Feed experiences

all generative content is generative ads

Recently Hearing Things interviewed a musician who, under the name Catbreath, posts songs to Spotify with such titles as “True Crime Podcast,” “Cozy Fall Playlist,” and “Gym Bangers.” As the interviewer, Andy Cush, points out, this is meant to trick plays out of distracted Spotify users, some of whom may be operating their accounts with voice commands, so “Siri, Play my liked songs on Spotify” may yield Catbreath’s “My Liked Songs.” It’s clever in an Adbusters sort of way and I can see wanting to laugh with him at the sort of people who fall harmlessly into his little traps. He also seems to make more than I make a month writing this newsletter, so clearly my content-creation strategy is all wrong.

After listening to some of Catbreath’s music, Cush came to the conclusion that he was “less like a scammy opportunist and more like some sort of minor folk hero wielding a bong and a scuffed-up guitar” and frames his intervention as “an effort to disrupt, in some small but real way, the seamless surfaces of a streaming system that exploits musicians and encourages listeners to think of their labor as worthless.” It would be especially great if the songs were Gang of Four–like metacommentaries on their own form, with lyrics that critique the culture industry or the instrumentalist use and commercialization of art, that interrogate the idea of why “gym bangers” have to exist under our socio-technical conditions. But Cush reports that most of the songs are “about getting fucked up and feeling dissatisfied with society’s unfairness” — perhaps those dialectics won’t break bricks.

I assume Spotify does what it can to purge the Catbreaths of the world from their catalog, but who knows? Possibly hackers like Catbreath are doing the arbitrageur’s work of identifying exploitable weaknesses or inefficiencies in a system which can then be systematically eradicated. Perhaps Spotify is in a monopolistic position in which it doesn’t have to care if people are tricked into their play counts. Or perhaps Catbreath is contributing to the mounting overall failure of online search as we know it and helping usher in an era where automated systems make even more decisions for us without our even attempting to have an input.

Catbreath urges more people to post fake songs and undermine the possibility of search as if this would sort of protest would lead to a more equitable distribution of attention and thus payment for musicians. “Streaming is messed up,” as he says, but not necessarily because the most famous artists make most of the money, as he frames it in this interview. It would be a more coherent critique if the point of Catbreath’s actions were to show that anything distributed on Spotify results in an alienated, arbitrary experience, and that Spotify doesn’t care about music at all but brokering audiences to advertisers, like all other media companies. But Catbreath’s intervention instead seems to target users, who, instead of being taught a lesson about convenience’s place in music appreciation, may decide that “searching” or “asking for what they want” is inherently inconvenient and will give over even more trust and agency to the tech company to fix this for them.

Would it be so surprising if, in the near future, people are so deskilled in the practices of aesthetic investment that they will have forgotten that it was once possible and desirable to search for something they wanted before their AI “assistants” told them what to want? (See what happens when you ask for what you want? You get Catbreath in your face. Don’t be so presumptuous. Leave the heavy thinking to the machines — just lean back and enjoy what’s in store.) Spotify has more to offer advertisers if Spotify users are in the habit of just listening to whatever the company decides is best for them (i.e. what is best for Spotify’s bottom line).

Spotify probably wishes all the artists it had to deal with were Catbreaths — interchangeable performers who supply a product that fills logistically determined product niches — rather than established musicians with leverage of their own. And then even better, replace the Catbreaths with generative models that process user behavior and pre-emptively supply machine-made content that will placate them. This would ideally divert listening practices away from its social use of organizing communities of like-minded users sharing an aesthetic experience, and away from its psychic use as a means for imaginative projection, vicarious participation, and aspirational orientation. Instead “music” could be for harmonizing a user with an app that performs like a mood ring, revealing to the individual to themselves and providing the most convenient, expedient means of managing that unfortunate condition.

That is to say, Spotify could take the route that Meta has taken, from supporting (and then cannibalizing) users’ social ties, to using algorithmic feeds to incentivize humans to produce content that could negate and replace social connection, to having machines supply individual users with optimized asocial content to close them off in a perfectly solipsistic loop. (From “Music is my subculture” to “searching for mood music” to “the app pipes in Muzak to stop me from thinking about wanting to hear music.”)

Tech analyst Ben Thompson argues here that Meta is the company best situated to profit from generative AI long term because of “the obvious benefit of applying generative AI to advertising.” He can hardly contain his glee at discovering that “generative ads and generative content; they’re the same thing!” Yes, both generative models and advertising deprecate subjective agency. Thompson writes:

Meta will work across their vast troves of data in a way that is only possible using machine learning-derived algorithms to find the right targets for an ad and deliver exactly the business goals requested.

That is euphemistic language to describe automating the process of finding human vulnerabilities (the “targets”) at scale and making the kinds of manipulative content optimized to take advantage of them. It can generate as many ads as necessary — ads made spontaneously for individual users to exploit whatever vulnerabilities their tracked behavior has revealed — and show them to as many people as necessary until the desired number of “conversions” are met: “advertisers can provide Meta with broad parameters and brand guidelines, and let the black box not just test out a few pieces of creative, but an effectively unlimited amount.”

The long-held dream of an infinite amount of advertising suffusing the world has never seemed so close — enough ads to make us forget to even imagine any other kind of sensory input. Why should our eyes look or our ears hear something that cant’t be measured in terms of commodity exchanges? Why should sense perception be exempt from profitability? Every piece of surveillance can be used to create an ad for someone before they have a chance to even think about looking at something else. They can then be flattered by this bespoke attention and benefit from the luxury of not having to ideate interests of their own. Who doesn’t prefer catalogs to magazines? What are pop songs but advertisements for themselves? What is participating in pop culture but deciding which ads most speak to who you really are?

Thompson notes that AI content is more expensive to the company than the free content its users supply, but the free stuff doesn’t sustain enough engagement to allow for probabilistic ad targeting. At the same time, Thompson writes, “the real impact of AI is to make everything inventory.” It can analyze “authentic content” and re-create it as advertising, just as Facebook promises to re-generate ads and put your face in them.

At 404 Media, Jason Koebler notes Meta’s recent move toward more AI content in feeds and more generative ads. Mark Zuckerberg’s comments on an earnings call made the strategy behind this plain:

If you look at the big trends in Feeds over the history of the company, it started off as friends, right? So all the updates that were in there were basically from your friends posting things. And then we went into this era where we added in creator content too, where now a very large percent of the content on Instagram and Facebook is not from your friends. It may not even be from people that you’re following directly. It could just be recommended content from creators that we can algorithmically determine is going to be interesting and engaging and valuable to you.

After that, there is an “opportunity for AI to help people create content that just makes people’s feed experiences better.” The pipeline of content goes from friends to “creators” to generative models. It is understood that algorithmic feeds can compel people down this path, teaching them that it is more convenient to prefer the kinds of satisfaction that the feeds can consistently supply and that distraction can be understood as a kind of snacking on content that is more fun than the work that goes into maintaining social niceties or researching specific topics. Why have life experiences when you can have better “feed experiences”? What more would you want out of life anyway than steady feeding? Play me my most liked songs.