A company called ISBNdb is positioning old, printed books as a superior source of training data for AI models. They argue that books offer curated, peer-reviewed human knowledge that web crawls cannot replicate, and importantly, are free from the "AI slop" of AI-generated text that could lead to model collapse. ISBNdb also notes that authors can actively sabotage AI models by intentionally producing flawed content. AI
IMPACT Highlights a novel data sourcing strategy for AI training, potentially impacting data acquisition and model quality.
RANK_REASON Article discusses a trend and a company's offering related to AI training data, but does not announce a new model or research.
Read on Mastodon — fosstodon.org →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
