When GPT-4 came out, I tried for a while to use it to write a story. It was kinda fun, but ultimately I decided that the story was bad, and using GPT-4 was making it worse. (I'd be happy to give more details/examples in the comments if curious.) When GPT-5 comes out I'm going to try again.
If, in my subjective opinion, it's a potent tool that actually helps me write fiction, this market resolves YES. If I judge it to be insufficiently good, like I did for v4, it resolves NO. This market only resolves at the point where we have a model from Open AI that is either called GPT-5 or is "clearly GPT-5 with a different name" according to my subjective opinion. The market also only resolves once I've had the opportunity to use it for a while to try to write fiction. (I expect to try this shortly after launch, but I might be busy such that it takes me some months to really try.) If OpenAI ceases to exist this resolves NO.
I won't bet.
Update 2025-08-12 (PST) (AI summary of creator comment): - Threshold: Creator considers Claude to meet the "YES" bar for being a helpful fiction co-author.
Sufficient condition for YES: If GPT-5 is roughly similar to Claude or Gemini at helping them write fiction, they will resolve YES (decision likely in the next few days).
1,000People are also trading
@traders I've been playing around with the new 3.5 Sonnet and while it's still not quite there, it's good enough that I now think this market is underpriced.
Are your main problems with GPT-4 the word choices (which is my main problem with it, it has a certain bias/tone in what it says which can contaminate even if you ask it to rewrite some of your own text) or idea generation?
(Since I expect tone to improve, like it did from GPT-4 -> GPT-4o, but probably not be solved completely, and I expect it to get better at suggesting ideas and handling complex ideas)
@Aleph My main complaint, though I have several, is that it's garbage at juggling tension well. If there's a setup, it'll either lampshade it super hard and then do a "payoff" a few paragraphs later, or it'll forget about it entirely. Good storytelling is about walking the line of uncertainty and making the reader ache to find out what happens next, and I almost never find that in current AI prose beyond the line level.
It also is very tropey and cliched if I don't pressure it into being more original. The word choice can be tonally off (as you mention). The AI can't seem to write from a confused/disoriented viewpoint. And it's pretty garbage at pacing (for reasons related to the big complaint, above).
I could work around some of these flaws, but eventually the tool isn't worth the effort to use it well.
@keltan I usually write science fiction. I don't want to commit to anything at the moment. You may have overheard me talking about my previous novels http://crystalbooks.ai and perhaps how I'm working on audiobook versions. https://youtube.com/playlist?list=PLO1uFvmwxzgrh3R2qeezvXdDBzYn7HtXl&si=_LSTSxAXnjkA6pfn