When I was watching this video by Computerphile on the OpenAI’s new language model and its masterpiece about the English speaking unicorns, this comment just jumped out and.
I don’t even want to know who upvoted this comment (I lied. I knew. I upvoted it after I took this screenshot.).
Just, my dear, GPT-2 needs huge data set as its training data.
Speaking of a billion triangle lovestories with hatred, angst, dusted with dub-con and other spices you are using… I really would like to peer review your data set and your output.