A brief history of the future of publishing

Artificial intelligence (AI) is allegedly revolutionizing industries, but is this the revolution we want?

Automated content generation is driving down rates for writers and authors, but also expands the need for outlets to have robust editorial departments. This is largely due to the quantities of content that large language models (LLMs) can crank out. 

Welcome to the slush pile

A slush pile is what those in the trade used to call the unsolicited submissions sent to a publisher. For many a sci-fi or speculative fiction magazine, this used to require a handful of actual staff whose job it was to read through all the schlock sent in by would-be authors. In a way, this was critical to the literary endeavor, because part of learning to write well is seeing lots of examples of bad writing along with good writing. The slush pile was a way for a burgeoning author or editor to cut their teeth, learning what works and what doesn’t. Then they could pass their choices up the editorial chain for review and selection by more seasoned professionals.

With the advent of LLMs, the slush pile is near infinite. AI makes composition easier, but that also makes everyone more prolific, which means bigger slush piles and a generation of readers who are faced with a glut of content and no more time to consume it than they had before. 

Authorship will be all but irrelevant

Once these models can reliably and convincingly generate content, some people will still seek out human-generated content at a premium. But for the most part, the skill in publishing will shift to reviewers and editors with a keen eye for plucking out the AI-generated needle that will resonate with readers from a proverbial haystack of content that varies wildly in quality. The publications that survive will be the ones whose staff can find the best content, not make it. 

For editors, the biggest challenge will be wading through the sea of AI-generated slop to find anything worth publishing at all. Because LLMs lower the bar of writing, large numbers of people who wouldn't otherwise have tried their hand at it are more willing to jump in. These works exist on a spectrum of legitimacy, from real authors using AI to supplement their creative process and overcome writer's block, to cynical grifters whose only goal is to have AI crank out a flood of stories which they submit as their own work in hopes of getting paid by the word.

The sheer scale of the problem is already forcing some outlets to take drastic measures. For example, the sci-fi magazine Clarkesworld was forced to temporarily shut down submissions because it was being overwhelmed by the volume of slush-pile content it was receiving.

From makers to finders

One of the biggest concerns surrounding AI in publishing is the question of authorship and who actually owns the work they produce. I don’t think it will ultimately matter, though. What we’ll have is a system where “writers” will generate, edit, and log articles, whether they use AI or not. For those who do, most of that job will be reading what the LLM they’re working with produces and deciding what to send to their editors. They’ll be slush pile readers. Their job will be to find or produce a quota of articles per week. Their editors will post them, and a fractional amount of ad revenue and syndication fees will be transferred to each of them every time someone reads it. Once published to the internet, a series of contracts will dictate how the piece is syndicated and how much everyone who touched it will earn.

Publishing as an industry will undergo a shift. It isn’t going to be about creating the best content, but about finding it in a vast sea of mediocre content. These are radically different skill sets with entirely different workflows. Perhaps AI companies can use their vast sums of venture capital to subsidize retraining by paying outlets to use proprietary LLMs while their staffs learn to use them. 

Or, perhaps, AI will be tasked to deal with the problems AI is creating. It's likely that publishers will be tempted to use LLMs on the receiving end, using it as a robotic editor to winnow down their slush piles and passing along only the best candidates (as judged by the AI's own inscrutable criteria) for human review.

This isn’t a particularly utopian vision of a future for creatives, who will be reduced to something akin to digital coal miners. They will slog through mountains of content for rare gems of insight. Their pay will be meager except for an exalted few. 

But hey, it’s a living.