March 20, 2018

Editorial: Artificial intelligence: coming to a medley near you

Piping and drumming is a rich environment for original compositions, arrangements and scores, thanks to the innate creativity of our musicians combined with the competitive need to come up with new and exciting content for competition medleys.

The percussion side is creatively even richer, with virtually every drum section having to develop all-new scores.

At the risk of oversimplifying things, writing tunes from nine generally unflinching notes and a set amount of embellishments for fewer than a dozen established categories of tunes and time signatures makes things straightforward relative to a symphony. A piper can typically write an original tune within a year or two after starting the instrument.

Snare drummers are completely capable, too, of trying the hands at a score for a tune or two.

The rub, of course, is creating a good tune or score that people want to play that doesn’t sound like something else. Plenty of composers crank out stuff no one will play because it’s unmelodic, derivative, or just doesn’t “go.”

Enter artificial intelligence. It’s here and improving, and companies like Aiva or Amper are using AI to create music for movies or dance clubs. And the amazing thing is, it’s pretty good.

These companies access millions of existing musical compositions and create a bunch of mysterious algorithms. Essentially, all that has to be done is input the parameters of the type of music they want, and out comes the composition or song.

If you’re like us, you’ve now got the heebie-jeebies. While this stuff is fascinating, it’s also unnerving because it has the potential to be so useful. We can see a day shortly when a major Hollywood studio uses AI to score a feature-length movie. We’ve already heard music that is purportedly the “perfect” pop song. It’s okay, and it’s a start.

Using AI for a pipe tune or a drum score should be easy if they can do that. Picture it: tell the gizmo that you need a singular tune of a certain style, key, tempo, and degree of difficulty to start your medley. Or maybe it’s that elusive two-part strathspey that everyone can play, that isn’t stripped of technique, that’s as compelling as “Orange and Blue” – or better.

Or you need an original snare score to a tune, but one that suits the band’s style and the competency of all players. Easy-peasy.

There is no longer any pressure to find not only new but also good content. There is no longer squabbling between composers in the band, jockeying for airtime, or egos running amuck. Just dial up the machine that never talks back to the pipe major.

We would imagine that some boffin boy or girl piper or drummer someplace in the world is already on this. They’re feeding in every pipe tune they know and writing “rules” for what’s good, bad, mediocre, easy, hard, and so forth. Statistical analytics of what’s played and popular will inform algorithms on what has been done and overdone.

After all, twenty or so years ago, a professor working with Panasonic in Japan created an entire robotic pipe band. The metal fingers of the automatons clack away together at “Scotland the Brave,” played on pipes blown by air compressors.

Artificial intelligence (an oxymoron if there ever were one) is coming to pipe and drum music someday. And it could be a lot sooner than we think.

What do you think? Feel free to add your thoughts using our comments system below.



  1. It’s been done for Irish trad:


    About 98% of the output sounds like imitations of Irish trad by someone who’s heard it but not really understood it. About 2% throws out ideas that are genuinely interesting or authentic sounding.

    The trouble is with the bagpipe that it’s not hard to write authentic sounding tunes; it’s hard to write *original* authentic tunes. Coming up with something as sonically unique as, say, Roddy MacDonald’s Good Drying is a challenge I suspect is still well beyond any machine.

  2. Calum, I totally agree with you with regards to the derivative nature of most of these AI outputs. I think some of the most interesting bits are the scraps or “mistakes” these machines make. I sent a YouTube of some of these to a friend a couple days ago and he thought some bits sounded like Aphex Twin. I also hear Raymond Scott’s electronic music, though part of that is simply the tinny orchestration. Some great computer-generated names too, reminiscent of pipes|drums “bandnamer” tool. Anyway, enjoy these crazy, crooked tunes – they get pretty wild by the end! https://youtu.be/RaO4HpM07hE

  3. AI for music is still fledgling. As usual, it’s all about money, or filthy lucre, as the case might be. The movie or music industries might see AI as a way to eliminate a lot of time and money by, in effect, replacing the middle-people with technology. It’s happening in virtually every industry where corners can be cut with automation, and thereby more money might be made, though not without intangible costs and value.

    As we’re seeing with social media, it’s paradoxically perhaps the most antisocial thing ever contrived, and beggars the question: Who **asked** for this stuff?

    But AI in this regard is inevitable. Someone will work at AI for piping/drumming composition as a side-project, maybe as a PhD thesis. It might well produce decent tunes and scores at a figurative push-of-a-button, perhaps a bit like the pipes|drums TuneNamer and BandNamer tools: you never know when something actually good might pop up.

    But, again as might eventually occur with social media, AI composing could start as fun, and end in regret, or at least a “What ever were we thinking?” realization.

    A prediction: there will be a prominent band to be the first to showcase an AI tune expressly “written” for their needs. And then there will be the first band to play an entire AI-created medley. And then of course there will be the first foosty adjudicator to squash the whole premise . . . just because he/she can.

  4. Music is an art form…at least most of it. (I still have difficulties with some genres.) Piping is heavily dependent on melody. Whether the tune was written by a master composer, like Donald MacLeod, or a piper who happens to write a good “one-off” tune, or by some computer, the “art” will be to decide what is musical, and what is crap. It has to sound like something. It has to achieve what the Pipe Major wants…the feel…the mood. That’s where the art comes in. A well-constructed medley should (IMHO) incorporate “something old, something new,” A medley comprised of nothing but new unheard, unfamiliar tunes has a challenge. I would imagine that electronically, computer-generated music would also face this challenge…..maybe more so.



Forgotten Password?