The Hand Made Web

Or, "How I Stopped Worrying & Learned to Love the AI Apocalypse" (sort of...)

A few days ago,

Caroline Ross

drew my attention to an essay about the way that generative AI is wrecking the internet. “This isn’t what everyone feared,” the author wrote, “which is AI replacing humans by being better – it’s replacing them because AI is so much cheaper.” This got me thinking about the longer history of work and cultural production, and also about the changing way in which I use the internet, and the result is the essay that follows.

Meanwhile, a warm invitation to join me on Sunday (10 March) at 8pm CET when Caroline will be my first guest in the season of “overheard conversations” that we’re calling the Sunday Sessions. To join live on Zoom for these fortnightly sessions, take out a paid subscription to this Substack.

The Cuyahoga River on fire in the 1950s. Is this the state of the internet today? And if so, what’s worth doing?

A long time ago, when I was far more involved in and enthusiastic about tech than almost anyone reading this will remember, I used to say: what I care about isn’t the World Wide Web, it’s the Hand Made Web.¹

That phrase came back to me with force the other day, as I read

Erik Hoel

’s post, ‘Here lies the internet, murdered by generative AI’. Because it struck me that I have a clearer sense now of what these words could mean – and because, given what he’s describing, the Hand Made Web may soon be the only habitable version of the internet that’s left – and because some of us are already inhabiting it.

You’ve probably noticed what Hoel is getting at, how large parts of the internet are increasingly polluted to the point of becoming unusable by the products of AI: the wordjunk that fills up search results, Amazon listings and inboxes, produced by a technology that simulates intelligence, and pumped out by people who have found ways to squeeze money out of this.

The worst of it – Hoel points out – is the hellscape of YouTube Kids, where a great wave of AI-generated nonsense floods the still-forming brains of our youngest children. People will look back on this with the disbelief with which we look back on Victorians sending kids up chimneys, or newspaper adverts for the health benefits of cigarettes.

Hoel concludes with a call for a Clean Internet Act, along the lines of the Clean Air Act of 1963 – a societal response to the pollution of the common space of the internet – and this is where I want to take a different tack. Because I think there are a couple of things worth questioning in his read on the situation. Asking these questions may lead to a different picture of where we find ourselves and what is worth doing.

1. Hack Work

If you pick up a copy of one of David Walliams’s endless series of The World’s Worst Children, Parents, Teachers or whatever the comedian-turned-children’s-author is onto now, it will come with a cover quote from a broadsheet newspaper declaring him “the new Roald Dahl”. I’ve rarely felt as offended on behalf of an author.

Whatever you think of Dahl, he was a genuinely inventive author whose books have range and depth. Danny the Champion of the World is tender and moving, my favourite portrait of a father-and-son relationship in children’s literature.² Even when he’s phoning it in – and George’s Marvellous Medicine reads like it was written in a week, without a plan – it’s like witnessing a great improv performance, sharing the performer’s delight in the free play of imagination. Think of Ross Noble on a night when he’s really on fire.

I can’t say I’ve explored Walliams’s contribution to children’s literature as thoroughly as Dahl’s, and maybe his earliest books go some way to explain the comparison. But I’ve read or listened to enough of his recent output to say that he is doing what a cheap stand-up does: churning out variations on the same basic material, with no genuine playfulness, but a well-calibrated targeting of his audience’s buttons.

And here’s where this connects to AI: you could never replace Roald Dahl with a bot, because his work went on surprising the reader, whereas David Walliams could already have been replaced and I’m not sure anyone would notice, because he already writes as though on an algorithmic autopilot. In that sense, the distance between Dahl and Walliams is greater than the distance between Walliams and the AI-generated pap that fills YouTube Kids. That the publishing industry can’t recognise the difference – or is too cynical to care – tells you how deep into this trouble we already were, before the AI-bots arrived to multiply the volume of crap exponentially.

Here’s another example: Hoel highlights the proliferation of AI-generated “workbooks” or “guides” to actual books, filling up the Amazon results for the reader who wants to order the book itself. As it happens, I’m currently rereading

Matthew B. Crawford

’s The Case for Working With Your Hands,³ in which he tells the story of his own experience, more than thirty years ago, producing a similar kind of “content” in what can best be described as a human-powered bot farm.

