AI Firms Are Destroying Books For Their Training

There's a bit of a buzz, and some concern, recently about how big AI companies are getting the information they need to 'learn'. It's come out that some, like Anthropic (one of the major players in artificial intelligence), have been buying millions of books – even rare ones – and then scanning them into digital format. The tricky part is, once they're scanned, these books are apparently being destroyed.

Now, why would they do this? Well, to make AI smarter, these programs need to read and process huge amounts of text. Imagine trying to teach a child everything about the world; they'd need to read every book in a library. AI is similar, but on an even grander scale. So, they're gobbling up books to feed their AI models — the complex computer programs that power things like ChatGPT — so these AIs can understand language, facts, and even how humans think.

The 'destructive scanning' part is what's raising eyebrows. Critics are worried about the loss of original books, especially rare ones. Our libraries and archives hold unique historical records and cultural treasures. If these are being physically destroyed after being scanned, even if a digital copy exists, we're losing the original artefact. This is a big deal for historians, collectors, and anyone who values physical books.

For everyday Australians, this might seem a bit far off, but it touches on bigger questions about how AI is developed. We're seeing these powerful tools become part of our lives, from helping small businesses with marketing to assisting students with research. Understanding where the data comes from, and whether it's gathered responsibly, is important for everyone. It's about ensuring that the pursuit of smarter AI doesn't come at too great a cost to our shared knowledge and history.