> But unfortunately, this wont change the fact that the classical recording industry is dying and that the sector is only surviving thanks to public and private subsidies.
Bet $100 that all classical music will be AI generated (synthesized and composed) in just three years. Supply side will be infinite.
I would bet heavily against that any such generated music to become popular (except maybe as bad filler music for video games music etc).
The truth is that classical music fan are very very picky about what they like.
Computers have been "better" musician than human for a long time: a midi file is always played "perfectly". Yet it's sounds horrible, mechanised to our ears. The controlled imperfection of a real human musician is what makes the music beautiful.
I don't see any computer composing anything worth listening to in the foreseeable future (and I'm bullish on generative ai).
What I'm saying: take a classical score as is and input it perfectly as a midi file and you are guaranteed to have a terrible output. Yet on paper, it's perfectly rendered as the composer wrote it.
This is why the act of playing music by a human is called a "performance", an "interpretation" in some languages. The musician is not just playing a score exactly as it is written, the musician gives a part of his/her into the act and make the music his/her own to make art.
And you don't think AI can't be programmed to "interpret" too? You haven't been paying attention much, in the last few weeks in particular
https://www.youtube.com/watch?v=yX9J4RIsvOA
So the performers are following some learned patterns of what they need to do with the written score to make it sound not-terrible to you. I disagree that's only possible for a human to achieve.
I don't have a stake in whether AI could or could not interpret a score, but it's still more than "learned patterns". Some performers imbue an interpretation depending on:
- the instrument
- the room or hall
- their personal experiences
- their personal philosophy of music (e.g., Gould and his anti-hedonistic philosophy)
- cultural norms and explicit respect paid thereof (e.g., in the interpretation of a Chopin mazurka)
- available recording technology and expertise
- available audio engineering technology and expertise
- what emotion they want to invoke or evoke to contextualize the music
- their patience
- their willingness to experiment
- etc.
It is not at all farfetched that a performer is taking most of these into account, directly or indirectly, to render a live or recorded performance.
It's not as if there's just this tradition of unwritten rules about interpreting a score that music students learn from their teachers, rather, it's that an interpretation is an amalgamation of experience, context, and on-the-spot reactive decision-making.
For an AI to successfully achieve this, they'll need the same variety of inputs and contexts to learn from.
>Why would anyone want to listen to it then? what would be the point? The acoustic equivalent of a screensaver?
Isn't that exactly how music is consumed nowadays by most people? They put something when driving, cooking, doing homework, etc. It's used mainly for mood/focus and they're not exactly paying close attention to it the same way one would to a audiobook of, say, an abstract algebra textbook. Especially considering how cheap and easy it is to steam music nowadays. I suspect focused, attentive listening only makes up a small minority of the total streams.
>I suspect focused, attentive listening only makes up a small minority of the total streams
The inverse is likely true with the classical audience. Just like with jazz, although there will be a portion of users seeking “vibey jazz” sort of playlists to use in the background while working, the majority of jazz fans consume jazz as albums and are still concerned with things like personnel and liner notes, perhaps even who mastered this recording. Classical fans are similar except I would say even more picky IME (worked in a jazz and classical CD store/venue for a number of years)
I wouldn't call myself a classical music fan, but if we're looking at Spotify stream numbers I wouldn't be surprised if a big chunk of classical is background noise too. Whenever I'm trying to get work done in a noisy environment I go for classical playlists - unless I'm doing something mindless like cleaning I find it very hard to focus with most music. Might just be that lyrics (especially in English) distract me, but classical is an easy go to for studying time IMO.
It might be useful, in both cases, to distinguish "fans" and "consumers" of those styles. Yes, the people who most consciously identify as fans of the styles are picky, but I'd be willing to bet that the vast majority of classical play counts by volume are consumed for muzakish purposes. Mozart for babies, Vivaldi for malls, etc.
The same may be true for Jazz, though in that genre there may be a bit more of a separation between jazzy background mood music and jazz as consumed by fans. There are some overlaps ("The Girl from Ipanema", "Take Five", etc), but they may not be quite as central to the canon as their classical counterparts.
these are fair points, I suppose I was more addressing this with the "classical fan" user in mind given the context of the discussion, idea of a platform that serves these users. It definitely is true though that a huge portion of music listeners in general would rather throw a random playlist on than select an album and it would be interesting to somehow see how many classical listeners on say Spotify are doing so through playlists vs selected albums.
> I suspect focused, attentive listening only makes up a small minority of the total streams.
You haven't justified your assertion of the dichotomy between pure blind consumption and active, focussed listening. You must actually argue that the negation of one implies the other. Your comment is totally meaningless as it stands
> Why would anyone want to listen to it then? what would be the point? The acoustic equivalent of a screensaver?
I would think they would want to listen to it for the same reason they listen to human generated music.
Is your thought here that AI generated art is not art? Is it the nature of the creator that determines the listeners enjoyment? If the music generated by a computer is indistinguishable from human made music, how can this be possible? Moreover, at what point would the computer generation be sufficient to shift it from art to “screensaver”? Would it be one simulated member of an otherwise human orchestra? Would it take 50%. If you can’t draw a line, doesn’t that further indicate it’s irrelevant?
