Using generative AI for sub-tasks in text adventures?

Hi Jorn – we’ve been working on a platform that allows creators to leverage LLMs in the creation and interpretation of interactive fiction games. Try one of these:

The platform is still new and under development. But it shows a lot of promise, not the least of which is that it is fun as hell – we have players that spend hours talking with the parser and solving the various puzzles. “Last Audit” has over 1,200 game plays (thank you Reddit and Discord!) – which beats most IF games by a factor of . . Well, not to be uncharitable, by a lot in my experience. Don’t get me wrong – When I hit millions of game plays, I’ll actually have something to brag about. Nonetheless, we are encouraged as we see all the input on the server and can mine it for improved gameplay.

The other major advantage is that players can play right out of the box, versus a trad IF game, where the learning curve of ‘guess the verb” and “spell it right” is not only steep, but frustrating to many modern players. We tested classic IF games in a high school CS class versus an LLM-enabled IF game – the results? Not a single player stuck with an Inform game for more than 5 minutes before quitting – most less than that. It just doesn’t meet modern expectations of gameplay. (Don’t get me wrong – I still love it – but I also remember using punch cards)

I agree with you, I think using the LLM to produce narrative text or dialogue is a mistake.

We almost have the platform in a state where we can invite writers on it to experiment. Let me know if you would be interested. It is pretty shabby-looking, so we’re looking for writers and creators with a sense of adventure and innovation.

Thanks,

M

1 Like