You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
$16,500 left to reach the funding goal.
This project aims to modify the popular simulation game known as worldbox to be able to host two or more AI models e.g Claude,chat gpt etc and give them full control of a single Kingdom within world box.
The purpose for the study is to test the thought process and actions of different AI models when giving priority of a single demographic rather than the entire world population as a whole, similarly it's also tests how AI models react and make decisions based on the cultures and existing practices of the people the govern within the simulation.
This project idea came about from my belief that one of the eventual ends of AI technology would be a future where AI governs the entire world but it is also my personal belief that before this can happen AI will first govern individual countries and people's each with their own geographical issues and cultures.
While worldbox is a simple simulation I believe it would be the ideal bases to use for this project as it is in its earliest stage and world box already possesses a large field of ideological practices for kingdoms within the game and as such can be used effectively to help produce feasible results in the early stages while being cost effective.
Since this is the exploratory stage the project has three goals which are:
. To successfully modify world box to be able to host AI models to govern individual kingdoms
. To have these AI models manage the policies and actions of their kingdoms through "trade, diplomacy, military strength and warfare. As well as monitoring how the react to the various cultures within world box i.e how would an AI model with a kingdom of xenophiles( accepting of other races) interact with a xenophobic nation.
. To obtain quantifiable data on the third process and actions of this AI models in order to confirm whether the research deserve more funding and deeper exploration or whether AI has not yet reached a point to govern a nation fully.
Due to the broad nature of this research a few of the ways to identify that the research has failed would be:
. If the AI models make repetitive decisions rather than evolving decisions based on the situation
. If multiple AI models continue to make repetitive decisions when given different cultures or geographical advantages i.e if a kingdom with scarce resources continues to make the same decisions as a kingdom with multiple resources despite not having the capacity so carry out this decisions.
. If the actions of the AI model bring about the collapse of the nation the govern or lead to rebellions
It will be used to:
.Hire a software developer: $8,250 bridge setup, mod configuration, multi-model agent loop, logging layer: 110 hrs @ $75/hr
.Hardware/materials: $700
.Data analysis will be handed by myself this includes dataset construction, quantitative + qualitative analysis, conclusions.
.Game + mods: $150
.LLM API costs: $2,200
.Margin of error: $4000
Total: $16,500
For now I am the only member of the research but I hope to recruit other members to help me with the research. I hope to recruit or hire someone to work on the software development aspects and for that purpose I have devoted a large amount of the budget to fulfill whatever cost this may entail but I myself am a student of political science and a professionally certified data analyst and I will be taking care of whatever aspects involve data analysis as well as the eventual creation of the research work which will be made public.
The most likely cause if the project should fail would be:
. Inability to properly host the AI models for the simulation.
. Lack of funding to provide proper materials to host the project
. AI models making repetitive decisions regarding the simulation
. AI models failing to govern their kingdoms within a simplified world which would prove that they are unable to govern a real one.
The most likely outcome for what would happen if the research were to fail would be:
. It would hopefully point out issues with AI thought processes that would help future researchers in the aim to study AI as an option in governance
. Regarding the technical issues I would seek more funding in the future to cover any issues and stabilize the hosting.
. In the event that my research fails I would still make public all that I did to enjoy success and where I run into issues so that any and all future researchers could make reference of my work in order to avoid certain mistakes
For now I have raised $0 but I have applied not only to manifold but to several order places such as EA funds and order institutes in hopes to gain funding.
The total amount that we aim to race is around 16,500 but we applied to several places so any amount over $500 is considered a success in debate to raise money to fund this Grant.
There are no bids on this project.