Posted 4/6/2024, 10:40:00 PM
Meta Explored Using Copyrighted Data to Train AI Systems Without Licensing Deals
- Meta considered buying publishing house Simon & Schuster to obtain data to train its AI systems
- The company hired contractors in Africa to summarize books, some of which included copyrighted information
- Executives discussed continuing to collect data from copyrighted sources without licensing deals
- Lawyers pointed out ethical concerns with taking intellectual property without permission
- Meta lawyers believe the company can use books for AI training under fair use precedents from the Authors Guild vs Google case