Criticism of Spotify has traditionally focused on its poor economics for artists and its flattening of music taste (read Liz Pelly’s excellent Baffler series for the best of both categories). In the past month, however, the company has opened up two new fronts against its haters, rolling out an AI DJ feature and then, just a couple of weeks later, unveiling plans to TikTokify its interface in the form of a vertically-scrolling video “discovery feed.” Through those two announcements, Spotify managed to fatuously connect with both of the dominant trends in consumer tech, AI and TikTok: The DJ feature essentially rebrands what Spotify’s algorithmic shuffle already does, adding “computer-generated commentary on what makes those songs great,” which sounds pretty annoying. The TikTok-like discovery feed, on the other hand, is a major change, albeit one that nobody asked for—a seemingly desperate pivot oriented toward squeezing more value out of Spotify’s users in a hostile economic climate. Maybe someone out there is excited to see Spotify become more like TikTok, but it’s hard to imagine who.
Those same two phenomena—AI and TikTok—have recently combined to pose an existential threat to search, or at least the version of search that has formed the backbone of the internet for most of the 21st century. Each is encroaching on its turf from a different direction. ChatGPT, of course, swaps one text box for another, replacing the results of Google’s internet-wide trawl with the interrogation of a model of that archive, or what Ted Chiang calls “a blurry JPEG of the web,” a simulacrum of a simulacrum. TikTok, meanwhile, has grown into a robust content universe of its own, a walled garden where users can “find information uncannily catered to their tastes…coupled with a sense that real people on the app are synthesizing and delivering information, rather than faceless websites.” Both serve as correctives to the ads and spam that increasingly pollute Google’s search results—the junkification process that I discussed in a recent post. Neither exactly resolves the problem of Google seeming to show us what it wants us to see rather than what we are looking for, but each offers improvement along a different dimension. ChatGPT streamlines and expands; TikTok humanizes. In this context, Spotify’s implementation of AI is less superficial than it seems, allowing the platform to personify its amorphous (and possibly nefarious) “algorithm” as a benevolent “DJ,” the latest in a friendly lineage from Clippy on down.
The widespread concern about AI replacing humans in various capacities should be tempered by the new roles that it will create for us, and one of those roles is laundering the output of machines to make it more palatable for consumption by other humans. TikTok is evidence that we are already learning how to do this. If something feels relatable enough, we can forgive it for being shoved in our faces. And although I’ve been discussing AI and TikTok as separate things, it should be clear that they are deeply intertwined and complement each other well. TikTok has not become a popular search arena because it works better—wading through video after video is much less efficient and precise than scanning text—but because the universal presence of people in its search results makes it feel less impersonal than much of the internet, and Google in particular. The videos do in fact contain important forms of information that Google has trained us to undervalue (the kinds of information that are also embedded in real interactions with other people). By now, we are accustomed to the algorithmic hall of mirrors that constantly envelops us online, and have largely accepted it, however grudgingly; on TikTok, the algorithm dominates the user experience so completely that it becomes invisible, allowing us to fully submit to it. At a moment when AI promises to ratchet that algorithmic omnipotence up several notches, computers will need a human face more than ever, and social media is where we train for that job. What’s even better than an AI DJ? An actual person who follows directions well.
Become a paid subscriber to access essays that are EVEN BETTER than the free ones!!
Last week’s was about 20th-century Manhattan hotels as novel forms of public space and how the internet has detached urban life from the physical city.
Free trials are available if you want to test the waters first.
Reads:
Rob Horning on AI overreliance. “Part of what OpenAI is selling is its ability to foster ‘general trust’ in its product, despite its obvious flaws…No one need be too concerned with a model’s inner workings or its method of reasoning, because the world can be reshaped around good-enough, plausible-seeming results.” (Also, Rob just turned on paid subscriptions—his newsletter is excellent and one I’m more than happy to pay for. If you like Kneeling Bus, you’ll probably like his too.)
Kate Wagner on why every house is painted gray now. “The home is no longer seen as a space of personal expression or comfort, or as the backdrop of everyday life, but primarily as an investment and as an asset—meaning that enforcing one’s aesthetics is a financially detrimental decision.”
Jonathan Nunn of the Vittles newsletter on the decline of the restaurant review and its replacement by the restaurant map (paywalled for subscribers).
“computer-generated commentary on what makes those songs great,”
That sounds like a glimpse into a dystopian hellscape.
Loved the recommended reads in this one.