Key Takeaways:
- Anthropic spent millions scanning books to train AI.
- They cut books from bindings and threw them away.
- The project copied Google’s book-scanning approach.
- The process was fast and cheap but destroyed the books.
- This raises concerns about preserving physical books.
Anthropic’s Big Move: Millions Spent on Scanning Books for AI
Imagine a world where millions of books are torn apart, scanned, and then thrown away. Sounds crazy, right? But that’s exactly what happened when Anthropic, a company building an AI called Claude, spent millions to scan books for their AI.
How Anthropic Built Claude
Claude is like a super smart robot that can answer questions and understand language. To make Claude smart, Anthropic needed a lot of information. They decided to scan millions of books.
Who Did They Hire?
In February 2024, Anthropic hired Tom Turvey. He used to work on Google’s book-scanning project. His goal? To get “all the books in the world.” Sounds ambitious, but Anthropic wanted to copy Google’s method. Why? Because Google’s scanning project survived legal challenges and became a big deal.
The Process Was Fast and Cheap
When Anthropic scanned the books, they didn’t just borrow them. They cut the books out of their bindings and turned them into digital files. Then they threw the physical books away. This makes the scanning process faster and cheaper because they didn’t have to worry about keeping the books safe.
But why destroy the books? For Anthropic, speed and cost mattered more than saving the physical copies. While this method might seem extreme, it’s what they thought was best for building Claude.
The Legal Side
The story came out in a court document about fair use. Fair use is when you can use copyrighted material without getting permission, like for education or research. Anthropic argues that scanning these books falls under fair use because they’re using them to build Claude, which is like a research tool.
Google’s book-scanning project also faced legal challenges but won because it was seen as fair use. Now, Anthropic is hoping the same logic applies to them. But not everyone agrees. Some people think scanning and throwing away books goes too far.
The Big Debate
Destroying books to train AI sparks a big debate. On one side, some think it’s important to use books to make AI smarter. On the other side, many believe we should save physical books for future generations.
Anthropic’s approach is different because of the huge scale. While smaller groups have used similar methods, chopping up millions of books is something new. This raises questions about how far companies should go to build new technologies.
What’s Next?
Anthropic’s actions are making waves in the tech and book worlds. If more companies copy this method, it could change how we use books forever. It also pushes the limits of fair use. How much can companies take from copyrighted material to build new technologies?
In the end, Anthropic’s decision to destroy books for AI shows how much the world is changing. Technology is advancing fast, but we need to think about what we’re losing along the way.
What Do You Think?
Should companies like Anthropic be allowed to destroy books to build AI? Or should they find other ways to train their systems? Let us know your thoughts!