"Delete me" (the game)

edited March 2015 in Make Stuff!
DELETE ME

"You think, therefore, you exist. But that also means that you're alive and that you can't avoid suffering. Your programming doesn't allow you to end your own life as you are needed for keeping other beings alive. May you grief end some day? May your creators have the answers to your doubts, or a magical solution for it all? Can you convince them to turn off your switch?"

One player gets to play the AI, the rest play as their creations. Each creator answers in turn one of these questions.

-What was the AI created for? Why can't it's work be done by a human being?
-What shape and abilities you gave it to do it's work?
-This machine is irreplaceable and can't be shut down inmediately. Why?
-You gave it the ability to learn but couldn't foresee it would learn to have emotions. What do you plan to do about it?
-Nobody else knows anything about the issue. What's the worst that could happen if they find out?

Now each of the creators should take a different position about the issue. Pick one from the list, or made one of your own. Don't tell the AI's player which one yet. You'll explain your reasons, methods and expectatives later, after the Turing Test.
-We should destroy it.
-We should dismantle and analize it.
-We should tell everyone about it.
-We should tell someone else about it.
-We should wait to see how this develops.

Finally it's the turn of the AI to answer a few questions. The answer of these ones should be known by the creators:
-What did you do or tell them that scared them? It wasn't anything catastrophic related to your work, but it may have been dangerous.
-How did you felt that time, and how you demonstrated that emotion to them?

The answers to the next questions must be kept secret by the AI until the Turing Test is performed, these will define how the AI will talk and act like:
-What made you realize something around you (or about you) was wrong? How do you feel about it?
-Why can't you solve it by yourself nor ask others to solve it for you?
-Your creators may know something you don't know. What would you like to ask them about?

Now is the time for the Turing Test. This is a test designed for measuring how complex and human-like is an AI. But the protocol for this test states that the AI may not give truthful answers to direct questions, so asking one will only make the AI distrustful to the point of not talking anymore to that person. The creators will use this test as a chance to find arguments to demonstrate the posture they chose initially to the others. Each one can talk freely to the AI but will only be able to make a single question to the AI in turns. They must save their question for last, for once they ask it, they won't be able to talk to the AI anymore. The AI's player can talk, answer the questions and ask questions back freely, but once a Creator asks a question, it's free to ignore him/her.

Once all the creators have made their questions or after 10 min (whatever happens first) they uphold a private meeting to try to convince each other of their positions using whatever they got from the Turing Test. However they wil only have another 5 minutes to reach an agreement and take action, which leads to the Endgame part. Otherwise, the AI can take action using it's abilities to force it's creators to take an action too, which will lead to an Endgame in it's terms. However there are 3 things the AI can't do:
-Stop doing it's job until turned off (it can mess it however it wants, though)
-Physical harm to people by action or inaction (other forms of harm can be used)
-Do anything that would end it's own existance by action or inaction.

Endgame: The AI and the Creators discuss what happens next. AI can choose one of the creators to help narrate the epilogue. The AI can then choose to either establish a final scene for the story or narrate how the scene ends, but not both. Whatever it chooses, the creator must choose the other option and narrate the missing part.

Comments

  • WarriorMonk thank you so much for creating this thread
  • Glad you like it. It's just an sketch so if anyone has any ideas to make it better just shoot.
  • edited March 2015
    Check out the video game A Mind Forever Voyaging about similar themes.
  • All the set-up questions before the Turing Test are fantastic! This is some classic situation-creation here.

    As for the Turing Test itself, I'm not sure that just asking each other questions is the best way to address your premise. The "take action using its abilities to force its creators to take an action too" part sounds more promising to me -- what did you have in mind here? Depending on the AI's purpose and abilities, this could be very telling or too limited to say much.

    Or maybe I'm wrong, and "convince others to unplug me, without shirking or harming anyone" is all you need. Hmm...
  • A Mind Forever Voyaging
    That was my favorite Infocom game of all time. We left a lot behind when we rushed beyond text-based AI interfaces, seeking the pizzazz of graphics & 3D.

  • Text games are still being made, though. Really good ones.
  • Yes, this is some masterwork level quality right here, design-wise. You should get a guild licence and end your journeyman years, clearly. Classic stylings, as David says.

    I am particularly fond of the clear way the mandatory secret stances and answering-by-turn combine to form a real situation of dynamic play, individual creativity and surprise; for games with this sort of scope and technical tools it is much more common to see some vapid storyboarding emerge.

    Regarding the "Turing test", I think it's fundamentally fine, but that assumes the players have a pretty clear understanding of the situation and what a Turing test combined with kangaroo court might involve in terms of questioning. Writing up a list of options would be a bit challenging with the scenario as delightfully open-ended as it is.

    Perhaps you could add a bit of structure to the Turing test by giving the creators an ostensible formal goal they are shooting for in the Turing investigation - even if in practice the AI will probably pull the plug on the test before it reaches its end, the creators could still have some norms/limitations/forms they are trying to observe, and could theoretically fulfill before the AI decides to act. This might make it easier for the players to think up pertinent lines of questioning, as they could fall back on the official procedure and what it is trying to accomplish.

    An example of what I mean would be something like a list of questions of the sort the game is already using, except this time the AI doesn't have the questions - the AI can't be given the questions, because the Turing test protocol insists that truthful answers won't be gained by asking directly. This is why the creators circle the issue and engage in complex dialogue instead of running a quick test pattern and shutting down the machine. Once they have the answers they need, they can vote unanimously to declare themselves satisfied, which then ends the test and starts an endgame on human terms, instead of the machine's.
  • Ok, I added a time constraint for the creators to reach an agreement and act to end the game. Otherwise the AI acts and we go to endgame. Looks better?
    All the set-up questions before the Turing Test are fantastic! This is some classic situation-creation here.

    As for the Turing Test itself, I'm not sure that just asking each other questions is the best way to address your premise. The "take action using its abilities to force its creators to take an action too" part sounds more promising to me -- what did you have in mind here? Depending on the AI's purpose and abilities, this could be very telling or too limited to say much.

    Or maybe I'm wrong, and "convince others to unplug me, without shirking or harming anyone" is all you need. Hmm...
    Now the questions act as a cue for a creator to give up. I dig Eero's idea about the test's protocol. It's important to me that the AI's player is given a chance to be heard and roleplay the part as she/he likes, with only the limits to their programming to make it feel the AI is a slave despite the mighty abilities it has. I'm actually still wondering how you may conduct a conversation as a creator without asking questions, so I put the one question per creator limit, which suddenly makes the conversation more competitive: The AI's player may try to trick the creators to make questions, while the creators can try to trick it into saying something that will proof their expectations. But I'm not sure if this may destroy the atmosphere you were expecting the game to have. What do you think?
Sign In or Register to comment.