News for hackers. Stuff that matters.
Re:The Philosophy of Humor: What Makes Something Funny? (2022) by naasking on Sunday October 29, 2023 @10:58AM (#38059194)
You can hypothetically train the most likely outcome to be what people don't expect. I suspect the problem is simply that humour is not really in ChatGPT's training set. Most of its training data is strictly factual stuff like Wikipedia and textbooks. Where's the humour there?

I'm curious to see what would happen if you took a pretrained model and then specialized it on a large corpus of jokes.

There are no comments on this comment yet.