Fresh out of an M.A. in philosophy, Crawford is hired for what sounds like a stimulating job, where he will spend his days reading articles from academic journals. It turns out he is expected to summarise fifteen articles per day, rising to twenty-eight by the end of his first year. He’s forbidden to make use of the existing abstracts for the articles, since this would remove the “value added” from the database his employer sells to libraries: “It was hard to believe I was going to add anything other than error or confusion to such material.”

Looking back on this dispiriting experience, Crawford observes that the logic under which the work was being carried out “all but guaranteed that the work cannot be animated by the goods that are intrinsic to it.” There’s an echo here of a much broader observation, made by John Berger, in a 1970 essay on the history of oil painting:

The art of any culture will show a wide differential of talent. But I doubt whether anywhere else the difference between the masterpieces and the average is as large as it is in the European tradition of the last five centuries. The difference is not only a question of skill and imagination, but also of morale. The average work – and increasingly after the sixteenth century – was produced cynically: that is to say its content, its message, the values it was nominally upholding, were less meaningful for the producer than the finishing of the commission. Hack work is not the result of clumsiness or provincialism: it is the result of the market making more insistent demands than the job.⁴

To resist the general logic which leads to hack work, talent is not enough: some combination of good fortune and sheer stubbornness has been required, a constitutional inability to do other than be animated by the good that is intrinsic to the work, even at significant personal cost. This is the element of truth behind the Romantic image of the suffering artist.

The point of bringing in this longer perspective is not to imply that “It was ever thus!”, but to locate the deep historical roots of the proliferation of shoddy, lowest-common-denominator cultural production, lacking in meaning, demoralised and demoralising. It is when the logic of the market breaks its bounds and is allowed to dominate all other kinds of human relationship that we find ourselves in a world of hack work. This has been the case in more and more fields of human activity, in a process going back centuries.

If it’s reasonable to locate the new tide of AI-generated dross within the much longer history of hack work, that’s not to say that the quantitative explosion of this material is anything other than an enormous and unwelcome change. But it matters whether this change represents the ruining of a previously good state of affairs, or the intensification of an already bad state of affairs. If we’re dealing with something more like the latter scenario, then the intensification may bring things to a crisis, and the crisis may contain a strange kind of hope. This is the kind of conclusion towards which I am tempted.

Before we get to conclusions, though, let’s open up a second line of questioning, because I want to probe the appropriateness of the pollution metaphor in Hoel’s original argument.

2. Landscapes and Dreamscapes

In a world of demoralising hack work, we need places of refuge, spaces to bring the parts of ourselves which it would be unwise to bring to work. If we’re lucky, we may even find or help to create spaces within which other kinds of work remain possible, be they pockets of anachronism, pockets of privilege, or pockets of resistance.

One source of the sorrow you hear old-timers express about the way the internet turned out is that many of us first experienced it as a refuge, an escape route through which we found our way to pockets of meaning, and even sometimes to meaningful livelihoods. Whereas it’s long been clear that its overall impact is to intensify the demoralising effects of the unbounded market, whether in the way we are invited to show up on the big platforms which came to dominate the web, or in the impact of online activity on the local economies and civic spaces of the places where we live our offline lives.

By the time it spilled out of the military-academic complex, the network was already in need of spatial metaphors: sites, architecture, cyberspace, the Electronic Frontier. Yet it’s a strange kind of space we visit through our screens, unlike the surroundings that we inhabit the rest of the time. It is as if we had constructed a giant collective memory palace, or were walking in and out of one another’s dreams. And what distinguishes these online spaces from the geographical landscapes and physical architecture whose metaphors we borrow to describe them is that, when we stop visiting them, they dissolve like a dream. This why I suggest we think twice about the metaphor of pollution.

Industrial waste set the Cuyahoga River on fire, multiple times. Fallout left the Chernobyl exclusion zone unfit for human habitation. What comes afterwards may involve expensive remediation work, or animals moving in to fill the gap created by the exodus of humans. But the point is, when pollution affects a landscape, that landscape isn’t going anywhere. It is still there to be reckoned with. Whereas when parts of the dreamscape that makes up our online lives cease to be habitable, the best move is often to abandon them.

My sense is that some of us have been doing this for a while, more or less consciously – and, in the process, finding our way to a use of these technologies that is less vulnerable to the AI infestation, because it was already resistant to the logic of hack work.

