3 ms·
(2023 - 1851) * 365 * 200 = 12,556,000 Yep, so a few million ripped off articles is plausible.
by engineer_22 3y ago
(2023 - 1851) * 365 * 200 = 12,556,000
Yep, so a few million ripped off articles is plausible.
- zozbot234 3y agoEverything from 1851 to 1927 ought to be in the public domain, though. If the goal of training an AI is just "to mimic a style" there are absolutely humongous amounts of text that's totally free of any copyright restrictions.
- hnarn 3y agoYes, there is large amounts of public domain text available, but does anyone believe this is a restriction that was imposed when feeding the models?