It's way less complicated than that. Music and art in general belong to the social context that produced them. A lot of our appreciation doesn't really come from the technical prowess of the artist or the music theory behind that, but from the cultural hooks we can find in it.
If you want a quick example of what I'm saying, look online for Music for Installations by Brian Eno. It's a "almost-generative-but-not-really" music album made for audiovisual installations. It's... kinda nice but it's 6 hours long and literally gets boring after less than 30 minutes.
I can't imagine listening to something even more abstracted from a human composer for more than 20 minutes. I mean, I'm pretty sure you can generate something even quite pleasant with AIs. Most people will just listen for 15 min, say "uh, cool", and go back to regular music.
Brian Eno was 100% right about ambient music, but he was just wrong about what form it would take. Basically, his music is so unintrusive to the point of being sleep-inducing for many, thus at least a little more intensity is needed. What really became the ambient music was 24/7 lofi hiphop streams. That kind of music is still very abstracted from a human composer, it just happens to have slightly different sonic qualities and sociocultural context than Eno, so it appeals to a wider audience.
EDIT: Should note that Eno's music is still very influential for a lot of people doing work. He composed Neroli after a lot of people doing intellectual work asked him for a new piece geared specifically towards it. Also, Discreet Music was historically used on many maternity wards. In general it is not as if he failed in the popular sphere, Eno is an extremely popular musician
Totally agree with this, but I think it's very likely we'll see AI-written pop hits. They've basically been churned out factory style for years anyway.
And if that does happen, expect a far more exciting backlash movement, like alternative rock in the 90s was to garbage excesses of the 80s.
Many people are so starved of any kind of creative education that they see art in purely a utilitarian point of view. It doesn't matter to them if a human made it, the outcome is the same. To them, yeah pretty much all art is like a screensaver
Relating to it as a creative process. Connecting directly with the artist, mind to mind, through the art. Being liberated from my current perspective. Story telling. Revolutionary mental states. New synergistic paradigms. All the actual reasons why people have been making art for many 1000s of years.
You should not admire artists, but I'm sad for you that art is just a product for you, and not a source of your own expression. Some people naturally gravitate to a subservient, non-creative position.
It already exists, and is known as elevator music or muzak. I can certainly see the value in generating endless copyright-free background music, because so many human effort has been put into composing it already.
Alright, but why would I actually spend time to actively listen to it? Op was specifically talking about Classic music being replaced by AI in listening events, not about staying on hold on a phone calls or waiting on a parking spot.
You wouldn't actively listening, and that's why generated muzak would be okay. "The acoustic equivalent of a screensaver", as someone in the thread called it. The idea that classical performances would be replace human composers in concert context is quite ridiculous. Maybe as a novelty, once, but that's all.
No one with enough bored capital to patronize the arts is going to go to see and be seen at a performance of a synthetic symphony - or not more than once or twice, anyway, for the sake of the thing. Philharmonics need not fear for their concert halls, or not at least for this reason.
Perhaps I should. An infinite future of the Berlin Philharmonic under Karajan, producing performances only slightly more perfectly machined, needs some antidote.
Given that you’re “building the future AI cloud production studio”, it seems you have a business interest in this coming true.
Maybe there’s more to art than business though. I bet there will still be a market for unique and novel interpretations of classical works played by real humans capable of being moved by the compositions they’re playing.
I'd have to disagree. Classical music would be the LAST thing AI could do well. Hard to but technique, nuance, opinion, interpretation, style and the greatest utterances of our troubled civilization into an algorithm. Pop, dance, rap, blaalads....maybe. Classical and Jazz? Never going to happen.
Lmao, classical and jazz are the two simplest forms of music. Without any lyrics to generate its basically just generating some simple patterns in basic instruments.
> classical and jazz are the two simplest forms of music
Uh, have you actually listened to any of it? At all?
There are some non-Western traditional types of music I'd agree can have a level of complexity not usually encountered within the Western classical or jazz genres, but it's fair to say all over forms of Western music are vastly simpler in terms of harmonic language, tonality, form/ structure, instrumentation etc.
None of which I believe would make it harder for AIs to generate, as computers can manage complexity rather well.
What I expect AIs to not be good at is to conjure a truly original and distinct sound world significantly different to anything that's come before, but that still captivates audiences. Which is arguably what the greatest human composers & musicians have generally achieved, in any genre.
The "over" was a typo for "other" (hope that was obvious!). But that primarily refers to the pop/rock/folk genres.
One point I'd agree on is that it will take longer for AI technology to produce a satisfying simulation of the human singing voice than it will for purely instrumental music. In fact despite the leaps and bounds in speech synthesis I've yet to hear any sort of convincing demonstration of synthesized singing. But I can't see why there's any real reason it won't happen sooner or later.
Classical and jazz are actually the primary Western music genres where you encounter some deep music theory.
