Clarkesworld’s Dilemma Is A Specter On All SFF Fiction

For Clarkesworld, a monthly science fiction and fantasy magazine that publishes short stories and non-fic-ion pieces, submissions are its lifeblood.

Above is the graph that forced Clarkesworld to cut it off.

On February 20, editor-in-chief Neil Clarke went to Twitter to share this graph, and explain what it represented: “the number of people banned … for sending in machine-generated spam.” It had nearly surpassed the number of actual submissions before they shut them all out.

Immediately, this spawned talk of Artificial Intelligence (AI) replacing creative professions and whether machine outputs can be considered “art.” However, missing in the conversation is how particularly intense the issue is for Sci-Fi/Fantasy (SFF) literary magazines, with how the spam affected the industry all at once, meriting an aftershock reaction from other SFF publications just days after Clarkesworld’s announcement.

This dilemma is certainly not exclusive to SFF magazines. However, it is a uniquely sensitive one to them. From SFF’s underrecognized origins, to its progression into the mainstream, to its persistent critical discourse, at stake is a fundamental threat to the genre’s reputation that refuses to go away quietly.

Previously, under normal times, Clarkesworld had a policy of always being open to submissions. Despite that, they maintained some seemingly picky submission requirements:

Science fiction need not be “hard” SF, but rigor is appreciated. Fantasy can be folkloric, contemporary, surreal, etc. Though no particular setting, theme, or plot is anathema to us, the following are likely hard sells:

  • zombies or zombie-wannabes (seriously, I’m not kidding)
  • sexy vampires, wanton werewolves, wicked witches, or demonic children
  • stories about rapists, murderers, child abusers, or cannibals
  • stories where the climax is dependent on the spilling of intestines
  • stories in which a milquetoast civilian government is depicted as the sole obstacle to either catching some depraved criminal or to an uncomplicated military victory
  • stories where the Republicans, or Democrats, or Libertarians, or . . . (insert any political party or religion here) take over the world and either save or ruin it
  • stories in which the words “thou” or “thine” appear
  • -talking cats or swords
  • stories where FTL travel or time travel is as easy as is it on television shows or movies
  • stories about young kids playing in some field and discovering ANYTHING. (a body, an alien craft, Excalibur, ANYTHING).
  • stories about the stuff we all read in Scientific American three months ago -stories about your RPG character’s adventures
  • “funny” stories that depend on, or even include, puns
  • stories where the protagonist is either widely despised or widely admired simply because he or she is just so smart and/or strange
  • stories originally intended for someone’s upcoming theme anthology or issue (everyone is sending those out, wait a while)
  • your trunk stories
  • stories that try to include all of the above

This list characterizes all of these tropes as signifiers of subpar writing, but not all are made equal. Some, such as those invoking excessive violence or hamfisted politics, could definitely be distasteful to Clarkesworld’s readers, but some are widely accepted in popular media – like zombies in a recent prestige HBO show. However, it is that popularity that I think Clarkesworld has an aversion towards. A niche magazine claiming to publish idiosyncratic stories would understandably avoid tropes that appear derivative or repackaged from other works. That aversion is not unwarranted … as SFF was once characterized this way as a whole.

Of course, science fiction and fantasy novels existed before 1900. However, in the first half of the century, SFF as a mass number of banned submitters media label was inextricably associated with pulp magazines, which originally published short stories of all kinds. While pulps later differentiated into niches such as romance, detectives, noirs, etc., lumping together genres in this way continued to be a path of criticism, or outright derision.

“The science thus discussed is idiotic beyond any possibility of exaggeration, but the point is that in this kind of fiction the bending of light or Heisenberg’s formula is equivalent to the sheriff of a horse opera fanning his gun, the heroine of the sex pulp taking off her dress.” Benard DeVoto, 1939, review of “Doom By Jupiter”

This can’t be written off as uninformed snobbery. The magazines were sold with the assumption they would be thrown away within hours — in fact, they were called “pulps” for the cheap wood pulp paper they were printed on. Because of that limited time frame, like how one immediately recognizes what a celebrity tabloid is about when sighted on the grocery checkout shelves, the pulps had to be eye-catching and effortlessly consumed. It shows in the magazines’ priorities: The cover art always got the higher quality paper, and sometimes was commissioned before the story lineup was even known.

