EVENT
Le Guin's Works in Training Data
The posthumous irony: Le Guin's copyrighted works—including her warnings about commodification—were scraped without permission to train the systems she opposed.
Dozens of
Ursula K. Le Guin's books, including "
The Ones Who Walk Away from
Omelas" and
The Dispossessed, were found in pirated datasets (Books3, others) used to train
large language models including GPT-3, GPT-4, and their successors. The woman who wrote that "the profit motive is often in conflict with the aims of art" and who defended artists' rights to control their work's use had her life's creative output consumed without permission or compensation to power the AI systems she had warned against in her 2014 National Book Foundation speech. The
appropriation is not metaphorical; it is the literal instantiation of the structure she described: creative labor extracted, processed into a commodity (training data), used to generate profit for people who did not create the work and do not value the practice that created it. Her estate's 2023 participation in the Authors Guild lawsuit against OpenAI sought remedy, but the legal outcome cannot restore the
epistemic violence—the conversion of a lifetime's artistic practice into statistical patterns available on