Thanks, I didn't know either what exactly I was doing and the result. And is not extensively tested. I really wonder how is the experience for others :D, since depends on how each writes and what.
Viewing post in The Uncanny Valley jam comments
Do you really want me to break the magic? :)
Yes, it relates to what you write.
The simulation works as follows:
There are a series of pools of sentences and questions (12) sorted by topic or "mood".
There is a mood selector and a set and sorting of those pools is created (4 of them)
There is a mood score, that varies based on your previous answer (if it found a match to your answer in the pool or not).
For each answer, it goes through the current text pools available in the mood selector and gives them a score based on the words given by you, and if there is some coincidence. There is also a "topic" control variable, which gives more weight to the sentences that contain the longest word contained in the previous prompt is displayed, to help simulate continuity.
The answer that scores the highest from those pools gets picked. If the answer is not a question, it picks a random question as a follow-up.
It also can detect certain rude words.
If no sentence from the pool scores, there is a series of fall-back checks triggered to save the situation, and if none is found it comes with the "insufficient cooperation" answer.
Thanks for the thorough explanation, that's awesome! I understand having the generative AI rule for this jam, but I am absolutely positive that large language models in particular will change how we perceive videogames. I am puzzled as to why there is so much resistance, especially in the indie game dev scene, over genertaive AI.