Thus, it was a commercial prerogative that pulp fiction could be reduced to easily recognizable and approachable tropes, even if it risked being synonymous with unoriginality. It makes sense that Clarkesworld, which uses a similar format but sidesteps the disposability issue with an online medium, would want to avoid that reputation.

The entire conceit of AI generation assumes that quality fiction can be generated from its constituent tropes and plot elements, summarized within a prompt — but this seems to me as backwards as the practice of designing a pulp cover before deciding what stories it will depict. There is room to input only so many keywords. As a result, the generation process is itself an act of reduction: finding what in its entire pool of training data matches the keywords, and then stripping their blurred commonalities down into a surface-level uniqueness.

Maybe that mutual preconception is why there are people who think that SFF magazines and machine writing should accept each other.

After all, the old pulps weren’t picky about submissions. Often, they paid authors less than a cent per word. At a biweekly or even weekly release schedule, pulps took what they could get and authors sent what they had. Pulps were not found in esteemed libraries, but were left on newsstands, hoping to be picked up alongside the morning paper. The pulps were understood to be mass media offerings, competing with the “slicks” and other, more established magazines. Thus, the fiction within was regarded as comparatively lowbrow.

Yet to some extent, the pulps always had quality writing, from now mainstay names in SFF like Lovecraft, Asimov, Sturgeon, Heinlein, Dick, Clarke … it’s not an exaggeration to say that without the pulps as a platform, the science-fiction genre of today would not exist.

Nonetheless, the pulps would not last against a general cultural shift towards mediums of greater literary appearance. Digests — magazines of smaller size and higher quality paper — removed the internal art and started arranging text in single columns to better emulate novels. Popular magazines like Playboy and Harper’s began publishing the fiction they used to discard. The introduction of the paperback novel was the final nail in the coffin, a path for SFF short stories to be published as anthologies without betraying their pulp origins. By the time Ray Bradbury expanded on his digest-published short story “The Fireman,” releasing it as “Fahrenheit 451” in both paperback and hardback forms, SFF had managed to escape the defunct pulp pigeonhole and become broadly accepted as “real” literature.

I lay out this historical context to point out three details. One: It wasn’t so much that the writing changed, but rather the medium changed for SFF to begin reaching its current mainstream acceptance. Two: that shift to the mainstream was ironically tied to an appearance of gaining a more literary, curated, and discerning taste. Three: the pulps weren’t antithetical to this trend, but just as they were to the dime novels of the 1800s, an evolution.

Clarkesworld maintains a few practices of the pulps: its stories are republished in non-magazine mediums, it still pays authors by the word (thankfully, 12¢ nowadays), and it still considers all submissions that it receives. AI generation now threatens these principles, prompting well-meaning fans to suggest asking submitters to leave a refundable deposit, or compiling a list of trusted authors to limit submissions to, among other suggestions, as an ex-post deterrent to AI-spam grifters. Neil Clarke has rejected them all out of hand.

For Clarkesworld, it is a matter of self-respect to cast a wide net and then uplift authors to an accessible medium, motivated by the same reasons that pushed previous SFF authors from the pulps to the paperbacks to achieve their literary reputation today. For others however, this historical disinterest in gatekeeping SFF to the works of already respected authors binds the genre to the stereotypes of a defunct medium. No matter what authors got their start in the pulps, those magazines were seen as lowbrow reading because they didn’t filter out all the junk.

So then, if even one AI-generated submission gets accidentally published in Clarkesworld, will that mean AI-fiction will have earned literary respect? Or will it bring back the lowbrow reputation that SFF took decades to extricate itself from?

