Infinitely Wide Culture

A lot of this is still unresolved in my mind, but I think there is something interesting happening at the intersection of generative AI, art, style, and entertainment. Rather than letting it gestate until more fully formed, I figured I’d just post some preliminary thoughts and come back to this at some later date.

The main reason I’m thinking about this now is the ongoing debate about how generative AI will impact creative fields, such as writing and design (as well as white collar jobs more broadly). Arguably there have already been some pretty dramatic effects, such as the sci-fi magazine Clarkesworld being suddenly overwhelmed by spammy submissions. At the same time, it’s hard to know the extent to which these disruptions may end up being transient phenomena that broader systems will adapt to.

What interests me, however, is what this all says about the different functions played by art, considered broadly. On the one hand, artistic work can be a highly personal expression, arising more from an irrepressible urge to create, rather than something more instrumental. On the other hand, artistic production is inherently tied to markets, with enormous demand for certain types of products, like music, movies, and games. Obviously this is a continuum, with every individual creator navigating their own personal path between pure artistic freedom and serving the market. In addition, while many artists walk many paths at the same time, weaving together a diverse set of income streams from sources such as direct sales, teaching, contract work, and other jobs, there also exist relatively professionalized forms of certain types of labour that overlap strongly with artistic production, such as graphic design or game development.1

The big question at the moment is how and to what extent generative AI will disrupt these fields even more. Already there is some evidence of a short term impact on labour markets, at least in terms of freelance job opportunities on a particular online marketplace studied in a recent paper. Almost certainly the long term overall effect will be complicated, with generative AI being incorporated into work processes for people in many fields, including coding, writing, image making, music production, etc. (not to mention places where the impact seems harder to predict like education and health care). Although we might expect a short term replacement of labour, there are many historical examples of automation actually having the opposite effect, due to a corresponding increase in demand due to a lowering of costs, at least in the medium term.2 For arts and entertainment, however, things seem somewhat bottlenecked by the fact that people’s potential consumption is limited by our available waking hours.3

Already it has long been impossible to keep up with all of the information and entertainment that is available, even if we narrow in on somewhat niche interests. For certain types of products, like video games, it often seems like the problem is that games are too long, rather than too short, and the actual cost per hour of entertainment can end up being extremely low (if one wants to think in those terms). Indeed, the amount of information and entertainment that is available completely for free is now unimaginably vast. If one were somewhat less particular, there would be no need to ever rent or purchase another book, movie, album, or video game, ever again, given how much can be freely and legally accessed online, without even considering various pirated options.

Naturally, however, we don’t want just to read or watch or listen to anything. Our consumption patterns are driven by our interests and tastes, not to mention a desire for novelty, and wanting to be part of the current conversation. The prevailing sentiment seems to be that most streaming services now have vast selections, but abysmal quality, on average, meaning that there is very little that is actually worth watching. And yet, the desire to stay up to date with whatever is au courant keeps people paying their monthly fee.

At the same time, tastes seem to be changing (or perhaps not changing) in interesting ways. There is a compelling narrative that nostalgia now reins, whether we look at the enduring popularity of music from the 20th century, the average age of male movie stars, the rise of pixel art styles reminiscent of classic games, or the nostalgic themes that drive so many new shows. Whether it is familiarity or quality or manipulation, something seems to keep a lot of our attention focused on things that by this point seem like they have always been with us.

In some ways, I think that music might be the cleanest example for thinking about how the impacts of generative AI might play out. Indeed, more than most other popular art forms, music is something that we not only can, but often want to, have as an accompaniment to other activities. While reading, playing games, or watching films demand (or at least request) something close to our full attention (although admittedly some people seem to be surprisingly comfortable with multitasking in their consumption of entertainment), music is both easy and pleasant to listen to while doing other things.4 While there is tremendous reward in forcing ourselves to put everything else aside, and devote our full attention to listening, there’s also nothing wrong with having it as a background accompaniment.

In part because of this, there is clearly a demand for music of a certain type that doesn’t require a deep investment in quality, history, or narrative arc. It’s extremely easy to imagine that we will see a rise in the popularity of generative music deployed in public settings, or perhaps private ones. Obviously there is a long history of generative music, one that I hope to write about some other time, but at least at the present moment, most businesses seem to still be willing to pay for authentically human produced music, as do people in their own streaming preferences. But once it becomes widely available, I fully expect to at some point hear an endlessly shifting stream of generative classical music or jazz or whatever genre in some public setting. Some people will also no doubt subscribe to services which provide generative “music to study to” or similar, much like the genre of video that is already extremely popular.

The big question is how this will intersect with people’s preferences, in terms of our nostalgic attachments, our desire to affiliate ourselves with particular creators, and our need to be current. One can certainly point to examples of purely synthetic musical figures that are nevertheless popular, at least within certain cultural domains, but my guess it that most people still tend to connect with art in part because of the human stories involved in its production. People don’t just like Taylor Swift’s music, they are intensely wrapt up in her whole narrative. Similarly in different domains, such as painting or film, part of the delight of appreciating art is obsessing over the particularities of a specific individual person, such as director or painter, wanting to know everything about them, and trying to understand how and why they did what they did.

On the nostalgia question, an obvious possibility might be that we end up with a lot of generative music that sounds like it could have been recorded by a particular artist, like the Beatles or CCR or whoever. This will likely end up being a primarily legal question, as I’m quite sure that the market demand is there. In terms of newness and staying current, it’s somewhat unclear what “new” even means, when music can be infinitely randomly generated. For any particular model, there is basically an infinite catalogue that could be produced from different random seeds, and hence those songs are already in some sense lurking below the surface, ready to be called into being. Rather than an endless progression of gradual production over time, we might end up with infinitely wide slices of culture at any point in time, ready to be explored, discovered, and shared by devotees.

Although each market has its differences, it feels like similar principles apply everywhere. Various cultural products claim to be the first book written by AI, or the first film produced by AI, and so forth, and these mostly tend to be pretty boring, succeeding (to the extent they do) only on the basis of a supposed claim to novelty. As that wears away, perhaps we will see things like generative romances or generative sci fi or other such products fulfilling certain market demands, whether in books, or movies, or games, or some vast continuum that merges all media together.5 One can also imagine a more personalized form of this, where novels no longer exist as the same type of shared cultural product, but are rather more like infinitely available on-demand text that has been customized to your particular interests (or at least claims to be). At the same time, I strongly suspect we will continue to see clear human individuality in all of these markets, both because of people’s inherent desire to create, and due to our love of affiliating and obsessing over the work of others.

  1. For an excellent book on how some artists manage to make a living today, see William Deresiewicz’s The Death of the Artist↩︎

  2. The classic example of this is bank tellers, which actually became a more common job class (with different responsibilities) following the introduction of ATMS, because the decreased cost of running a branch led banks to compete on number of branches. ↩︎

  3. In an oft-quoted remark, Reed Hastings noted that Netflix’s biggest competitor was sleep, something which is quite dramatically no longer the case↩︎

  4. For some thoughts on combining media and commentary in different formats, see also my post on tightly-woven cultural commentary↩︎

  5. On the devotees of particular media styles, it would be worthwhile to connect some of these ideas to what Ursula K. Le Guin has written about literature and genre fiction↩︎