AI & You — The Generated Content Debate

by Mandira Pattnaik

Andrea De Santis

Sometime last year, in his Google office, software engineer Blake Lemoine was having a conversation. He was simultaneously preparing a transcript, and at one point asked the one he was talking to, what was to be afraid of. To this, they engaged in a dialogue about rights and personhood.

About 2000 miles away, a creative writing professor was grading essays, and his wife, an editor with a popular literary magazine, was reading submissions. Outside her window, the weather was perfect, a lovely summer morning, and she was convinced everything was perfect with the world. She upvoted six of the twenty she had been assigned.

Lemoine was convinced too. He didn’t delay to report his findings. That April morning, he sent a Google doc titled, “Is LaMDA sentient?” to his bosses. Lemoine’s engagement with LaMDA (language model for dialogue applications) however did not find favor with Google and other leading scientists. Dismissing the findings, they said LaMDA was simply a complex algorithm designed to generate convincing human language. Lemoine was subsequently fired.

The Editor and her husband continued to read and vote and grade creative writing for the rest of 2022. They were proud to champion great new voices, happy to discover originality and talent. When these writings were applauded by others, they were delighted their time and labor didn’t go in vain. Later in the year, they read about the Davos conference on the internet. IBM’s chairman and chief executive Arvind Krishna predicted a wave of job cuts from AI. “You should worry more about the clerical, white-collar jobs than the physical [jobs]. A large number of them will get replaced. So the question is: ‘What jobs do you create to replace those?’” Arvind Krishna said, according to a report.

Somewhat worried, the professor and his wife still continued to believe in the power of originality, and the uniqueness of art and creativity. If they imagined the community of writers, copyeditors, magazine editors and publishers as a solar system in a universe of content creators, the stars would be ‘Originality.’ Stars — light and heat — the spectacular mechanism of generating something incredibly powerful from within itself.

This changed last week. The couple woke up to the shocking news about AI-Content-writing flooding submission queues of Clarkesworld Magazine, reported in The Washington Post, The Guardian and Wion.

The guidelines of the publication clearly state that it doesn’t want AI written or assisted works. Neil Clarke, editor of Clarkesworld, said in a Twitter thread, that in spite of such clarity: “They (the submitters) don’t care. A checkbox on a form won’t stop them. They just lie.”

Most genuine writers, admittedly tech-challenged, had not even heard of ChatGPT, let alone AI Content writing until then. It emerged that with AI content generators, one can choose the right tone for unique content from over 20 options. One just selects the use case or template, enters some input for context, and lets Rytr do the rest! Rytr is one of the platforms offering the service, initially for free. One can also enter one’s own sentences to guide Rytr along, highlight text and edit it using the “expand,” “shorten,” “rephrase,” “append,” and “write operation” functions, plus, the generated content can be edited and formatted to suit needs.

Predictably, the revelations sent shockwaves. What were we reading all this while? How much of it was not human-original? Did we champion a fake writer? Suddenly, the editor questioned herself if she was reading AI generated content. Asked, how many fellow writers and people within the community were not ‘originals’ and simply faking it?

Not a great week to go by.

It also emerged that the detection mechanism is being discussed and developed, not foolproof yet, meaning there’s the possibility, however slim, that a genuine writer could be banned for something they didn’t do.

This February, the editor sitting and reading at her desk can empathize with the editor of Clarkesworld. Neil Clarke’s concerns, understandably, go beyond the human-versus-machine debate. She can imagine the scenario where writers and editors are worried that an AI-generated text is next in line for the Booker Prize. As of mid-February, Amazon was reportedly selling over 200 AI-generated novels.

But the greater debate is AI-driven spam could silence voices. One suggestion that gained momentum during the initial revelations was to use some sort of barriers, like only allowing writers with n number of publication credits. Also, perhaps, making all opportunities fee-based. This is precisely the most feared reaction to the development. Clarkesworld is so extensively affected because it is free-to-submit, with quick responses, just like at trampset and a host of other literary publications. These places attract the best voices — Submissions are received irrespective of the writers’ paying capacity or publication credits, and from diverse places from across the globe. Such a measure would prove devastating for these writers.

Another point to examine is that any creative work carries an ownership and responsibility, other than the aforesaid originality. AI tools cannot take such responsibility. Particularly for nonfiction or research-based writing.

This is certainly a developing story to watch out for. And for all stakeholders to ponder about.