Some authors and critics may have already sidestepped this conundrum by establishing their own reputation, separate from the commercialized labels of sci-fi and fantasy. They choose to call their work “speculative fiction”. Horror, alternate history, utopia/ dystopia, and so many other genres can fall under this umbrella, making this term nebulous.

However, the label inherently invokes an air of self-importance, asking the reader to treat a story’s premise as a “speculative” thought experiment, designed to be extrapolated to its real-life context. For example, Ursula Le Guin’s The Left Hand of Darkness incorporates both a fantasy world with the social structure and mysticism of medieval Europe and an intergalactic ambassadorial main character visiting this world, but these elements are unified under the same experimental question: “what would a world populated by people of shifting biological sex?” Or: “What would a more flexible understanding of sex and gender in our world be like?”

Le Guin was part of a generation of feminist writers who found a home in speculative fiction. This makes sense: The thought experiment was a means of envisioning alternatives to the patriarchal society, and the novel a means of bringing mainstream awareness to said alternatives.

That self-importance, however, would sometimes drive mutual conflict, as speculative fiction writers unsurprisingly had different, yet deeply personal, conceptualizations of what these alternative societies should be like. Le Guin herself would bow out of a 1975 “Women in Science Fiction” symposium after extensive discussion of her “honorary male” reputation in the publishing world, accusing colleagues of “writing John Wayne’s wetdreams with the sexes reversed.” The specific stances aren’t the point — simply that speculative fiction draws in authors who ascribe existential and personal significance to their work, to the results of their thought experiments.

To this day, SFF authors still receive similarly scathing criticism from writers and readers of their own kind. Back in January 1st, 2020, Clarkesworld published a short story named “I Sexually Identify as an Attack Helicopter.” The title invokes a transphobic internet meme that mocks the idea of non-conforming gender identities. After a reactionary response accused Clarkesworld of letting alt-right rhetoric slip into their magazine, the story was retracted at the author’s request.

The author’s name was Isabel Fall. She was a trans woman herself. By the time she would speak publicly about how the story was partially an experiment in anonymously exploring her gender identity, she had given up on both that name and identity.

I was going to leave it at that. But in the middle of writing this article, I realized that I had learned about this controversy long after it had been publicly “settled.” I had never read the story myself.

So I did.

I know when I wake up tomorrow, I will feel differently about (the now retitled) “Helicopter Story.” Sometimes, I’m enraged by the implication that identity-questioning people are easily manipulated and brainwashed by political pressures. Sometimes, I’m thrilled by the author’s

willingness to call out gendered social constructs, from assumed roles in warfare to internalized taboos about our own bodies. Sometimes, I feel this story simply regurgitates the same existential questions about the melding of body and machine that the cyberpunk genre has exhausted over decades. Sometimes, I feel this story is the only thing I’ve seen dare to ask: “would a transphobe’s perversion of our world make questioning identity any less important?”

In the end, all I have to note is that an inconclusive experiment is not the same as a failed one. Some hypotheses, like what a fully-realized “feminist” world should be like, will probably never have a certain conclusion. That doesn’t make such experiments any less common or important.

When I think about what those spamming Clarkesworld are trying to accomplish, I can only think of two reasons. The more charitable reasoning would be that they seek to prove AI output can be indistinguishable from legitimate literature. The more likely reason is that they don’t seek to prove anything, but are here to steal a quick buck, at 12¢ per word. In either case, they are either unwilling at the outset to accept inconclusiveness regarding whether AI-fiction is “real literature” … or never cared about that justification to begin with.

What is happening to Clarkesworld is inevitably going to be a dilemma that SFF, the publishing industry, and all literature will face. At least there is solace in the understanding that, if AI-proponents refuse to judge and care about others’ art in the same way real artists do, their demands for respect do not have to be taken seriously. After all, without the willingness to accept the full range of possible conclusions, their theoretical world that openly accepts their work can only be described as … a failed thought experiment.

Jason Liu ‘23 studies in the Olin Business School. He can be reached at jliu1@wustl.edu.

Share your thoughts