Not sure about this. I’m behind it morally, but if a player wants to ask an LLM to process or reformat some gamestate because they don’t have the ability to do that for themselves in Python or Excel, I’m not sure that it’s my place as a BlogNomic player to say that they can’t and (since it will probably give garbled results) shouldn’t.

Ruling out AI banner images? Yes, BlogNomic should take a view on whether its own website supports that aesthetic. Banning AI-generated text? Yes, Nomic is a game of discussion and we want to know whether posts and comments are being written by humans. But using AI tools to make or support decisions might be a bit outside of the circle.

I agree completely with everything that Kevan said and because of that I am leaning

Good to see you again jjm

I guess the disagreement I have with Kevan is the idea that it’s python, Excel, chatgpt or nothing. This dynasty I have exclusively used pen and paper and while I’m not likely to win I do have, at a rough estimate, around 20% win equity. The thing that differentiates winners from losers is effort, and I don’t think we should be treating access to the often-wrong shortcut box as privileged.

Plus the IP thing. I really don’t like the idea of someone feeding my stuff into chatgpt; I understand that it’s the public internet and the Corpus will eventually scrape it all anyway but why make it easy? I would rather have an opt-out.

Oh and yes: hi jjm! Good to see you again 😊

LLM tools are a very broad category, while it’s reasonable to ban AI-generated game content (in a “writing dynasty”, for example, generating text with AI feels like cheating, just like you’re not really playing chess if you feed a board into Stockfish and do what it says), regulating personal use seems outside the scope of the game. Like, people have written scripts to generate a sequence of moves, or analyze data, and that’s generally seen as legitimate play. And if someone wants to use ChatGPT to clean up the wording on their proposal or otherwise support *their own* writing, I don’t have a problem with that.

Now, there’s an interesting gray area in my opinion with regard to something like looking for scams. If you throw the ruleset into an LLM and tell it to look for loopholes, that feels like the computer playing for you, but again, how is it philosophically different from using a solver, spreadsheet, or custom script to optimize resources and actions? The only difference seems to be accessibility, modeling out game actions requires some math/data analysis/programming skill of your own, while anyone can use an LLM. But is that a reason to ban one but not the other?

As far as data collection goes, I think that’s ultimately futile. It’s a public site on the Internet, it already gets scraped, indexed, and spidered, and I wouldn’t be surprised if my BlogNomic posts are already in a training dataset somewhere.

Unfortunately, I can no longer reliably tell whether the content I’m looking at during web research is LLM output.

Hey, Vovix! Love seeing friendly faces in a recussant proposal :)

I’m very split on this but lean in this direction

@Kevan I’m almost convinced the solution to that problem is to ban excel and python too

COV its clear from the votes on this proposal https://blognomic.com/archive/one_step_further certain players like Chiiika want an unfair playing field where they can use their skills with computers to get advantages in dynasties, but don’t want a more level playing field that that could be offered by generative AI

I’d rather just ban everything, but better to keep a more level playing field that gives people more options than to create an unlevel one

I very much dislike the technology, and the folly of thinking that an LLM would actually be correctly performing the analysis that it was confidently claiming to have done (cf. this BoardGameGeek thread I mentioned somewhere else), but I don’t feel it’s my place to say whether or not other players should have the option to use it, when pen and paper isn’t enough.

This is predicated on the understanding that an LLM could do many things that some players would be able to do with scripts or regexps or spreadsheets (the LLM doing these things badly, but at least well enough to somewhat inform a person’s gameplay), but isn’t yet at the point of “computer, calculate the odds of each player winning the game based on this ruleset and gamestate, page, and what my next move should be”, and may never be.

Just to say that I’m grateful that this proposal has been considered, for the most part, soberly and seriously and with an eye to the long-term health of the game, and am particularly grateful to the idle players who have cast a vote on this for their input. I suspect that this issue will be open for re-examination as the technology and its impacts progress.