I recently plunked down $2 to read Andrea Grimes' excellent piece for FlyTrap media, "Hate-Reading Wirecutter in the Age of Nascent Era of AI Slop." I'm not going to excerpt the many, many banger lines here because, frankly, Grimes and Flytrap deserve to be paid for good work, but I do want to include this bit because it is worth expanding:
I’d almost rather get a bad recommendation from a real person than a good recommendation from a bot (which, let’s be real, probably explains my hate-read dedication to Wirecutter). Is that so much to ask? To expect that it’s really possible to go looking for a “simultaneously beautiful and functional” coat rack, and find it … because someone else found it first? That’s what Wirecutter promises, but in a media landscape where they are competing with paid influencers, AI bots, and anyone who can faux-test and synthesize anything anyone else has tried, I crave a human element that is less about what everyone might want and more about what one person definitely wants.
(Emphasis mine.)
What Grimes is advocating for is what we used to get in magazines: a distinct point of view. To become a regular reader of a magazine is to join a community of shared taste, whether unironically or in a sort of reactive way. It’s to absorb that editorial voice and respond to it.
For example, the only thing Cat Marnell and I have in common are that we are both bipedal hominids, but her kick-off to a product round-up devoted to the late Carolyn Bessette Kennedy lives rent-free in my head for its distinctiveness: "Carolyn Bessette-Kennedy was a party girl and I am a party girl and I know what beauty products she used and here is my dumb outfit and I think my brain is broken bllllarggggg."
In that one sentence, I knew ev-er-y-thing I needed to about the publication she was writing for at the time and what it stood for. By contrast, AI-generated content is as opaque and tasteless as warm rice pudding. Who is it for? What does it add? Nobody knows because no human perspective was required in the generation of the string of words and hyperlinks in the bland little bowl of slop.
SO WHAT
One of the amusing contradictions of the attempt to make fetch generative AI happen is that retail rises and falls on the development of trends, and the one thing AI will never, ever be able to solve is how to scale economies of taste and perspective. Generative AI is fundamentally a pattern-extrapolation machine that iterates based on training data and queries. Taste, on the other hand, is derived from highly personalized and idiosyncratic decisions.
To give one example: the husband-and-wife team of Ann and Sid Mashburn run an echt-preppy empire in the American South. (Sid is responsible for having designed the original J. Crew barn jacket in the 1980s, so imagine how wild it must be to see how other retailers are nostalgia-riffing on his work.) However, whenever Ann writes about how she's inspired by her husband's style, she writes about the African beads he mixes into his outfits, or the Birkenstocks he was wearing. You'd think those wouldn't vibe with his witty interpretation of southern prep, but they only make it more memorable, precisely because the person wearing those things knows how to play with contrasts and outliers.
Above: the OG barn jacket.
Is that sort of thing scalable? No -- because taste doesn't rely on data. Taste creates one data point. It takes multiple data points to create a trend.
Or as writer William Gibson's coolhunter Cayce Pollard explained it in Pattern Recognition:
"The ‘cool’ part—and I don’t know why that archaic usage has stuck, by the way—isn’t an inherent quality. It’s like a tree falling, in the forest. … It’s about a group behavior pattern around a particular class of object. What I do is pattern recognition. I try to recognize a pattern before anyone else does.”
“And then?”
“I point a commodifier at it.”
“And?”
“It gets productized."
One can see where generative AI could be tailor-made for this type of thing. It's good at identifying patterns. But it's not good at instigating the particular class of object that sparked the entire pattern in the first place.
WHO CARES?
Anyone who is at all interested in living in an objective, factually-based reality, for one. And anyone who is paying attention to how carefully and quickly the modern media ecosystem is sifting people into the haves and have-nots.
Google's had an outsized impact on journalism in the past two decades, first by incentivizing publications to make it their primary audience (where else does SEO come from but by appealing to the algorithm before the human?) and now by slurping up original human reporting and writing, then regurgitating it as a characterless summary at the top of any search page.
The thing is, people don't love slop. They really do love finding those communities of taste; the specificity allows them to either see themselves in that community or to define themselves in response to that community.
Why do you think Substack has taken off? It allows people to launch their tiny empires of influence -- things like Ruby Redstone's Old Fashioned, which goes maximalist glam. And it's no coincidence that lifestyle retailers are launching their own media-esque imprints. People love to bag on Gwyneth Paltrow, but the Goop newsletter is nothing if not a distinct community of taste.
However, let's look at the sifting. Substack subscriptions add up; it's pricey to be tapped into tasteful media, while it's still cheap to google something and see what ungated content you can read. As documented extensively in Elizabeth Currid-Halkett's The Sum of Small Things, the new upper class is organized around informational capital:
The restructuring of the global economy prizes a meritocracy, who own the means of production through their minds, not land ownership. These labor market elites (many of whom are members of the aspirational class) believe in upward mobility and want the same for their children. Their dominant ethos—working hard and acquiring knowledge—is also the dominant cultural hegemony and spills over into all walks of life.
[...]
Finally, and most importantly, material consumption (particularly post-Recession) is less valuable than investing resources into the consumption that counts, like education, retirement, and health care, all of which price-out ordinary people but are critical conduits in the reproduction of aspirational class position and further separating the rich from the rest.
The people who can afford it will still retain access to information that is by humans and for them. For everyone else: when a search engine becomes a simultaneously bland and imprecise content-recapping machine, it is impeding your ability to access information from primary sources committed to something greater than appeasing a data-matching algorithm. Flattening out the authority that comes from either institutional expertise or individual experience makes things less useful.
Specificity is what makes media valuable. It's just too bad it doesn't scale.
YOUR POP CULTURE RECOMMENDATION OF THE DAY
I went to see the new Superman movie and liked it a lot, but I'm in the tank for James Gunn's whole thing. What I want to recommend instead is one of the graphic novels I suspect was used as raw material in the script: Justice League International Book One: Born Again, which features a collection of C-listers led by the Green Lantern most likely to use the power ring to put truck nuts on a Cybertruck, Guy Gardner.
The team of Keith Giffen and J.M. DeMatteis cranked out some of comics' funniest tights-and-flights teams, and anything they do is worth reading -- I also recommend Formerly Known as the Justice League. Reading this will likely prep you for the James Gunn era of DC movies better than anything else will.