Can we start over? Ferrytale's use of ClubFloyd transcripts.

Technical Development
21

as somebody who joined the Discord specifically to yell at you (a lot, with caps lock) about this, I genuinely appreciate you owning your mistake and trying to fix it & proceed in an aboveboard way; that’s not easy to do and not everyone does it

opt-in is a good start but I feel like the project would be better served by avoiding existing material altogether and working on a way to combine:

  • an agent that’s been trained to provide an “interactive fiction” kind of experience, presenting bespoke writing where it’s available and extrapolating where it’s not, and meaningfully keeping track of game state
  • bespoke writing & game design created specifically for this agent

my guess is once you got the agent down, it’d be faster and easier for authors to create games of this type for your system, since they’re skipping the coding and simply providing object rules & sample writing

two big problems you’re dealing with regardless: first, the “what if the AI bad confabulates,” like it’s out of your hands if Google screws up Gemini such that every character starts uncharacteristically advocating for the eating of Irish babies etc.

there’s also zarf’s point that being able to jump around and eliminate friction destroys the capacity for puzzles, which are friction and a lot of what is fun about games that include them

anyway I genuinely think it’s an interesting design space and I’m curious where you go next

6 Likes
22

As part of that conversation, you kept bringing up how this is “progress”, how it’s “inevitable”, and important for accessibility, how IF is in decline, mentioning poor sales of IF games on Steam, and this is what will help bring many more players and financial success to the medium.

I do believe it’s progress even if still rough, and I do believe it’s inevitable that AI-augmented IF ends up being enjoyed by many That’s all personal opinion, and it’s fine if you disagree. I could have worded things in a more positive light - rather than suggesting IF has been in decline I should have said the player base isn’t growing much and I believe there’s opportunity for more folks to enjoy IF. I continue to believe that too.

What you made is an experimental toy. You’re exploring if something is “interesting”. Don’t jump to imagined future scenarios like “playing interactive fiction while commuting” or “allowing people not familiar with IF to play them”.

I’m not sure what you’re trying to say here, but to me this comes across as “stop dreaming”. Dreamers are always getting ahead of themselves, shooting for something. I’ve admitted to my mistakes and tried to make amends. Despite the tone you choose to take with me, you and I are not actually at odds, and only time will tell how it all plays out.

Those are a lot of assumptions about what’s important to authors and players, and the capabilities and potential of your tool.

I made assumptions about current authors, which I’ve copped to. As for why IF doesn’t have a larger player base, I continue to believe I’ve got a strong hypothesis and next steps. My actual goal isn’t to convince everybody I’m right today - it’s to work with authors who are also interested in exploring AI-augmented IF and hopefully pull more “normie” players. I love IF’s immersive potential.

Suppose you gave a human who’s a creative professional writer/actor a transcript of a game, like you do with the chatbot, asked them to read it carefully, internalize it, and then interact with you, based only on that transcript. Do you think that’s an interesting/useful/amusing/novel thing to have instead of actually playing the game?

Honestly, yes that sounds amazing. These are known as immersive experiences, and they’re pretty popular. From DND to GTAV RP to hosted murder mystery dinners among friends to successful theatrical performances in NYC like Sleep No More, it’s pretty clear that it’s fun and there’s a lot of public interest. A lot of people love interacting with characters, being a character themselves, uncovering storylines and going off script.

And if you think IF has faults you can help with, maybe work with the community on the individual problems, instead of thinking a single prompt to a chatbot is the solution to everything.

I’ve been trying to constructively point out what I see as opportunities for making IF fun for non-players. I’m not entirely sure why you try to diminish the work I’ve done - that’s one thing I would never do to the authors in the IF community. If you don’t like it, that’s totally fine.

23

as somebody who joined the Discord specifically to yell at you (a lot, with caps lock) about this, I genuinely appreciate you owning your mistake and trying to fix it & proceed in an aboveboard way; that’s not easy to do and not everyone does it

I appreciate it. I didn’t see your message because I think I was already banned. Thanks for taking the time to bring your feedback here for me to see.

my guess is once you got the agent down, it’d be faster and easier for authors to create games of this type for your system, since they’re skipping the coding and simply providing object rules & sample writing

100% I’m right now working with an author to create materials specifically for Ferrytale. It focuses on both teaching the AI what good prose looks like (the purpose transcripts served previously) as well as hashing out character motivations, storylines, and themes (which were missing from my initial attempt). Interestingly game state is handled basically perfectly over long playthroughs due to very large context windows and the concept of compaction.

two big problems you’re dealing with regardless: first, the “what if the AI bad confabulates,” like it’s out of your hands if Google screws up Gemini such that every character starts uncharacteristically advocating for the eating of Irish babies etc.

This made me laugh. While it’s possible, I’ve never experienced it firsthand in my many hours of experiencing game transcripts through Ferrytale. When you attempt to do something untoward and outside the scope of the game, instead of getting “I’m sorry I can’t do that” you usually get characters successfully defending themselves or other characters coming to their aid. It’s actually really endearing.

there’s also zarf’s point that being able to jump around and eliminate friction destroys the capacity for puzzles, which are friction and a lot of what is fun about games that include them

I totally get that. I think I need to focus on capturing the moment the player would have gone to a walkthrough and instead give them the subtlest hint to moving forward.

anyway I genuinely think it’s an interesting design space and I’m curious where you go next

Thank you Jenni!

24

I think you might have a fundamental misunderstanding about how much of a piece of IF appears in a single transcript.

4 Likes
25

I think you might have a fundamental misunderstanding about how much of a piece of IF appears in a single transcript.

I honestly do not - I’m familiar with Inform and have dabbled in it myself. I understand that transcripts do not exercise all game conditionals/storylines or reveal all author text. A playthrough transcript will not be the final format for AI-augmented IF. It is an interesting way to show its ability while following the main storyline as explored by ClubFloyd though. The use of transcripts alone is not where this is going, but resembles part of the format at least. Large context windows (~1mil tokens these days but pretty inexpensive) allow games to be defined through author writing alongside themes, motivations, paths, rules, pacing, potential interactions. It’s not dissimilar from the original Inform writing and code. That’s why I’ve called Ferrytale an “interpreter”.

26

In that case, I should caution that I don’t think many authors will be happy about rewriting what they’ve already written based on how an LLM thinks it should go.

7 Likes