The Ugly Death of Curation

It's the algorithm's world, we just live in it

Aug 10, 2023

Hello!

This week it’s all about that pesky algorithm. Yes, yes, a tired topic but not as exhausting as actually being on the internet dominated by For You. It’s something I’ve been thinking about a lot, and suddenly everything came into focus as I read a few articles (linked below) and a Threads post by (our king) Ezra Klein. Through the death of Twitter we’ve arrived at a new internet, void of meaning but abundant in contrivance. Cool!

This isn’t just me complaining, although there’s a bit of that, I really wanted to get to the crux of how we got here and it really does feel like an inflexion point where things could go in a much better direction sooner than you think. I feel like the silent majority are all on the same page, but lack the platform(s) to express or act on that concern. Or am I an eternal optimist and the algorithm is too big to fail? See what you think and as always, please hit me up or leave a comment.

Declan x

The Week in AI

Google has teamed up with Lupe Fiasco to create a very cool new tool for rappers (and anyone else) to help with ways to connect words, phrases and sentences called TextFX. It’s easier to just watch the video and try it for yourself than to explain how it works, but it’s exactly the kind of tool we love to see – augmenting creativity instead of replacing it.
This week the FT reported that Google and UMG are in talks to create licensed AI models of artists from Universal’s roster. It’d be a huge deal if it did go through and, in my opinion, would change the landscape and mood around AI models from threat to opportunity.
Meta has launched AudioCraft, an AI text-to-audio and music tool. Meta is open-sourcing its paper for research purposes “to help advance the field of AI-generated audio.”

The End of Curation

The internet is getting more annoying. As the age of curation becomes the age of For You, I find myself questioning what I’m seeing more and more. Why am I being shown this? Why did someone make this? What is the point of this? Who is this for? This has always been part of the online experience, but over the past few months, it feels like it is the only version of the Internet we are permitted to experience. There is absolutely ‘good’ stuff behind the messy chaos, but it feels harder and harder to find.

These were the thoughts dominating the discourse of my mind until I saw a Threads post by everyone’s favourite softly-spoken NYT journalist Ezra Klein: “I find myself exhausted with the algorithmic internet and longing for the curatorial internet. Individual people telling me what they found, and liked.” (The replies are worth reading for loads of Substack reccs.)

And that made me wonder, why is this exhaustion happening now? What’s changed in the recent past that’s shifted our internet experience to rely on what platforms want us to see over what we choose, at the expense of our own sanity? And then it hit me: the collapse of Twitter.

Feed First

Twitter was, of course, a fucking nightmare. No one is going to sob openly at its funeral. But Twitter represented one of the last bastions of the early internet – a curatorial feed of links from people you choose to follow. Of course, there were RTs and QTs, but they were only ever one step removed from your own curation, and because of its strong Mute and Block game, it was pretty easy to avoid any topics, people and trends you didn’t want to see.

Twitter’s slow and painful death over the past year has highlighted that lack of curation and the heavy reliance on the For You feed that’s come to dominate social media thanks to TikTok’s runaway success. And it’s exhausting.

I am absolutely not the first person to think this, but as we sit in this inflexion point between the end of Twitter, the scramble to launch Threads and the uncertain future of feeds everywhere, it feels like a good time to step back and wonder what may lie ahead. To do that, it’s helpful to look to the past.

*‘Traffic: Genius, Rivalry, and Delusion in the Billion-Dollar Race to Go Viral’ tells the fascinating story of the early internet’s chaotic quest for traffic*

Human Traffic

I’m currently reading a book called Traffic: Genius, Rivalry, and Delusion in the Billion-Dollar Race to Go Viral by Ben Smith. It explores the rise of virality through the stories of The Drudge Report, The Huffington Post, Buzzfeed and many other early websites that set the tone for our modern-day internet. I should point out that I’m only about a third through, but the reason I bring it up is that Ezra’s Threads post coincided exactly with the part of the book where a plucky disrupter named Jonah Peretti starts Buzzfeed while working at Huff Post, and with it, begins dissecting and hacking how Google crawls their site. In doing so, Peretti was a pioneer of what would later become known as SEO.

