blog about the orchestra business, provide a platform for arts insiders to speak their mind, keep track of what people in this business get paid, help write a satirical cartoon about orchestra life, hack the arts, and love a good coffee drink.

Related Posts

4 thoughts on “Apparently, It Is Obvious”

  1. Drew,
    I listened to both clips and could tell immediately which one was the real thing…but, I do have a background in classical music. I played the clips for my wife and a friend of ours (both who have zero formal music training) and they both had no trouble telling the difference. They picked it out after listening for about 6 seconds.

    So – I think the message is that although your reading audience is predisposed to be musically inclined, and thus will likely be a bit better than average at picking out the “real” performance, even those who have no formal training are clearly capable of picking out the human vs. android orchestra.

    An additional observation: if we were allowed to listen to only a very small sample of the sound clip (like 1/4 of a second), I think the numbers would be drastically different. The actual tone doesn’t sound all that different; it’s the subtelties, the way the orchestra swells and recedes, the accenting of specific notes that provide context for the meter, and so forth. The real orchestra “plays” the rests – the android orchestra just suspends animation.

    My two cents.


  2. This experiment reminds me of some research I uncovered for my Master’s thesis. I did work on timbre theory, approaching it from a neurocognitive perspective, and read a lot about various timbral subtleties that allow the brain to differentiate sounds (mainly to identify them).
    Incidentally, the human brain is capable of identifying discrepancies in synchony of attacks up to 15 milliseconds! In other words, a difference of 15 milliseconds between the entrance of one player and another can be heard. I think that particular aspect plays a significant role in distinguishing the two ensembles in your listening samples.

Leave a Comment