3. The Hand Made Web

Tech people have a strange way of talking about trust. They often use the word to mean its opposite. The philosopher

Andrew Taggart

pointed this out to me, years ago: when a tech enthusiast talks about “creating trust through smart contracts on the blockchain”, or whatever, they are really talking about constructing impersonal substitutes that make it possible to operate in the absence of what humans generally understand by trust: it’s precisely when I don’t trust you that I need to get your commitment in writing, witnessed by a third party. The trouble with trust is that it takes time and doesn’t scale easily.

What I’m going to call the Hand Made Web is the way that some of us are using the network, not against itself, necessarily, but against the logic that tends to dominate its larger spaces. The web that I inhabit today is smaller and slower than the one that I was part of five or ten years ago, and it is woven through with enough threads of actual trust to limit the need for impersonal substitutes. The quality of expression is such that the words I read or listen to don’t seem especially vulnerable to replacement with AI. And the scale of audience for most of the voices in this Hand Made Web is small enough that I can’t see these spaces attracting actors whose modus operandi is to flood a space with low-grade content and profit from a tiny conversion rate, which is the logic behind much of what Hoel describes.

Let me sketch out what this looks like, knowing that your version will be different. There are maybe half a dozen blogs I visit regularly, a couple of which I’ve been reading most weeks for the best part of twenty years. Then there’s a longer list of Substacks, some where I’ll rarely miss a post, others which I dip in and out of, or keep half an eye on. There’s a group chat with a dozen members whose interests meet around one side of the themes I write about, a couple of smaller chats like that, and various folks who I get together with on a regular or semi-regular basis on a video call. A couple of podcasts I listen to fairly steadily, while others come across the horizon and set me on a deep dive through their archive for a week or two. I have a podcast of my own that gets a thousand or so listeners, and this Substack with a few thousand subscribers and a couple of hundred paid supporters.

These are the ways I spend my time online. The pattern is sharpened because I’ve been on a digital fast since the start of Lent, which focuses attention on what’s worth doing with my limited screen-time. So I spend five minutes most days skimming the headlines on a mainstream news site, I may pop up once a week on Facebook and Instagram, if I’ve put out something worth sharing with folks who still spend time on there. Aside from that, you’ll find me in this little corner of the Hand Made Web.

Looking back, I started moving this way in 2017, when I withdrew from Twitter.⁵ And maybe this is just me getting old, but I think there’s more going on. For a start,

Yancey Strickler

made some of the same points, in a darker key, five years ago, in The Dark Forest Theory of the Internet, which clearly struck a chord with a lot of people. Around the same time,

Richard D. Bartlett

got me thinking about trust and the way we live our lives online, when he wrote, “I no longer seek my accountability from strangers.” In the comments on Hoel’s post,

William Collen

describes a conversation with a twenty-year-old who “thinks his generation will abandon the internet altogether”.

So I’m curious what others are seeing, whether my description fits with a broader experience, and how this plays into your reaction when you read about the damage that generative AI is wreaking on the internet at large?

Finally, if you’d like to hang out in my corner of the Hand Made Web, you’re warmly invited to join me this coming Sunday (10 March) for the first of these Sunday Sessions, a season of “overheard conversations”, where I’ll be getting together with people whose work has been nourishing my thinking and writing.

My inaugural guest will be

Caroline Ross

of

Uncivil Savant

who, as it happens, was responsible for bringing

Erik Hoel

’s post to my attention. So join us for a conversation about beauty, friendship, conversation and “using the false to cultivate the real”. These sessions are open to anyone with a paid subscription to

Writing Home

and I’ll send round a full announcement about the series in the next day or so. Meanwhile, here’s where you subscribe.

DH

1. I was, for my sins, co-founder of a web startup called the School of Everything, inspired by Ivan Illich’s Deschooling Society, which was the main focus of my work for three years between 2006 and 2009. So, for a few years, I was in the thick of the London tech scene.

2. When

Elizabeth Oldfield

wrote a while back about how often children’s authors need to “get the parents out of the way” for the story to start, Danny was one exception that sprang to mind.

3. Published in the US as Shop Class as Soulcraft, a title that doesn’t cross the Atlantic.

4. John Berger, ‘Past Seen from a Possible Future’ (1970) in Selected Essays.

5. Though writing this, I think of friends who report that they have made a corner of Twitter into something resembling what I experience as the Hand Made Web.