Up until that point, the majority of leading websites were simply links to other places. It was pure curation – you experienced the internet through the lens of a person or org you trusted and in turn, they grew in their influence as they sent traffic to other blogs and websites.

There was no algorithm: everyone saw the same version of every site. That’s not to say there weren’t dirty tricks being played as CPMs became a core revenue stream but ultimately there was no algo to be gamed; it was a mix of tabloid-esque baiting, crude celeb gossip and ‘takes’ traditional mainstream media wouldn’t touch and a race to being ‘first’, as bored office workers hit refresh multiple times a day.

Everything changed when a colleague of Peretti’s – Andrea Breanna – created a new type of dashboard that monitored traffic in near real-time, split into categories. “That was the moment we went from mushroom hunting to farming – and when we opened Pandora’s box,” Breanna would recall later.

At the time, Breanna’s dashboard was controversial – when visitors came to the Huffington Post office, the screen would be hidden from view. “Serious journalists sneered nervously at the idea that you’d allow your news judgement to be replaced by crude clicks,” writes Smith.

A combination of Breanna’s work and other analytic companies like DoubleClick and Right Media slowly began to reshape how content was commissioned. Internet aggregator site Digg, similar to Reddit, also played its part: “Digg’s power came from an opaque blend of community and algorithm, and it was beginning to shape not just what got read, but how the news was written,” Smith concludes. “The tail had begun to wag the dog.”

*DeepMind’s office within Google HQ at King’s Cross, London*

Two Types of Internet

We’ve come a long way since the early days of traffic hacking, but Klein’s labelling of these two ‘types’ of internet – curatorial and algorithmic – is really helpful in thinking about how things have changed. But while Klein, myself and many others might long for a more personalised version of internet existence, the reality is that algorithms are only going to become more dominant as a tsunami of AI content requires filtering in the coming months and years.

In her paper, ‘The Algorithmic Internet: Culture, Capture, Corruption’ DeepMind’s Christina Lu summarises the damage the algorithmic internet has caused. “What do we become when the machine smooths over the need to think?” she writes. “Algorithmic ease destroys our capacity for sustained attention and with it our capacity for care, empathy, and cultural innovation.”

Ultimately, despite condemnation of its impact, Lu accepts the need for some kind of mechanical interference at the window of the ’net. “Salvation will lie in neither a datacenter-smashing anarchy nor an orderly, fully comprehensible system of computation, but in a deliberate corruption of the dream of scale and ease.”

The Art Formerly Known As

The algorithm's negative impact on how we experience current affairs and politics is fairly obvious, but how it affects our experience of music, art and culture is a more subtle story.

In January 2023, an exhibit called The Algorithmic Pedestal choose to explore these themes by uploading a series of images from the Metropolitan Museum of Art’s Open Access collection to an Instagram account. Next, researchers at the Oxford Internet Institute followed the account and made a note of the order in which Instagram decided to display the images in its feed. They were then displayed next to a series of images that artist Fabienne Hess collated under the title ‘Dataset of Loss’. The idea was to show how a person might arrange a series of images that went through a very human process, and a very human selection criteria, versus Instagram’s ‘black box’.

“Many of these algorithmic platforms, such as social media platforms like Instagram, were not created with the intention of artistic display,” said Laura Herman, a doctoral researcher at Oxford Internet Institute, University of Oxford. “They have very different goals: enabling connection between friends, selling ads, gaining attention, serving as a marketplace, and so on. This means that the underlying formulas – that is, the algorithms – are not tuned to artistic considerations of aesthetics, beauty, novelty, or even creativity.”

