This was going to be a follow up piece on my AI-assisted little introduction to spectralist composition, but is rather more (if briefly) about generative AI itself. I asked my, er, "research assistant" to pull together a short list of spectralist and post-spectralist work featuring classical or electric guitar, with a brief commentary. Here's what it came up with (my comments on its comments in italics:
Fausto Romitelli - "Dead City Radio" for Electric Guitar and Ensemble: This mesmerising piece combines the electric guitar's unique timbre with spectral orchestration, creating an otherworldly sonic experience. OK, not a bad start.
Tristan Murail - "Vampyr!" for Electric Guitar and Ensemble: Murail's innovative use of spectral techniques shines through in this composition, where the electric guitar interacts with the ensemble in a haunting exploration of sound. Actually, this is a solo electric guitar piece, so I'm marking you down for that - although it's fantastic!
Brian Ferneyhough - "Kurze Schatten II" for Solo Electric Guitar": This technically demanding piece challenges the electric guitar's capabilities, combining post-spectral elements with extended playing techniques. Mmm... it's for solo classical guitar, and Ferneyhough is in NO WAY a spectralist. So nil points on that one (although it IS great and I will inlcude in a later post about New Complexity composition).
Michael Levinas - "Le Rouge et le Noir" for Solo Classical Guitar: Levinas' composition explores the expressive potential of the classical guitar, with spectral influences enriching its emotional depth. I've found nothing on this other than a bunch of scholarly articles, I'm afraid.
Philippe Leroux - "M" for Electric Guitar and Orchestra: Leroux's imaginative work features the electric guitar as a soloist in dialogue with the orchestral forces, showcasing spectral orchestration and the guitar's versatility. Ditto.
Magnus Lindberg - "Era" for Electric Guitar and Ensemble: Lindberg's post-spectral language beautifully intertwines the electric guitar with a diverse ensemble, creating a dynamic and engaging musical dialogue. Ditto.
Benjamin Thigpen - "Metal, Skin, Wood" for Prepared Electric Guitar and Electronics: Thigpen's electroacoustic exploration includes prepared electric guitar elements, combining spectral and electronic soundscapes. Ditto. Although I ended up enjoying an article about the future of aluminium supply.
Javier Torres Maldonado - "Cyber Guitar" for Guitar and Electronics: Torres Maldonado's piece features the guitar in a captivating conversation with electronic elements, exploring spectral timbres in an electroacoustic setting. Er, ditto - this is getting boring, frankly.
Kaija Saariaho - "NoaNoa" for Flute and Electronics (or Flute and Live Electronics): Though not specifically written for guitar, Saariaho's use of electronics and spectral techniques provides an inspiring example of how the guitar can be integrated into similar works. Well, thanks for that, but not exactly what I was after.
OK, back to me, in non-italics.
So this was, to put it mildly, a disappointing exercise. To date I'd been less than impressed with ChatGPT's ability to write posts for me, or at least in anything approaching what passes for my style, but I have been broadly pleased with the results of my experiments in terms of research. But it really fell over here. For one thing, it completely missed this piece for classical guitar by one of spectralism's greatest figures, Tristan Murail (although in fairness, it did find "Vampyre!", albeit mis-describing it):
My guess is that a lot of the failing here is that in emplying a large language model (LLM) approach, ChatGPT is better able to search, say, scholarly articles than it is on, well, just browsing YouTube. And in the case of this piece of research, the latter appraoch would be rather more beneficial. Although I stress that this is just a hunch, as this is very much not my area of specialism, and I feel like I'm reverse-engineering its methods. That being said, my own "manual" Googling on the topic hasn't turned up much more either, so it's more than possible that I've set it an impossible task in this instance. I'll keep trying, and perhaps set something a little less challenging next time.
As for the "tone" thing, it's just not getting "me". I've stripped out some of the more egregious examples of it mangling my tone in this post, but really, I don't think even the 25-year old Simon Hopkins would have used terms like "captivating conversation" or "the expressive potential of the classical guitar"*. My suspicion here (again in the spirit of reverse engineering) is that I'm not giving it a large enough of a sample of my writing to get to grips with, so I'll try harder on that front next time, too. More soon!
* I'll have more to say about my 25-year old self in a couple of days.
Did the machine make up any pieces of music i.e. tell you about things that don't actually exist?
Posted by: nick | August 10, 2023 at 09:54 PM
Not quite that bad - it was just somewhat economical with the truth...
Posted by: Simon H | August 11, 2023 at 10:17 AM