I got about 3000 words in and lost interest but I will say this - I always wonder with posts like this whether you are seeing examples that are representative or cherry-picked. I assume the latter.
Which implies that we are reading about what humans select, not what the machine learning is selecting, artificially enhancing how interesting the output of these algorithms really is.
That's actually a point the author makes - he sees an AI as a tool for the artist, not as an creative subject on its own.
I found it highly interesting to read about a writer's perspective on text producing AIs. Especially the part where its output suddenly was "I don't know what you want me to do".
Part 2 - where he presents a sci-fi short film with a plot written by AI - is also a interesting read.
That's actually a point the author makes - he sees AIs as a tool for the author, not as a author itself.
I found it highly interesting to read about a writer's perspective on text producing AIs. Especially the part where its output suddenly was "I don't know what you want me to do".
Part 2 - where he presents a sci-fi short film with a plot written by AI - is also a interesting read.
Having played around with neural networks for generating text, I strongly suspect that most of the funny/deep examples that people share are cherry-picked (I'm also guilty of this but typically also post a million characters+ of raw output in addition to any samples).
Cherry-picking or not, this is a pretty awesome project
> Unfortunately, I’ve had trouble making it say anything interesting about language, as it prefers to rattle on and on about the U.S. and Israel and Palestine.
And the simulated version shares the same problem?
Which implies that we are reading about what humans select, not what the machine learning is selecting, artificially enhancing how interesting the output of these algorithms really is.