Scrutable

There was a post in a drum forum I'm in about AI in music that I wrote a response to. Given the interest in AI from some in this group I thought I'd share it here. I think the point can be made about other creative areas.

"I’ll start with the caveat, because it’s important. I’m impressed by what AI can do, but I’m also genuinely worried about what it means for musicians. None of this comes from a place of entirely cheering it on.

But I do think there’s a lot of complacency in the arguments against it. So many of them focus on what AI supposedly can’t do. It can’t capture feel. It can’t be creative. It can’t sound like a real player. I just don’t buy that. I think those claims are going to age badly, and probably within the next decade if not much sooner. At least in a recording context. Live performance is a different conversation.

Saying “AI won’t ever be able to do X” is a dead end.

Even on creativity, AI will get there in two ways. One, it’ll eventually be able to generate new styles on its own. Two, people will use it as a tool to create new things themselves. That’s exactly what happened with synthesizers and electronic music. People said they’d kill creativity. Instead, they ended up creating entire genres. You don’t have to like electronic music to admit that it massively widened what was possible. If you’re making the same argument about AI that people made about synths 50 years ago, odds are you’re going to be wrong.

And honestly, most of us musicians aren’t as wildly original as we think. A lot of what we do is basically pattern-matching. We absorb stuff we like, tweak it a bit, and call it “influence.” In that sense, plenty of us behave a lot like an AI already.

Which is why I think we need to separate out two things: what AI can produce, and what impact it will have. Those are not the same. If we want to make meaningful arguments against it, they need to be about the impact on musicians, on the industry, on society. That’s the part people need to care about. And that’s the part that’s much harder to get across, because most people just don’t/won’t care.

So yeah, bit of a downer. But if your argument is “AI won’t be able to do X, Y, or Z,” instead of “AI will have serious consequences and here’s why that matters,” then you’re setting yourself up to lose."

I have read that music channels are being filled with AI slop. Moreover, people with an eye for an easy buck were uploading AI music onto public media, and then getting bots to "listen" to it to, so earning money from nothing. I read that some social media are starting to chuck out AI slop on a large scale, at least in part for this reason.

IvanV wrote: Tue Nov 11, 2025 12:04 pm I have read that music channels are being filled with AI slop.

My argument is that within 10 years, and most likely well before that, it won't be slop. It will have the capability to be indistinguishable from human generated music and will likely also be it's "own thing" creatively too.

Whether that's a good thing or not is a different argument. But the capability will be there.

and will likely also be it's "own thing" creatively too.

Whether that's a good thing or not is a different argument. But the capability will be there.

I think I agree with you that indistinguishable is close if not already here - at least with books. There's a an author who tests this over the last few years (can't remember who though?!) and for short stories with a starting prompt AI is already very close. It's actually a difficult challenge for writers.

However that's not being 'pwn thing' creatively and I can't see how they could do that. Fill in the gaps between what's already been created using other patterns in as yet un-copied ways, yes sure. Find a new pattern/idea/concept no, simply doesn't not have the tools to do that, and doesn't even have the means to make those tools, or even be 'aware' that such tools could be made.

science_fox wrote: Tue Nov 11, 2025 3:50 pm

and will likely also be it's "own thing" creatively too.

Whether that's a good thing or not is a different argument. But the capability will be there.

I think I agree with you that indistinguishable is close if not already here - at least with books. There's a an author who tests this over the last few years (can't remember who though?!) and for short stories with a starting prompt AI is already very close. It's actually a difficult challenge for writers.

However that's not being 'pwn thing' creatively and I can't see how they could do that. Fill in the gaps between what's already been created using other patterns in as yet un-copied ways, yes sure. Find a new pattern/idea/concept no, simply doesn't not have the tools to do that, and doesn't even have the means to make those tools, or even be 'aware' that such tools could be made.

Here's how i think AI could end up being it's own thing musically, with some initial set up. Within a 10 year timeframe there's little reason that you couldn't have an AI system that is trained on all known music that then:

- Has a novelty-seeking objective
- Generates 100 new tracks a day
- Clusters them into emergent styles
- Publishes the most promising ones to a platform
- Monitors listener feedback to refine its direction

So there is human direction but it isn't musically creative at all.

- The person directing it is just telling it "create something new" - similar to a record label exec.
- The person who created the AI wasn't doing anything musically creative either, they were a programmer - similar to the sythesizer builder.
- The person who set up the platform that it releases to isn't doing anything musically creative, they were a businessman - similar to the guy who founded
Spotify.
- The listeners providing feedback aren't doing anything musically creative, they're just listeners - similar to listeners who decided whether a song made it into the top 40 or not.

So there's human input at all stages, from building the infrastructure, building the system, turning it on, to consuming it. But none of that input is truly "creative" for any meaningful sense of the word in a musical context.

science_fox wrote: Tue Nov 11, 2025 3:50 pm

and will likely also be it's "own thing" creatively too.

Whether that's a good thing or not is a different argument. But the capability will be there.

I think I agree with you that indistinguishable is close if not already here - at least with books. There's a an author who tests this over the last few years (can't remember who though?!) and for short stories with a starting prompt AI is already very close. It's actually a difficult challenge for writers.

