Some composers are facing challenging times. Not me. I don't even write music for the public, and I don't require the income. Since my retirement, I've simply been trying to keep myself busy, especially during the autumn and winter months. I enjoy riding my bike, but not when it's cold and windy. Oh, how I miss my pupils. There's no one left to bully anymore.
It's not an ideal time to be a young composer aspiring to carve out a small niche in the music world. In the past, your competitors were other young composers. Nowadays, you have to contend with a ballroom full of high-end computers and the most sophisticated software ever developed. And while you're getting older, the software is getting better. In the future, you might need a walker, and AI will create the accompanying music.
Unless you write purely for the fun of it. Like me. That makes you an amateur, but let me assure you, leading an amateur's life in the arts is quite peaceful and fulfilling.
Replies
What can I say?
Puts a whole new meaning to "heavy metal"...
Are they playing this tempo rubato?
I recall vividly growing up with smug assurance from experts that a computer would never defeat a grandmaster in chess. Computers had the raw serial processing power, but lacked the sort of parallel processing faculty of the human brain, they rationalized.
Then in 1997, Deep Blue defeated Garry Kasparov, now relegated to the depreciated honor of [human] World Champion.Todays engines can defeat Magnus Carlson decisively. One can assert that AI is senseless machinery, mimicking human input, but the matter has become more nuanced, as AI's capabilities accelerate. Chess.com monitors games hosted on its platform for cheating (ie, using an engine on another devce to advise moves) and employs the world's leading experts at that particular task. They specifically look for a patern of "engine-like" moves which contrast sharply with the manner in which human beings think and strategize.
I'd wager composition is just the next watershed, and that we're largely in denial of the inevitability of AI supremecy in every intellectual arena. In short time, AI will compose works that make Bach's seem like Chopstix in perfection of part writing, complexity of counterpoint, and sophistication of architecture.
The evolutionary tipping point will come when man-made AI becomes capable of engineering AI-made AI which is superior to itself. Mankind will have to reexamine its theological conceptions once demoted to the rank of Inferior Creator.
But, I can putatively out-think amino acid sludge, so this all seems fairly unsurprising to me.
But as if to prove the point I'm about to make, chess is still going. Computers haven't made human players obsolete, because humans tend to prefer achievements made by other humans.
I'm still a skeptic of the putative AI revolution. Current AI models are actually nothing new: the underlying algorithms are decades-old, and there's no new insight or breakthrough that the hype would have you believe exists. The only material difference is that now we have access to a large amount of data -- much larger than we've ever had in the past -- to feed to what's at its core nothing but an interpolation algorithm. Given enough data points, you can smooth out your approximation of a curve to a precision that you stop noticing the wrinkles and seams.
But the problem with this scenario is that what you get out of the algorithm is only as good as what you put into it. As they used to say in the old days, garbage in, garbage out. There have been studies to show that ChatGPT has significantly degraded in quality since its initial launch -- after interacting with hundreds of thousands, perhaps millions, of your typical internet user, well-known for generating large amounts of input with low average signal-to-noise ratio.
That's not all; other tests show that the algorithm can only answer what someone somewhere out there has already answered; it does not actually understand the data that was fed to it nor the output it generates. Just this past week I read an article about how someone prompted chatGPT to generate a list of words whose 3rd letter is F. ChatGPT proceeds to generate a list of words containing F, but with a few entries in the wrong position. Then it corrected itself, apologizing for the error, listing what it claims are erroneous entries -- among which are words with F in the 3rd position. Afterwards, it confidently proclaimed what is now the correct list: now with different entries containing F in the wrong position.
Now I'm sure after this incident, the engineers behind ChatGPT will monkey-patch this particular flaw and train the machine to generate correct answers to this particular query, or perhaps a class of questions about letter positions in words. So I wouldn't be surprised if now it starts giving you the correct answer. But this whole fiasco shows that the dream of AI with human-level intelligence is still distant. What we have today is something that can generate output that has uncanny resemblance to human output, but which is at its core no more than a philosophical zombie.
I've listened to AI-generated Beethoven symphonies, and while the machine was able to generate orchestration and melodic contours that resemble existing Beethoven works, I was not able to detect the master's spirit in it; it resembled a patchwork of Beethoven quotations stitched together in a way that did not convey any meaningful musical thought. Quite unlike real Beethoven works, that have the characteristic inner "fire", an emotional impulse, that drives the development of the music from start to end. An overarching musical logic across the span of the entire piece. Such was missing from the putative AI-generated Beethoven symphony; the only thing I heard was the musical equivalent of a text generated by the Dada engine: full of eloquent, well-formed grammatical sentences that fit together in the requitite structure of an essay, but devoid of any actual thought or message behind the words -- it's just a bunch of randomly-selected words slotted into a bunch of randomly-selected sentence structures chosen according to some abstract model of how an essay ought to be structured.
If that's where the future of music is headed, I will gladly continue to compose "by hand", so to speak, and avoid any so-called "music" that has "AI" on its label.
Rosemary Brown wrote like Beethoven (sort of) and she didn't even have a computer.
Chess made another decisive breakthrough with Alpha Zero. This engine learns chess simply by playing against itself and the only criterion for evaluation is how likely the game is likely to be won from that position. There is no chess "knowledge" at all. This way, the engine (and others like it like Leila Chess Zero) can often play what seems to be more interesting and daring chess than humans as well as better. I'm quite sure AI will soon be able to write better fugues than Bach because this is primarily a technical exercise (I've already made my own relative lack of interest in Bach, fugues and Baroque in general pretty clear so am perhaps a bit biased) but they will not be able to write better operas than Janacek or symphonies than Bruckner which to me is a higher art form and more importantly, a more human one. But then we're getting into dangerous territory here as I seem to be among Bach devotees....
-
1
-
2
-
3
of 3 Next