I don’t carefully track the numbers, bit it seems like various sectors of the media have been hit pretty hard by layoffs recently. Earlier this month, Disney announced that it was shuttering FiveThirtyEight, the data journalism outlet that Nate Silver started what now feels like a lifetime ago. According to the Wall Street Journal, the site only had a staff of about 15 people, but the move is part of broader consolidation and job cuts at the company.
I didn’t personally spend a lot of time reading FiveThirtyEight, but they were close to my heart for their pioneering work in data journalism, and for their use of interactive data visualizations (which is a subject I’m currently teaching). Along with the New York Times and others, outlets like FiveThirtyEight did a remarkable job bringing visualization to the masses, while also advancing the form in meaningful ways. For a while, it even seemed like there might even be a trend towards fully reproducible data journalism articles, with news media possibly approaching the standards I believe people should strive for in science. That turned out to be a bit of a pipe dream (both for news and for science), but it was still a worthwhile experiment to run.
The other major example that is on my mind is the video games industry, which seems to be in the midst of a major shakeup. Most recently, Warner Bros. announced that they were shutting down three games studios, including one which had recently had a massively successful hit.
There are clearly numerous factors leading to all of this retraction. Certainly we should not discount the degree to which the current administration inexplicably seems hell bent on forcing the US economy into a recession (something which I hope will contribute to an increasingly disgruntled electorate). There is also the ongoing fracturing of the attention economy broadly, with some of our mindshare now being captured by podcasts and streaming and tiktok and whatever else I’m not even aware of yet. The main explanation for the layoffs that is often given, however, is recent advances in AI, and the potential for that to augment or displace humans in the production of content.
This is of course nothing new. The Associated Press, for example, has been using various forms of technology that would now be called AI since at least 2014, for tasks like summarizing financial earnings reports and language translation. Making video games rather obviously involves many different types of software, and so has inherently involved some form of automation since their inception. In my mind, there is a strong continuum here, with recent developments fitting in nicely with a gradually shifting technological landscape. Use cases of AI like generating concept art are already happening, and clearly allow for creating things with less human effort. My guess, however, is that the layoffs we are seeing have less to do with an expectation that AI will be a wholesale replacement for human labor, and more to do with worries about it becoming increasingly hard for anyone to profitably compete in the space of selling news and entertainment.
As I’ve discussed before, in some sectors, automation can actually lead to an increase in demand for labor, (at least in the short run), but this seems unlikely to hold true in media. Unlike some other sectors, media is fundamentally limited by attention. There are only so many hours in the day, and we are already feeling the strain of trying to keep up with everything we would like to care about.
Games in particular are interesting, in that the lifespan of a game can be much much longer than other forms of media. News articles are typically consumed immediately and mostly fade within a few days. Movies may experience a few waves (e.g., first in theaters and then on streaming), and a few classics will bear many repeated views, but most quickly fade into relative obscurity. The games industry, by contrast, has a considerable number of massively successful creations that have been actively maintained for years, and are still going strong. There is of course huge competition to try to become one of these games, with some recent and notable failures. But the existence of enduring game communities means that new entrants to the market are not just competing with other new releases, but also with long-running alternatives.
As I noted in my earlier post, it seems very likely that we will get lots of successful low cost replacements for certain forms of background entertainment, especially music.1 But those are somewhat different from things that we engage with more intentionally. I do feel confident that some media companies will figure out ways to both incorporate AI into their production processes and stay ahead in terms of being market leaders. My general expectation, however, is that as more and more companies adopt these tools, the more everything will start to feel the same, such that it will be harder for a company to justify selling their own particular branded version of a thing.
Unfortunately, an escalation of the use of AI into media production does not necessarily mean we are entering an age of breathtakingly impressive media output. Rather, I expect we’ll enter an age of an abundance of “good enough”. Lots and lots of movies that are fine. An infinite supply of games that are all kind of samey. The ability to generate our own versions of things that are maybe not great, but just about as good as we can get anywhere else. For that matter, why would I pay for a professional summary of recent events if they are mostly just using a model that I can prompt myself?
The counterargument is that by making it easier and cheaper to produce media, we will see a profusion of creators catering to niche markets, and an increased return to having a truly distinctive authorial style or voice, with findability solved by even better recommendation systems. Freely available software will enable small teams to make award winning creations, like Flow, with small teams and relatively little money. I do think we’ll continue see some version of these projections, but it also does not sound like a recipe for anything like a stable career for most artists, which limits how much creativity and effort people will be able and willing to invest. It also remains to be seen when we’ll get more examples of people pushing the limits of what is possible with recent advances in AI.
As my friend Ari Holtzman recently wrote, the kinds of things that people have been so far trying to build with AI have been overall disappointing. It’s somewhat unclear at this point whether this is more due to technical hurdles or people’s lack of imagination. I have a theory that because LLMs have the most dramatic effect on making it much easier to do things that we already know how to do, those kinds of use cases overwhelm any attempts to put them towards more unforeseen applications. It could be that we’ll see something totally revolutionary come out of how a news or game company tries using AI. But before then, and in much greater abundance, we’ll see a massive increase in familiar and uniform output, in a way that will likely further bury more interesting possibilities.
In other words, it may be that companies are being totally rational in laying off employees, even from profitable divisions. But I suspect it’s not because we’ll soon have AI producing amazing and scintillating content, without the need for much human labor. Rather, it’s because we’ll soon be suffused with AI generated slop, in which no one can quite justify paying to produce an actually impressive and unique version of a thing, when the cost of even distinguishing oneself from the ambient background noise becomes just too high.
My hope is that we’ll still have plenty of people making incredible art, perhaps only for their own satisfaction and inherent need to create. But my guess is that overall we’re headed to a world in which there is an even clearer and stronger distinction between art and content, especially for information and entertainment spheres like news and games.