Dead Substack Theory is coming ...
Thoughts arising from the Velvet Sundown and AI content in general
In case you didn't know already, a popular band on Spotify is confirmed to be synthetic. The Velvet Sundown was described as a 'provocation' by its creator/s and people have certainly been annoyed about it.
I wasn't annoyed. I thought it was funny. Especially the pictures.
It all reminded me of the K-Hole Trends stunt, which I loved.
The episode raises some ethical issues, such as robots taking a slice of streaming revenue while human artists struggle to earn a living. But I find that hard to feel mad about. For mid-level and independent artists and bands especially live performance is a more reliable source of revenue than streaming, downloads and physical format sales. If we concern ourselves with other people liking what they like, even if we think it's rubbish, I wonder what's really going on with that.
Still, I'm guessing that artists feel threatened by technology capable of producing a simulacrum of music plausible enough to gain time-limited consumers' attention.
Paradoxically it's because music has always been core to my life that I didn't care very much about this story.
The music of the Velvet Sundown meandered past me in the same way that AI-authored Substack essays do, with their uncanny valley quality of connecting predicted text strings in a way that feels hollow, despite executing all the correct rules.
The writer Will Storr recently came up with 'the Impersonal Universal' as the expressive AI mode in written form.
"There is a white-noise generality to its insights, an uncanny vagueness that makes the mind glaze over. It is never funny. There are no surprises, true confessions or controversial moments. It is a description of the human average."
That's what The Velvet Sundown sounded like. But not uniquely so. It sounded the way that Sabrina Carpenter's tunes do, produced as they are by teams of professional writers who know what tickles the auditory parts of the temporal lobe, prefrontal cortex, limbic system and so on.
The Velvet Sundown is on a continuum I've known since childhood, when hit singles were often the work of someone other than the people you saw on the record sleeve.
The Wombling Song wasn't performed by actual Wombles, guys.
Yes, that's flippant, but actual fakery isn't new in music and the Velvet Sundown merely told us that people will like things, however they originated. We knew this already.
I do understand what irks people about 'content' produced by agentic AI. Sharing those feelings is why I don't use it for writing, or even editing. My output doesn't appeal to a mass audience and I struggle to make ends meet in the real world, but I stick to high-effort output that's homeopathically profitable because I'd feel ashamed to outsource my words in the way that some of Substack's biggest sellers do.
Maybe it's a hard-wired human tendency to appreciate high-effort work by entertainers and artists.
Appreciation of effort is probably one of the reasons why people get annoyed about abstract art, like this.1
But I wonder if half the fuss around 'AI slop' is partly motivated by jealousy, because many 'writers' on this platform are achieving high status and serious income for doing next to no writing.
Some of the big names with genuine talent are increasingly bold in suggesting that the success of these people is undeserved, because they're prompters and editors at most.
But I'm sceptical.
'It is not our part to master all the tides of the world, but to do what is in us for the succour of those years wherein we are set - Gandalf/J.R.R. Tolkien (The Return of the King)
All this maundering about AI reminds me of a point often made in Dan Williams's discussion of cognitive biases. He observes that when we talk about cognitive biases we're solely concerned with other people's cognitive biases.
In this case I'm thinking that the underlying frustration is that other people aren't seeing what the rest of us can. Or that they don't care that little effort go into 3,000 words on their favourite topic, just because it makes the right noises. Every day.
It's only human to think that we're above the weaknesses we see in others. When I look at the proliferation of fawning comments under essays that someone obviously got a machine to write I feel disdain for the low sophistication of those readers who are so moved by it. But also a certain reluctant admiration for the person who got them to pay them for it. But my scorn is still mainly reserved for the consumer. It's the same contempt I feel when anyone gravitates toward the lowest common denominator in their choices of music or literature.
And yet, objectively, it's none of my business what other people listen to and read. Or what others publish. So it's really none of my business that someone else thinks an essentially vacuous string of successfully predicted tokens is profoundly meaningful.
Yes, when it comes to the creators it's cynical to pretend that you're doing something you're not, but if people enjoy it, what then? Ban it? Impose our choices instead?
There's a trend on YouTube for videos like 'At 79 Jimmy Page revealed what he really thought of Eddie Van Halen' and they're obviously synthetic machine stitch jobs. There are no insights to be gleaned, but they attract lots of positive engagement. After being disappointed by 2 or 3 or these I learned to spot the signs and avoid them.