*The music industry’s genuflection before the algorithm has left fans and artists more disconnected than ever*

Streambait Pop

The same can be said, of course, for music. Much has been written about the homogenous effect of Spotify algorithms on music, including the creation of tongue-in-cheek genres like ‘streambait pop’, coined by Liz Pelly. “Musical trends produced in the streaming era are inherently connected to attention, whether it’s hard-and-fast attention-grabbing hooks, pop drops and chorus-loops engineered for the pleasure centres of our brains, or music that strategically requires no attention at all—the background music, the emotional wallpaper, the chill-pop-sad-vibe playlist fodder,” she wrote in 2018.

“All of this caters to an economy of clicks and completions, where the most precious commodity is polarised human attention—either amped up or zoned out—and where success is determined, almost in advance, by data.”

In a piece in The Guardian in July 2023, Dan Brooks pondered the death of ‘selling out’, writing: “By generating massive amounts of more precise and almost totally accurate data … streaming gave artists, labels and platforms a much clearer perspective on what sells and what doesn’t. And because this perspective is (a) provable and (b) quantitative in ways that ideas such as ‘selling out’ or ‘authentic’ or ‘good’ are not, the people who decide what music to promote are pretty much obliged to decide according to one metric: what audiences like, defined as what they have liked before.”

“To a greater or lesser degree,” Brooks continues, “this phenomenon has for the past two decades shaped every field of popular culture: not just music but also television, movies and online publishing.”

One of my favourite explorations of this topic in recent months came from YouTuber Venus Theory in this 18-min dissection of how we got here, titled ‘The Playlistification of Music’. Ultimately, Venus Theory accepts that although the system rewards echo chambers, the financial incentives mean that it works for almost everyone, except the artist.

“It’s in the music industry’s best interest to create stuff that does well on streaming services, it’s in the streaming service's best interest to find the music that resonates best with its audiences and keeps them on said platforms and it’s in the audience's best interests to use a streaming service to sort through this massive catalogue of stuff into something they’re going to enjoy.”

The conclusions Venus Theory’s video comes to are no more heartening than you might expect – essentially suggesting to game the algo until you get an audience and do what you want afterwards. A fair take in an era of perpetual noise.

It was, then, painfully ironic that major labels recently came out and proclaimed that it’s harder to break new artists than ever. In an article in Billboard this week, one “senior executive” lamented: “Nobody knows how to break music right now. I think they’re all lost.”

The article continues:

“It’s common to hear grumbles about young acts who have hundreds of millions of plays of a single but can’t fill a small room for a live performance. ‘It’s easier [today] for folks to be passive fans,’ J. Erving [a manager and founder of the artist services and distribution company Human Re Sources] says. ‘For [an artist] to consider [themselves] really broken, people need to care about you beyond the song. Where is the connectivity? Are people really dialled in a deeper way?’”

No comment.

What Now?

The curatorial internet, for all of its flaws, still felt peer-to-peer to some extent. There was a middleman, but it was a facilitator, not a dictator. What was posted was seen, and what was seen was reacted to and it felt like the whole world could have been in that one place at that one time, simultaneously experiencing the same thing.

In For You’s world, we all exist in a silo, presented with our own watered-down version of life with no shared experience or meaningful connection. Or as Christina Lu puts it: “We have been psychologically hijacked by corporations who prey upon our desire for belonging while short-circuiting our ability to satisfy it, leaving us stranded in restless myopia.”

But to simply say everything the algorithm suggests is evil and malicious and everything shared by a human is joyful and sincere is obviously very wide of the mark. Algorithms have been extremely useful in sorting through the shit, and they’ll be even more necessary in the age of generative AI. But it’s the lack of control that leaves me feeling helpless and frustrated. It’s about balance, and right now it’s heavily swayed in the machine’s favour.

So where do we go from here? With the algo working so well for attention and retention, are we stuck with For You, forever? What does the future look like for personalised curation?

