AI Companies Are Hoarding Old Books
Seeking clean data sources for AI training.
1AI Companies' New Data Source
Companies seeking clean data to train AI models are turning to printed books published before 2022. These books are free from AI-generated content, making them ideal for training purposes.
2Risk of Model Collapse
Most data scraped from the internet now contains AI-generated content, which can lead to 'model collapse'—a decline in model performance. Old books eliminate this risk entirely.
3ISBNdb's Role
ISBNdb is a platform that supplies books to AI companies. It emphasizes that books contain curated, peer-reviewed knowledge with a structure unmatched by web crawls.
4Privacy and Ethical Concerns
AI companies demand strict NDAs to keep their identities hidden during book scanning. Ethical issues, like the physical destruction of books, are also under scrutiny.
5Authors' Backlash
Some authors oppose the use of their works in AI training. Some even create specially designed texts to sabotage the models, complicating the data collection process.
6Future Data Sources
Clean and reliable data sources are becoming increasingly critical for AI companies. While old books meet this need, they also bring ethical and legal debates to the forefront.
Original article: Slashdot — Yapay Zeka Şirketleri Eski Kitapları Topluyor