Feeding Books Into AI Risks Cultural Void, Essay Argues
Get the Culture newsletter
Daily culture — film, music, books, the trends and ideas worth your attention. Free.
- The author, an unnamed science journalist and science fiction writer based in San Francisco, argues that training large language models on books, movies, games, and art could leave future researchers with only model outputs to reverse-engineer, dubbing the era "the Void Ages."
- The author draws a parallel to the Bronze Age — when humans transformed global civilizations but left virtually no written narratives — warning that AI training could deliberately recreate that information loss.
- Samuel Pepys's 350-year-old diaries of the 1666 plague provided the author personal solace during the 2020 pandemic, cited as a concrete example of how first-person human records transmit experience across centuries.
- George Eliot and H.G. Wells are invoked as examples of 19th-century fiction that captured the anxieties, biases, and psychological tics of a rapidly industrializing era — things facts alone cannot preserve.
- The author distinguishes acceptable AI uses, including code development and big-data analysis, from unacceptable substitution of LLMs for human writing, journalism, and cultural creation.
Why it matters: The essay reframes the AI-training debate from a copyright dispute into a civilizational memory problem: once originals are subsumed into models, the human perspectives, biases, and first-person experiences those works preserve become unrecoverable. Writers, journalists, and publishers are the stakeholders most directly exposed, because the record of early-21st-century human thought — what Samuel Pepys did for the 1660s — is what gets lost.
Ask SkimNews
