Overview
This game tackles the topic of appropriation by combing a medium that cannot exist without appropriation – AI generated images – and appropriating the rules of pictionary and the rules of drawing telephone games such as Gartic Phone. In this game however, the creation of pictures are completely left up to the AI with the players only writing the prompts and guesses.
Rules and Gameplay
- The game can theoretically be played by any number of players, but at least 4 players recommended. Additionally 1 person is needed to be the game master.
- The game takes place in a chatroom or a Discord server, and the game master is in charge of managing the players’ prompts and images, as well as stringing the final results together.
- At the start of the game, each player writes down a phrase of no less than 5 words, and send it to the game master.
- The game master then sends each phrase to the next player, and each player enter the exact phrases into an AI image generator (craiyon is used for playtesting). Each player chooses an generated image and send them back to the game master. The players must not alter the images in any way, and it does not matter if the generated image does not resemble the input prompt.
- The game master then sends the images to the next player, and each player describe the content of the image as reasonably detailed as possible (they are also allowed to twist and remix their own interpretation of the image if they wish).
- The players sends their new descriptions back to the game master, who pass them to the next player.
- After at least as many rounds as the amount of players have taken place (the game can go on for longer if there are few players or if the players wish to), the game master presents the result of the telephone game.
Artist’s Statement
When I think about the word “appropriation”, my mind immediately jumps to the current hot and controversial topic of AI image generators and how all of them are based on the countless amount of harvested data from across the internet to generate new images based on these data. As a hobby digital artist, I definitely agree with the many ethical problems raised with generative AI, especially in how their data are overwhelmingly gathered and used without the original artists’ or photographers’ consent. However, I also think that there is one area that AI is almost perfectly suited for: making memes and “funny internet pictures”. Since memes already appropriate by nature, AI mashing them together to create something new is basically the next step in memes’ evolution. One particular game that lends well to this are the pictionary type telephone games such as Gartic Phone or Jackbox Civic Doodles, which is the other part of the game I appropriated. I think letting AI do the image making spoofs on the concept of a drawing game and also enhances the collaborative nature of telephone games – in this case collaborating with all the unwitting data used to make the AI image generation possible. While not explicitly stated in the rules, I trust the players will try to write the most ridiculous prompts and choose the weirdest AI generations. After all, to laugh at the results is the point of the original telephone games.
Originally, I wanted the players to draw the pictures themselves, but only closely following that the AI has generated; as I thought this would create extra variance in the game to derail the telephone chain. However, after observing that there are some aspects of AI generated images – the overall blurryness, the weird human hands and the smudged text – that cannot be easily conveyed by drawing in a reasonable timeframe; and the fact that the AI can and will misinterpret the input and generate something completely unexpected, I decided to let the AI completely take over the creation of images. Unfortunately, I was unable to find a online telephone game that allowed direct image uploads, I had to switch the setup to be somewhat clunky and require a game master.
There was another idea I toyed with when iterating with the game, that being instead of needing the players to input descriptions or prompts, the players would put the generated images into AI image identifiers and use that to generate descriptions for the images. I thought it would be a fun parody on the endless cycles of online data harvesting and the corruption of digital data and their meaning. However, currently the AI image identifiers can only use words and not sentences or phrases, this idea had to be scrapped. But I think this idea may be revisited with our current rate of advancement in AI.
Playtest Results
2 sample telephone game chains
