📚 The research accuses that OpenAI may have used O'Reilly books without authorization to train GPT-4. The AI Disclosures Project (co-founded by Tim O'Reilly, CEO of O'Reilly Media) released a paper stating that OpenAI...

📚 Research Alleges OpenAI May Have Used O'Reilly Books Without Authorization to Train GPT-4o

The AI Disclosures Project (co-founded by Tim O'Reilly, CEO of O'Reilly Media) released a paper stating that OpenAI most likely used O'Reilly's copyrighted books to train its GPT-4o model without authorization. Researchers used a method called DE-COP (a membership inference attack technique) for detection and analyzed 13,962 paragraph excerpts from 34 O'Reilly books. The results showed that compared with older models such as GPT-3.5 Turbo, GPT-4o could "recognize" significantly more O'Reilly paid content, indicating that these copyrighted materials may have been used in its training data.

(Tech Intelligence)

via Teahouse - Telegram Channel