A fun weekend project would be to utilize GPT-2 [2] to model HN comments; quite the challenge considering the usually insightful comments here when compared to other sites.
I agree the content is actually funny in that reddit tread. I wonder how good it can become with a really large good database. I read in some article of OpenAI a dataset gets better by for example replace names by pronouns.
A fun weekend project would be to utilize GPT-2 [2] to model HN comments; quite the challenge considering the usually insightful comments here when compared to other sites.
[1] https://xkcd.com/810/ [2] https://github.com/openai/gpt-2