We placed a tracking device in a shipment of rare books to see which AI company was buying it, and found an Amazon facility where Amazon scans and destroys books.
They mean the PDFs of them or ePub or whatever they’re scanning them into. Yeah, the original is gone but imagine if the Library of Alexandria went up in flames but every ash contained a complete work for free to everyone to read at their leisure.
Storage space is valuable for their AI stuff. Why would they keep around copies like that on valuable storage space if they don’t think they’re going to be using them again once they’ve trained their model? I really hope they are scanning them to a format or something like that that. But I doubt it. These aren’t the most forward looking or intelligent people you’ll find. I wouldn’t be at all surprised to find out that they never did anything more than scan it into train the model and then flush it from the system. It’s 100% on brand for them.
You always keep training data because you never know what you need for the next generation. Ironically they might be preserved (privately though unfortunately) pretty well
They mean the PDFs of them or ePub or whatever they’re scanning them into. Yeah, the original is gone but imagine if the Library of Alexandria went up in flames but every ash contained a complete work for free to everyone to read at their leisure.
Oh, and Ram.
Storage space is valuable for their AI stuff. Why would they keep around copies like that on valuable storage space if they don’t think they’re going to be using them again once they’ve trained their model? I really hope they are scanning them to a format or something like that that. But I doubt it. These aren’t the most forward looking or intelligent people you’ll find. I wouldn’t be at all surprised to find out that they never did anything more than scan it into train the model and then flush it from the system. It’s 100% on brand for them.
You always keep training data because you never know what you need for the next generation. Ironically they might be preserved (privately though unfortunately) pretty well