*Threads was a very elaborate owning of Musk, but where does it go next?*

Threadless

Meta’s Threads launched with a flurry of ‘I Told You Sos’ in July, but quickly dropped off as its basic features tired quickly. To be fair, it’s barely one month old so it’s not entirely clear if some features are bugs or if some bugs are features.

Threads is fine, but it has a fundamental problem that I don’t see ever being solved: Meta doesn’t want hard news. They’ve had their fingers burned too many times (for good reason) and I cannot see them walking back on that policy, especially after all the goodwill the new app seems to have gifted them. Adam Mosseri, Meta’s Head of Instagram, essentially said as much on a Hard Fork podcast special.

By abandoning any serious content that might actually make it a useful tool for journalists and citizens alike, it’s unclear what Threads is actually for.

Come for the Elon mocking, stay for the… actually, let’s just go.

Stacked in our Favour

It’s not lost on me that this newsletter is being sent through a completely algorithm free, 100% curated platform in the form of Substack. And that’s good! In fact, that’s what Ezra was asking for more of in his post. The inbox, after all this time, is still a fairly sacred place – frozen in time and ready when we are, rather than an endless FOMO-powered stream of banalities. The inbox is apolitical and without agenda. But it’s also siloed, lacking the social context that can give recommendations a new level of meaning. That being said, forwarding emails was the first iteration of virality online – maybe it’s time to bring back email groups. Or is that basically what newsletters are?

*Daisy Alioto’s Dirt is a great example of a new twist on traditional blogging*

Economy of Taste

In Daisy Alioto’s somewhat viral piece on the Taste Economy for her excellent newsletter and website Dirt, she makes the point that as content becomes even more prevalent than it is today (thanks AI), taste will become the new currency. It’s a compelling argument.

“[Print] magazines were containers for someone else’s taste,” she writes, “and when you read them you were inhabiting that taste. But you were also co-creating with the magazine, because by reading the magazine you were contributing to its fantasy of lifestyle.”
“Most investors see the financial success of a Mr. Beast,” she continues, “and work backwards, looking for the next vessel. But when I am served videos by someone who has been anointed with this stardom I don’t feel like I am inhabiting someone else’s taste but, rather, the taste of the algorithm.”

This is ultimately the crux of the point. The algorithm cannot deliver the intimacy of connecting with another person because you’re not connecting with anything at all. An algorithmic recommendation has zero emotional value compared to a recommendation coming from a trusted friend. By sharing art, music, articles and even memes, you create three new strands of value – between you and the friend, you and the art, and both of you and that art together.

A study in 2011 actually proved this, claiming “music can create interpersonal bonds between young people because music preferences can be cues for similar or dissimilar value orientations, with similarity in values then contributing to social attraction.” Algorithms segregate us by design – it’s how they tailor our feeds – but as the study showed, shared experiences, even on the internet, can lead to shared values which lead to deeper connections.

As our friends at Billboard outlined above, mass abundance with zero meaning doesn’t come without consequence.

The Fate of the Feed

As Shawn Reynaldo put it in his First Floor newsletter back in June: “I figured, or at least hoped, that once [Twitter] kicked the bucket, all that time and energy we’d all become used to investing in Twitter could potentially be directed toward new spaces—and new ways of communicating, preferably without the invisible hand of an algorithm pushing us all to become our seemingly worst selves.”

Perhaps there simply will be no Twitter replacement, and Substacks and newsletters, blogs, WhatsApp Groups, Discord servers and other communities will combine to fill the void. At the same time, I feel like curation’s stock is rising as Twitter’s core user base is left in no man’s land. Are the data-led strategies that have dominated newsdesks, publishing houses and major labels over the past 15 years starting to show their cracks?

As Tina Brown, former editor of The New Yorker and Vanity Fair once said: “To ride the zeitgeist successfully you have to know when it’s turned.”

Maybe the ugly death of Twitter can spark something beautiful.

Future Filter

Discussion about this post