Hacker News Reader

Comment by carlosjobim | Hacker News Reader

I think this is why the shadow libraries are the only spheres that will be able to resist being taken over by AI. If you're risking jail time to distribute the written word, you're not going to waste your resources on AI generated stuff.

View on HN

Running, protecting and supporting these shadow libraries is one of the biggest service to contemporary's humankind's written culture that can be done nowadays. Sometimes the discussion is just on "copyright" and "for knowledge to be free", but some older material is actually realistically impossible or extremely hard/time consuming to find out of these shadow libraries. Money is not the only problem that makes certain material inaccessible. Having these places shut down and their material disappear by copyright predator firms would be catastrophic in a way that has probably not happened since events like the burning of the library of Alexandria. The current state of AI just makes it even more important.

If you're risking jail time to distribute as much of the written word as possible, you're going to include AI-generated stuff as well. Because the alternative is to waste time checking every single book.

LibGen already has plenty of garbage: books that were already garbage when they were written, blurry or truncated scans, OCRed text in the wrong script without the original scans etc. etc.

If someone hacks Google Books and downloads the entire collection, are they going to laboriously filter out the AI garbage? No. Just throw it on the Pile.

Your comment makes sense, but we have a real world example to compare to that I think is relevant: YouTube videos. They are not uploaded to torrent sites or other pirate download sites. The closest we have to "pirating" are Invidious proxies. And I'd say that the reason is that the proportion of garbage to quality is so high on YouTube, that pirates don't really want to touch it. If AI slop becomes a problem, it's not going to be included.

The incentive for publishers of AI generated books is that they make money selling them to people.

As Youtube becomes less and less of a stable platform, archivists are in fact starting to do that, but knowing that it's free media provides significantly less incentive for leechers.

Do you know that they don't serve AI generated content or are you just assuming?

I am just assuming. I use many different sources for finding information, and this is how results have been for years:

Social media: Mostly not AI generated

YouTube: Significant part AI generated

Google: Mostly AI generated

Kagi: Significant part AI generated

Anna's Archive: Never seen any AI generated results

LibGen: Never seen any AI generated results

Google Books: Never seen any AI generated results

Google Maps: Never seen any AI generated results

Are there any other huge online resources for finding information that I've missed?

Considering that the amount of information available in a the shadow libraries rivals the general Internet in size, I think it is impressive how they have managed to keep relatively clean.

Why not?

Why isn't spam a highly effective attack by the copyright agencies?

Only the librarians can add books to the library.

I think that AI generated from the shadow libraries is going to be the only good AI

If you're willing to risk jail time to distribute an AI, you're not going to waste time on capitalist AI

Long live communist AI!

what shadow library?

Pirate websites to download books. They are right now the largest libraries that have ever existed. Anna's Archive is the largest, LibGen is the most famous one.

Their roots are from Soviet times, when underground librarians distributed books out of control by the government.

I already knew about them and appreciate some of the figures behind them, but when you lay it out like that, that's so badass.

Reply on news.ycombinator.com