Wikipedia:Reference desk/Archives/Miscellaneous/2023 February 13

= February 13 =

Using AI models to ease out the transition from shallow writing to in-depth writing
After a much discussion about the idea of hiring Marcus Bromander and CoreyNK (the one that wrote Ninja Kiwi blogs on Reddit) as story writers for a hypothetical Titanfall 3 on Entryway (formerly Qualitipedia) Discord server, the replies I got is neither of them are qualified for in-depth writing for story and lore for the fiction that something in-depth, thus I have an idea to use large language models like ChatGPT for these two in order to ease the transition to it, again, the replies that I got is LLMs have no spirit as humans do, which I agree with. Is it a right approach if the dedicated writers for the particular fiction are short? Is it better or worse than just hiring people with right expertise? 2001:448A:3041:7E63:D3E:5407:F6DD:3DF5 (talk) 02:16, 13 February 2023 (UTC)


 * LLMs cannot produce coherent texts that are longer than the context length they have been trained on. OpenAI has not published the context length used for training ChatGPT, but it is estimated to be maybe 4000 tokens. Presumably, you want the story line for the new game to contain more words than that. Since ChatGPT is accessible for public use, you can ask it to produce a synopsis and see how well it does on that. --Lambiam 08:25, 13 February 2023 (UTC)
 * Here are some links to articles with ideas and guidance about using ChatGPT to assist in creative writing:
 * Using ChatGPT as a Creative Writing Partner | Towards Data Science
 * How Kindle novelists are using OpenAI’s ChatGPT - The Verge
 * Can ChatGPT Write a Better Novel Than I Can? - The Washington Post
 * Writing a novel with ChatGPT : singularity
 * I hope you find these links useful. By the way, I found these links by entering this into the AI search chatbot perplexity.ai: "using ChatGPT to write a novel".  &mdash; The Transhumanist   12:51, 13 February 2023 (UTC)
 * Putting aside the question of whether it can the basic idea is just weird. Titanfall is a very significant work of one of the top video game publisher. Titanfall 3 if it happens is likely to have a very big budget a big team behind it. The chance that anytime soon the story will be done by two people who supposedly (to be clear, I have no comment on the accuracy of this claim) cannot handle an indepth story supplemented by AI is close to zero. If Titanfall 3 happens in the next 5 years, probably 10-20 years even we can be fairly sure EA will have no problems finding and hiring high qualified writers who can handle an indepth story. Whether they'll succeed in making a good story and a good game may be an open question but it's unlikely to be because there's a shortage of talent willing and able to work on it but instead a variety of other factors. Likewise if EA are unwilling to allow and fund the development of Titanfall 3, there are likely a variety of factors of play and the ability to hire two people who can't handle an indepth story supplemented by AI instead of proper story writers is unlikely to make any significant difference into their decision. Nil Einne (talk) 07:26, 15 February 2023 (UTC)