EG: Read something like George Russell's "The Lydian Chromatic Concept of Tonal Organization" (the basis from which modal jazz sprung, which includes one of the most famous jazz albums of all time, Miles Davis's "Kind of Blue"). Now add to the theory the ability to improvise around it as good as a Miles Davis level jazz performer can. It's not technically easy at all.
Even if other music genres are much technically easier, so much of music is the social experience anyways.
Take punk music. Though some parts got more technical later, much of it (particularly the late 1970s / early 1980s stuff) is, in my opinion, very technically easy; not too challenging to play, with very basic music structure (which was half the point, a return to rock's garage roots).
I'm guessing an AI can probably be developed (especially with today's fairly realistic sounding guitar VSTs) to make some "technically correct" old school punk rock, certainly much easier than it can be programmed to make "technically correct" modal jazz. An AI, however, cannot replicate the human parts, eg the social or community aspects of a music scene. Which with a lot of music is a huge portion of the point (certainly for punk it was).
You probably will be able to get imitation in the not too distant future. But a world where we just listen to imitations of 50s and 60s derivatives of bebop is a sad one. The most loved musicians are ones who are pushing things forward and don't just imitate Trane endlessly or whatever.
AI would need to be able to do something like create The Bad Plus in 1995. That's an even bigger mountain to climb.
We have the tech to be licensed (melodyne or zynaptiq for the polyphonic pitch data extraction), python libs for the analysis. Just needs the brains to execute it at this point...and someone to pay for this kind of compute
I don't think it is completely unreasonable in the future. But I'm personally good with what we've got in this style. We don't need another A Love Supreme. We've already got it. It was fun when Both Directions at Once was discovered, but it isn't like it was something that we needed in 2018.
People will still want to create their sound, and that'll lead to new music over time that isn't just imitation.
I have think you have defined the terms of battle. An improvisational jazz solo from a master vs one that is AI generated. I don't know this but I suspect that Monk had no idea where his solos were going when he started to play them. I like where they went, I'm just saying there were no directions.
It's how commerce works. As soon as AI can generate a commercially viable product, in less time and for less cost, human content ceases to have any commercial value, and as AI becomes ubiquitous it becomes normalized in pop culture. Appreciation comes with the inevitable cycle of one generation being born native to a paradigm shift rejecting the standards of the old guard.
And it won't just be classical, it will be all genres and all creative media. And it will probably take longer than three years. But it will happen.
> As soon as AI can generate a commercially viable product, in less time and for less cost, human content ceases to have any commercial value
Ever been to a craft show? An artist can sell a handmade bowl for $100 even though you can find an identical manufactured one at Target for $1.
Other examples: Chick-Fil-A markets their milkshakes as “hand spun”. Or when Dreamweaver started automating the layout of webpages, web devs started calling their websites “handmade”. It’s been possible to automate call centers for a long time now, but companies advertise the ability to speak to a human representative as a differentiating feature of their business.
There is value in having humans in the loop, despite automation. People just don’t trust robots.
We have machines that can produce clothing, yet there are people out there that buy custom tailored clothing. AI can generate images, yet people will go out there and buy something custom made. I think manufactories will become much more important and expand into more areas. Which means that hand-made music will just be much more expensive, and your "off-the-shelf"-production will be way cheaper.
Exactly - in fact, once AIs/automation are capable of achieving all the necessary production of goods/services for a universally high material standard of living, then I'd imagine choosing to handcraft works of art and share them with others (potentially in exchange for money) will become as popular as it's ever been, if not more so.
We do not have machines that produce clothing. All clothing in the entire world is handmade, which is why developing countries like Bangladesh specialize in it.
The idea that Bach composed the Well tempered clavier and similar works merely in a mechanical process is basically a fairy tale, and it's kind obvious once you read the music sheets.
No harm done. Chalk it up to appreciating the music in near perfect ignorance of its contemporary social context - I had no idea that'd been a controversial point until I went looking for why something I'd thought a harmless joke had crossed people.
Bach's oeuvre is practically inexhaustible and after a while gets to feel a bit samey in places. This led me to make a joke that, judging by a sibling comment of yours, has not landed well among Bach fans, who seem to have a history [1] of being annoyed by such badinage.
Apparently there's something about his work that's seen [2] as uniquely susceptible to automation, and I suppose I can see where the irritation would come from. Doesn't bother me any, but then I've always preferred Beethoven anyway.
I would have disagreed with you 5 years ago but actually I think you could be right. There are adjacent examples of transcendent human experiences being synthesized - eg people falling in love with that Replika chatbot.
I feel grateful that I heard the Mahler symphonies before the great flood of AI content, at least I will always have that. But perhaps AI could complete his 10th symphony…
You could make the argument that charting pop music will be AI generated, or Spotify will be pushing AI generated music through all their popular playlists to avoid paying music royalties - these don't seem to be too far from what happens currently.
But to bet that some of the fussiest and most discerning music listeners would prefer this seems foolish.
Bet $100 that all classical music will be AI generated (synthesized and composed) in just three years. Supply side will be infinite.