Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

How the flying f*ck did they collect all the data for GPT-3?



That work started back in the late 80s, early 90s, and people have been building and organizing corpora at an increasing scale. Cleaning a corpus is a hell of a lot of work. I guess they've been using older work to bootstrap that.


Scrape the web. Same way Google do it.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: