I meant that mostly as a joke, but it’s a real moderation technique. If you’ve got griefers in your online game’s community, rather than banning them, you silently alter the matchmaking to match them only with one another. Of course, I don’t think people submitting AI-generated content intend to grief anybody, but maybe you could ask how often griefing in general comes from a failure to appreciate community norms.
2 Likes
From my perspective, there seems to be a few different arguments going on:
File size
Not sure if this is really specific to AI, but I’d be tempted to decrease max file size to say 40-50mb – but then perhaps allow larger files on request (but only if there’s a really good reason for it). This does not rule out rich-media, folks just need to get better at compression! Perhaps more tutorials could be posted to help in that regard?
Quality vs Effort
There’s an assumption that all AI content is ‘slop’. Whilst I’d say that’s generally true, I think the increasingly worrying reality is that it’s possible to create some genuinely good stuff with AI. You have to put the effort in, though. Similarly, I think there’s a feeling that all AI content is low effort, single prompt stuff, and whilst effort != quality, I can understand that if folks are expected to devote their own time to playing it, then it should be representative of a lot more human effort to make it.
Copyright
It’s definitely true that pretty much all current AI has serious ethical question marks re. what it’s been trained on. I think this is where folks can take an ethical stand and simply say they do not allow or wish to play games made with AI - purely for ethical reasons.
But for the quality and effort reasons, I think there should be an allowance for just a little nuance. As an example, if you use tools such as Adobe Photoshop, gen AI is built in to an extent where it’s hard to say if AI tools were used or not (if I use advanced object selection, or extend an image by 10px, in theory AI is involved, but it’s still the image I created – is that allowed?).
But if there was a rule that said all text in the game (and any images used) were 99% the product of human effort, that may allow a little more leeway, without letting in the generic low-effort stuff everyone is worried about. Similarly, I use Grammarly a lot, again, it has AI built in (I think), but all I use it for is to tell me, where I put, all the commas, in the wrong places. It would be a shame to outright ban works that have a huge amount of human time and effort just because they may have been ‘near’ AI tools.
Anyway, ultimately, my feeling is that whilst it’s lovely that the comp runners ask for opinions, it’s entirely up to them what rules they want to keep or change.
8 Likes
I will just note that in the og post, Mathbrush said he’s not going to ban ai-assisted coding or spelling/grammar checkers. It’s even in the wording of the suggested rule!
5 Likes
I think the fundamental argument is intention and effort in the context of a judged competition.
If entrant A spent weeks writing and editing the prose in their game and entrant B pushed a button to generate finished room descriptions in five minutes, do they deserve to be judged against each other?
The prize isn’t given to the AI, it’s given to the person who entered the game. Even if the AI content is brilliant, is it fair to say a person who pressed a button and handed the work over to someone else - in this case a machine - without acknowledging that fact is just as talented and worthy of accolades as a person who spent weeks doing the work?
It’s one thing if you’re creating your game for fun and not having it judged against others, but pitting a machine-generated work against those created by an actual humans for a reward or accolades is potentially shady as someone who hands in an obscure poem they found on the internet written by someone else with their own name on it for a grade so they pass a writing class - which is ostensibly in service of teaching them to write well.
The counter-argument is that there are ways to legitimately use AI in the engine of a game creatively. In a simplified way, I have built games with random text-generation routines that spit out responses that surprised me and worked very well - but it’s all built out of data and randomization routines I structured and bits of phrases that I created. I have consulted online websites that provide “meaningful baby names” as inspiration to create a list of potential entries to put in the $name slot - but that was research I did which is still curated by me; I didn’t copy/paste the entire list.
I have paid human artists for game art, or commissioned it at no charge by agreement. I have used sound clips from freesound.org that were “no attribution needed” and “free to use with attribution” but I always credit the artist or mention where the sounds came from so people don’t think “oh wow, Hanon drew that/composed that music” when I didn’t.
If you’ve got a co-creator, they should be credited; if you use someone else’s work, even work done by an artificial entity, it should be cited as a resource.
Unless you yourself built the AI, the content that comes out of it wasn’t fundamentally created by you. Someone else programmed the tool you used creatively, and in my opinion I don’t think it’s a big ask when accolades and prizes are involved that this fact be transparently disclosed.
18 Likes
I’ve often wished, when playing AI-written games, that I could just read the prompt they gave the AI (like, “Give a description of a screwdriver in the style of a text adventure, with an emphasis on a cozy atmosphere in a workshop.”). It would save time and let me see the intent.
But that made me think, it would be fun to have a ‘paint by numbers’ style text adventure that was intended for AI use but didn’t use AI. So each bit of text, instead of being real text, is just a prompt that the player is meant to feed into the AI.
AI users could actually put each prompt into the AI, whereas none-AI users could use their imaginations.
8 Likes
This gives me a great idea for cover art.
6 Likes
Fully agreed. This is basically the idea behind a labelling requirement. I try to keep it short and not starting a long discussion again, still feeling somewhat tarred and feathered.
Really a good question. In SpringThing everyone selects the games to play and review. I think there is no minimum number of games you have to play for voting. I could imagine a certain level of self regulation, because people who hate AI will not play and review AI games and the technophiles will perhaps handle it exactly the other way around, I have no idea. Perhaps even separate categories might be a solution. We have the Backgarden? Perhaps there is space for an AI Lab, but this can only be answered by the maintainers. Situation would be different if a judge has to play and review every game, because you would be obliged to judge against each other.
I think - and of course I might be wrong about this - AI could support in making good games or lend a hand in starting with authoring games, but AI alone will generate „bog-standard“ games that will not last and after some time either improve or vanish. And the brilliant games?
Photopia, Counterfeit Monkey written by AI? No, I do not see this coming in the foreseeable future. This is where imagination beats 100,000 tensor cores…
I would be surprised to see a best in show ribbon on any AI-only entry soon, but things are in constant flux…
I enjoy running and chess, and it’s understood that when I run a marathon or play in a chess tournament, I’m not competing against e-bikes and cars and Stockfish.
It thus seems entirely reasonable to me for the organizer of a competition intended to highlight human achievement to place whatever ban on AI they see fit.
That said, it seems strange to me to treat prose and illustration as Art that must be human-generated, while allowing AI-assisted coding. I can see an argument that prose has a privileged position as the “soul” of a piece of IF; I have a harder time reconciling a ban on AI cover art with allowing AI game logic (even from a pragmatic quality-control perspective: surely nobody wants to review a buggy mess.)
12 Likes
If you look at it from ease of enforcement it makes sense. Text and art are public facing and can be vetted by anyone, whereas code isn’t necessarily accessible and requires a higher amount of knowledge to spot AI use. Any code ban would be on the honor system.
5 Likes
I am a little more convinced a notch toward the idea that AI can be useful when it’s used to supplement the parser during gameplay. Most people write their game and program the rules and hand over the “Dungeon Master” duties to a parser they didn’t create to begin with.
The one application that impressed me was an “AI Zork” prototype which was fully just Infocom’s Zork that could be played as normal, but the AI only came into play when the parser didn’t understand the player’s command. Instead of giving an error message, the AI could interpret the command and either feed a more logical version back to the parser, or improvise a response that was in-character for the game.
The examples that impressed me:
PARKOUR THROUGH WINDOW - I typed this in since very few games would even have this as an action synonym. The AI was able to determine that “parkour” was a form of locomotion and substituted probably ENTER WINDOW or CLIMB THROUGH WINDOW and the game progressed without giving an “I don’t know what that means.” refusal.
LISTEN FOR HOUSE OCCUPANTS at the front door of the House - This is an absolutely daffy command that no author could probably predict and rarely is LISTEN FOR [random thing] a command that would be implemented. Instead of going “Huh?” the AI gave a response to the effect that the house was abandoned and lonely in its remote area and the only sound from the house maybe was wind whistling through the eaves. This was all correct based on the normal parameters and world-content of Zork and didn’t make up any random details that would conflict with the author-prescribed story and setting.
7 Likes
Parser games are in desperate need of some kind of autocorrect/auto-thesaurus. I’m not as sold on filling in the blanks in the world description though. For instance, the next scene (in the kitchen) has a recently-used table, and the built-in message “the white house makes no sound” doesn’t entirely agree with the AI message either. There’s too much risk of the AI contradicting the game in subtle ways that would be difficult to catch in testing.
6 Likes
tads already has an autocorrect afaik…
3 Likes
Yeah, I think ADRIFT too? I’m a little surprised it isn’t more broadly used.
It does! TYPO will turn it on in How Prince Quisborne the Feckless Shook His Title, at any rate.
2 Likes
There was a bit more to that Zork version than handling unrecognized commands - There were some valid commands I couldn’t get it to recognize, but they were minor.
Honestly I’m not wild about games (or any software for that matter) that depend on external black boxes to function. Just because it works today is no guarantee it will tomorrow.
11 Likes
Totally agree with these (I’d entered school competitions even just as a beginner chess player for fun)
To build further upon these with the case of chess tournaments and this IF Spring Thing competition:
When entering a chess tournament, you are entirely entitled to expect that your opponent is not using a computer to choose their next moves. The organizers are allowed to ban the use of computers in chess tournaments.
The point of most chess tournaments is for two humans to play against each other in each match.
I think it would be super strange if a chess player entered a tournament and said, “I’m using a computer to assist me and decide what moves to make next, because currently my chess knowledge is less than those of kindergarteners”. If you’re going to enter a chess tournament, you gotta learn how to play chess—without relying on a computer to make the moves for you.
Same should go here with IF competitions like Spring Thing.
If you can’t play chess yet and still want to enter a chess tournament by using a computer to help make your moves, the actual logical step would be: why not learn how to play chess? Instead of focusing on how a computer ban prevents you from entering all chess tournaments.
If you’re going to enter an IF competition, why not take the time to learn the writing, drawing and coding skills needed for your game?
With the analogy of chess competitions I mentioned, there’s also no ideological prejudice in banning people from using computers to make their next move, in chess tournaments between human beings.
The problem to me is a point I will make below, where I will build upon mathbrush’s existing point:
I’ve consistently found that people who tend to be most impressed by generative AI (and thus want to use it a lot) heavily lean towards those who have little to no skill in the area that they’re using the AI for. If they could code, they would code, not use AI; if they could draw, they would draw; if they could write their own stories, they would write them themselves, not use AI.
Because they have little skill in the area, they also have less of an ability to judge the quality of the output, and are highly impressed just by anything that the machine spits out.
It’s like not being able to tell the difference between a dessert from Michelin star restaurants and junk food. And then you decided to try and enter a competition at a local baking fair, by submitting a bunch of $2 chocolates and uncooked cake mix boxes from your supermarket. And not only that, but then expecting everyone else and the judges to have to review your regurgitated unoriginal junk food which you didn’t even make yourself, and asking for other competitors to be able to do the same as you are and submit more mindless prepackaged food. If you could cook and loved it, you wouldn’t be using cake mix to enter a competition.
It might be blunt but it’s true. I’m generally sick of being told by non-creatively-literate people who are impressed by AI slop that they think this is good output, when it’s much less worthy of being proliferated or good enough to be entered in competitions, let alone actually rivalling the work of artists, writers and programmers.
As other expressed too, I also strongly feel that the most basic human-made thing would have more meaning. In this case, entering the most basic homemade cake to share with others – even slightly burnt – would at least have come from the heart, and the person who made it can use the experience to make more cakes themselves in the future.
With using gen AI, it doesn’t help you get better at writing, drawing or coding. You never have to learn how to formulate your ideas, how to creatively make use of analogies; you never learn about color schemes, proportions and composition if you keep getting the AI to spit images out for you; and you don’t learn how to debug, refactor and optimize with the code that AI spits out at you. Similar with using a computer to help determine your next chess moves–it could keep telling you what move to make, while you never even have to learn what all the pieces in chess are, and how each piece can move around the board.
You don’t need fancy tools to start learning how to play chess, write or draw.
I think it’s totally reasonable to ban the use of genAI by having a rubric of factors which are likely indicative of AI use, and rejecting entries that reach a certain threshold of those factors.
21 Likes
Personally I dislike all the flimflam and hype about having to have a cover picture as well as sound and graphics for the game. I THOUGHT IF was only words and your mind/brain supplied the rest. When you read a book, do you only read those that have pictures and/or sounds. It reminds me of my grandchildren.
Make the rule ‘you can’t have any pictures or sounds’ to be fair to everyone.
As has been said, where does it leave the person who isn’t an artist or composer? Don’t throw the baby out with the bathwater.
On the other hand, if people feel they MUST add pictures and sound because their writing isn’t enough, then let them but reduce their score by some fixed percentage.
The point of Spring Thing is to be the most experimental and freeform of the major competitions, so I don’t see the point of banning graphics and audio completely. If someone thinks illustrations enhance their writing, let them use 'em.
The reason for the proposed AI ban is to address a specific problem that’s arisen, and I don’t think we’ve seen a specific problem coming from the use of graphics and audio.
11 Likes
It would be interesting to participate in a No Image/Cover Image Jam, but that would be for a new conversation.
7 Likes
Visual novels fall under the category of interactive fiction. User interfaces are a part of interactive fiction – both of these things consist of or contain images. Making the rule “you can’t have any pictures or sounds” isn’t the measure of fairness you think it is, it’s incredibly reductive and excludes an entire category of creators who deserve a spot in the Spring Thing as much as anyone else.
You’re saying “don’t throw the baby out with the bathwater” – but illustrations are not inherently undesirable in interactive fiction, just as they’re not undesirable in regular novels. They’re additional, not required, same as a cover image. If you cannot make an illustration, a cover image, a soundtrack – it’s not like you’re losing out on some major component of IF. The writing is still there.
Furthermore, the idea that you should lose score simply for wanting images undermines the whole point of Spring Thing: to make more experimental and freeform games, and for the judging to be more freeform in turn with the audience ribbons. Banning images and audio wholesale is not a solution nor even getting close to the problem; the ban is not a matter of fairness as in leveling the playing field, it’s about fairness in that it requires integrity through the creative process and not feeding the work of other writers, artists, composers, and coders into the environment-damaging plagiarism machine.
12 Likes