However that's not being 'pwn thing' creatively and I can't see how they could do that. Fill in the gaps between what's already been created using other patterns in as yet un-copied ways, yes sure. Find a new pattern/idea/concept no, simply doesn't not have the tools to do that, and doesn't even have the means to make those tools, or even be 'aware' that such tools could be made.

Last year, I tried to get AI to turn my rough notes into good, compact prose. Because I'd been told it was good at that. I was told that the output was crap, in particular by the person who suggested I do it. In particular, it "read like AI". I've now been told (these suggestions not available a year ago) that I should give it a sample text in good style, and ask it to imitate that style in writing up my notes. It has also been suggested that I ask it for three different drafts, to try and get a good one.

So it seems to me that if AI writes good prose, then it must have some kind of a model for that, because it doesn't do it naturally without a model. It will need some help. It isn't going to be the next [Author whose distinctive new style you liked] or [next author whose distinctive next style you liked]. You can tell it to write like [Author whose distinctive style you like]. Or maybe you can try to get to try and synthesize multiple styles, but I rather suspect the outcome of that would be a bland mess. Like if you mix too many paints you end up with grey-brown.

But there are many ways of writing a short story. Maybe you intuit something from giving it some opening lines and hints of a plot. It will doubtless finish an incomplete plot in a predictable way. Though I suppose you can ask it for a plot twist or some surprises. And if you want a short story in the style of Saki, or whatever, you can give it some samples and it will do that. But do you want that?

A lot of us are happy with some fairly predictable entertainment of a particular style, if competently executed. If you have enjoyed reading a thriller by Lee Child, you might feel in the mood for that and want to read another one. I bet AI could produce some Lee Child type books, if you give it a bit of prompting some initial outline of a plot. Probably his style is quite easy to imitate. Probably his plot types are quite easy to construct from some initial hints.

Some people have already discovered this. We recently had the unedifying sight of a rather gleeful Alexander Boris de Pfeffel Johnson the other day seemingly having discovered that he could solve his problem of the unfinished book commissions - typically histories - by getting AI to help him out. It knows the facts of history better than him. It can reproduce his style. He can spend a couple of mornings going over the draft adding in a few points of his own, probably rubbish, but the kind of stuff his readers might want. Get AI to repolish and make consistent. And bish bash bosh, new Johnson book on the Boche, or whatever other bish-bash was required, quickly churned out.

I used to comment on the similarity of L Ron Hubbard to St Paul that both succeeded in authoring books after their death. But maybe now many more people can achieve this. Perhaps AI can finish off Dickens' incomplete Edwin Drood rather better than the various attempts by humans. Perhaps AI can revise and improve the later Pratchett books. But I doubt it can be the next Dickens, or next Pratchett. Or not just yet.

IvanV wrote: Tue Nov 11, 2025 5:52 pmOr not just yet.

THIS!

My frustration with the "it'll never be able to..." arguments are that they're likely wrong in the long term and in some cases they're already wrong.

10 years is a long time and if you look at the progress in a very short space of time the progress has been rapid.

But also, a lot of people critiquing it haven't kept up and actually listened to what it can currently do. Often they're saying "It'll never do" something that it already does.

Rick Beato creates AI music (youtube).

I mean, he's made videos about complicated music theory but then he also co-wrote Carolina (youtube) which frankly sounds like the sort of thing a lazy AI prompt would generate now.

ETA so I generally agree with the OP. Most music is already fairly derivative and forgettable. But AI won't replace the rewarding experience of me playing with a band, for example.

I’m terrified of what AI will do to all creative industries but I must admit I was very impressed with the NSFW parody songs it can do. Some of these are brilliant.

All NSFW

https://www.instagram.com/reel/DQpAAT3ggKI/

Stranger Mouse wrote: Wed Nov 12, 2025 11:17 am I’m terrified of what AI will do to all creative industries but I must admit I was very impressed with the NSFW parody songs it can do. Some of these are brilliant.

All NSFW

https://www.instagram.com/reel/DQpAAT3ggKI/

I posted my post somehwere else and had a very similar reply but about the AI recreations of songs in different genres (eg, a classic Slayer song done as 70's soul etc). Basically, "Yes I think AI is terrible, but this specific thing is quite fun isn't it?".

I wonder if that will become part of the normalisation of AI in music.

Tristan wrote: Tue Nov 11, 2025 11:21 am That’s exactly what happened with synthesizers and electronic music. People said they’d kill creativity. Instead, they ended up creating entire genres. You don’t have to like electronic music to admit that it massively widened what was possible. If you’re making the same argument about AI that people made about synths 50 years ago, odds are you’re going to be wrong.

Fair observation, I think - for all that digital synthesis has probably killed off quite a lot of session musicians’ jobs, there still are some, and there are plenty of people who still want to make and hear non-electronic music.

Also, without dismissing the real issues that digital distribution and streaming have created for anyone trying to make a living out of creating music, what to do in response to AI is probably going to be similar to what to do in response to digital distribution and streaming: focus on building a relationship with an audience through doing stuff on top of just releasing recordings, which might be live performance but also can include the social media type stuff. AI might soon be able to sound exactly like human made music but I’m holding on to the idea that people will still want a real person involved in order to turn out to a gig.