Maybe a badge on things that were synthetically created? I can see a case for that, so that the rest of us don't waste our time starting to read before realising that it's low-effort gruel.
The current trend of hand-wringing over the blight of algorithmically probabilistic music, writing or art begs the question of why any of it really matters. Our assumption is that it does. I just called it a ‘blight’. But shouldn't we be showing our working out when we assert this?
Dissecting my personal reaction to AI-derived entertainment (which includes 'serious' essays) reveals an arbitrary quality to my objections. I would never actively choose to listen to The Velvet Sundown because the music isn't interesting or affective for me, rather than because it's just some code producing auditory stimulation. If the music was good I'd feel differently.
I know this because I already enjoy 'machine music' and I do it myself. Generative and evolving ambient electronic music relies heavily on setting some parameters - including probabilistic variables - stepping aside and leaving the music to essentially make itself. I sometimes pay for music that I know has been created in this way.
This is hardly a million miles away from prompting Suno with text strings, although it does involve some prior knowledge and effort invested into learning some basic musical principles. Whereas Suno promises that anyone can make any music they can 'dream' of. I don't really know what to make of this difference.
So I'm pondering whether or not there's a 'gatekeeping' impulse in play, rather like there is when a certain class of person seeks to define misinformation on behalf of everyone else.
An old proverb (famously cited by Anaïs Nin, but apparently originating in the Talmud) says
“We do not see things as they are, we see them as we are.”
I suspect that lamenting the rise of popular machine-written text, visuals and audio, is as much an expression of personal frustration as it is a moral statement about the world. Us biological struggling writers are seeing our efforts eclipsed by effortless attention hijacking and it feels unfair.
Like many moral impulses, it's less noble than you think when you drill into it.
And yet I still share in the unease that this content produces in others. My personal angle on it is that machines cannot mean anything they say and this drastically reduces the value of their outputs.
It also contributes to a miasma of unreality that I keep wondering about, like here and here.
A few years ago a popular conspiracy-adjacent fringe hypothesis that gained some notoriety was the Dead Internet Theory. In short, this suggested that most content online - including human engagement - is synthesised.
I can see a Dead Substack Theory growing from this, as more blogs with serious scale turn out to be authored by no one really.
Then the bots will pretend to write about that.
There is no happy ending other than to ramp up your AI discernment skills.
— This essay isn’t just a sincere expression of enduring humanity - it’s epoch defining. Let me know if you’d like me to write another version tailored to a particular audience—e.g. journalistic, corporate, political, artistic or as a piece for your Substack blog Rarely Certain —
(Joking !!!)
It's a while since I recommended another blog here and all this talk of bot life has certainly made me appreciate some of the smaller-scale humans I enjoy reading.
One is
, who has introduced me to someone from ancient Greece, some of whose thoughts about the world I never knew I shared until recently.Pyrrho of Elis is now a personal hero and Doug's blog has been the catalyst for this.
Ataraxia looks quite like the principle I've been trying to grind out in Rarely Certain connecting epistemic humility with subjective wellbeing.
What's not to love.
It doesn't get much lower effort than those bricks. But another reason why people get annoyed about abstract art is that it's an elite status marker. However 'uneducated' or incapable of articulately critiquing some bricks laid onto the floor of a national gallery space the little people may be, they know when someone is provoking them, taking the piss and then looking down on them.
There's a part of me that wants to insist that the writing is the writing, or the music is the music, no matter who or what is creating it. If AI produces slop that a large enough subset of people like, then chances are those people would have liked slop created by real people too (god knows there's enough of it, and I've contributed to that slop myself). So if you like the article, does it matter where it came from? If you don't like the article....does it matter where it came from?
There's another part of me that wants to say that the AI slop is different, it's not "just any other writing"; it's dangerous because it combines some things at least somewhat rare otherwise: pretty solid "mechanical" skills (the prose won't set the soul on fire, but it's perfectly serviceable; in fact better than much of even professional prose), speaking primarily in cliches and well-trod patterns, which makes it, in many ways, easier to read, and it has an excellent capability to distill a bunch of ideas from the massive pile of all ideas relevant to a subject into a digestible form. In short: it summarizes ideas, arguments -- relevances -- very well. This means it's likely to drown out anything that's more creative, like corn becoming ubiquitous in our food supply, replacing many other ingredients, because of its abundance and low cost.