Meta torrented & seeded 81.7 TB dataset containing copyrighted data

#1

Post by gameshot911 » Fri, Feb 07, 2025, 11:26 AM UTC

Meta torrented & seeded 81.7 TB dataset containing copyrighted data
arstechnica.com

Re: Meta torrented & seeded 81.7 TB dataset containing copyrighted data

#2

Post by tremarley » Fri, Feb 07, 2025, 11:36 AM UTC

ebooks are a 1-2 mb each max. 81.7 TB are a lot of books, like 42-85 million books.

Re: Meta torrented & seeded 81.7 TB dataset containing copyrighted data

#3

Post by thunkingdeep » Fri, Feb 07, 2025, 11:38 AM UTC

tremarley wrote: post #2

ebooks are a 1-2 mb each max. 81.7 TB are a lot of books, like 42-85 million books.

I’ve got 70-80mb pirated books, I think because of the illustrations. Guess it depends on the book.

Re: Meta torrented & seeded 81.7 TB dataset containing copyrighted data

#4

Post by mateus1 » Fri, Feb 07, 2025, 11:39 AM UTC

thunkingdeep wrote: post #3

tremarley wrote: post #2

ebooks are a 1-2 mb each max. 81.7 TB are a lot of books, like 42-85 million books.

I’ve got 70-80mb pirated books, I think because of the illustrations. Guess it depends on the book.

I don’t think they’re using picture heavy book for LLM training, no?

Re: Meta torrented & seeded 81.7 TB dataset containing copyrighted data

#5

Post by panki27 » Fri, Feb 07, 2025, 11:39 AM UTC

They could have at the very least seeded some more, to give something back to the, uh, community.

Re: Meta torrented & seeded 81.7 TB dataset containing copyrighted data

#6

Post by moralestapia » Fri, Feb 07, 2025, 11:43 AM UTC

mateus1 wrote: post #4

thunkingdeep wrote: post #3

Earlier quoted context omitted.

I’ve got 70-80mb pirated books, I think because of the illustrations. Guess it depends on the book.

I don’t think they’re using picture heavy book for LLM training, no?

Yes they do, there's multimodal models.

Re: Meta torrented & seeded 81.7 TB dataset containing copyrighted data

#7

Post by mnsu » Fri, Feb 07, 2025, 11:43 AM UTC

mateus1 wrote: post #4

thunkingdeep wrote: post #3

Earlier quoted context omitted.

I’ve got 70-80mb pirated books, I think because of the illustrations. Guess it depends on the book.

I don’t think they’re using picture heavy book for LLM training, no?

For multi-modal models, why not? They would be probably some of the best data.

Re: Meta torrented & seeded 81.7 TB dataset containing copyrighted data

#8

Post by Havoc » Fri, Feb 07, 2025, 11:43 AM UTC

Really curious what the judges are going to do here.

Horse has functionally bolted on this already

I’m guessing slap on wrist despite courts going after individual for a couple of movies torrented pretty hard

Re: Meta torrented & seeded 81.7 TB dataset containing copyrighted data

#9

Post by WithinReason » Fri, Feb 07, 2025, 11:45 AM UTC

mateus1 wrote: post #4

thunkingdeep wrote: post #3

Earlier quoted context omitted.

I’ve got 70-80mb pirated books, I think because of the illustrations. Guess it depends on the book.

I don’t think they’re using picture heavy book for LLM training, no?

Presumably they didn't create the torrent

Re: Meta torrented & seeded 81.7 TB dataset containing copyrighted data

#10

Post by squigz » Fri, Feb 07, 2025, 11:46 AM UTC

tremarley wrote: post #2

ebooks are a 1-2 mb each max. 81.7 TB are a lot of books, like 42-85 million books.

It could be anywhere from a few million to a hundred million

https://annas-archive.org/datasets