Data-labeling firms like Scale AI and Appen are hiring writers and poets with humanities backgrounds to improve the literary quality and creativity of AI tools
Training data companies are grabbing writers of fiction, drama, poetry, and also general humanities experts to improve AI creative writing. Mastodon: @oisinmcgann@mastodon.ie . X: @iethics and @iethics Mastodon: @oisinmcgann@mastodon.ie : In an ad and design company I worked in years ago, our stupid boss came into the studio one day to say a client had asked for a copy of our very expensive design software. He was going to give them our tools. For free. — This is dumber than that. — https://restofworld.org/... … X: @iethics : “The companies say contractors will write short stories on a given topic to feed them into #AI models. They will also use these workers to provide feedback on the literary quality of their current AI-generated text”: https://restofworld.org/... #ethics #law #data #internet #writing @iethics : Let us not to the training of #AI / Admit impediments: “[H]igh-profile training #data companies... are recruiting poets, novelists, playwrights, or writers with a PhD or master's degree”: https://restofworld.org/... #ethics #literature #business #tech #LLMs #highered
Context & Ripple Effects
This is an early sign that language-model training is becoming a labor market for editorial and creative judgment, not only data annotation. It sits alongside coverage of AI-generated publishing that warned automated text was expanding across books and personalized articles, putting pressure on the same human-writing market now being recruited for training work.
Later coverage shows that this contractor model broadened from creative exercises to writers inventing chatbot responses and journalists doing fact-checking and prompt drafting. That makes the hiring pattern relevant to both model quality and the changing economics of freelance knowledge work.
First-order effects
- Scale AI, Appen, and similar vendors gain access to writers and humanities specialists whose work can supply higher-quality creative examples and evaluations for AI systems.
- Writers and poets gain a new contract channel, but their creative expertise is being purchased as training input rather than necessarily as publishable work.
Second-order effects
- Data-labeling competitors will have reason to compete for specialized editorial talent, not just large pools of generalist annotators, as creativity and style become model-quality differentiators.
- The same freelance workforce can be directed beyond prose quality: later reporting describes freelancers creating difficult safety-test prompts, widening the range of tasks AI-training intermediaries can sell to model builders.
Third-order effects
- If this model persists, human creative labor increasingly becomes an upstream, metered input to AI products—while AI-generated output can compete for downstream writing demand.
- The boundary between editorial work, data labeling, and model evaluation may continue to blur, making contractor standards and attribution central issues for the AI content economy.
The trend: AI development is turning specialized human judgment—from creative writing to safety testing—into a scalable but intermediary-managed input for model improvement.