I've just submitted my proposal for my final year computer science project, entitled:
Centralised and Individualised Machine Learning in a Multi-Agent Simulation
Sounds cool eh? You can read the detailed proposal here.
The above isn't too technical a document, but if you're just too lazy to read it, here's a quick overview:
I'm going to create a virtual world, and fill it with 'people' that are capable of interacting with the environment and one another. I'm then going to give these people one of twenty of so 'personalities' that I've defined, to give the world a bit more flavour and sense of reality. These are my 'dumb' people.
I'm then going to introduce some 'intelligent' people that will originally know nothing other than how things are currently going, and how they want them to go. They'll use reinforcement learning (a type of machine learning/artificial intelligence) to get from A to B. These guys will then interact with the dumb people as they see fit.
Finally, I'll have some kind of 'chief' person that all the intelligent people will report back to. It'll therefore be the most intelligent person in the virtual world, and will be able to give advice/issue instructions to the intelligent people.
Read the next paragraph
The example I gave in the document is a super intelligent police chief, controlling intelligent police officers, looking after a dumb populace - but I'm not definitely going to go with that. So, any other suggestions of scenarios that I can put my three types of people into would be Extremely Welcome. Seriously guys.
Also, any other feedback, experiences in this field, technical advice, or links to useful documents/books/resources/code would also be appreciated.
Xx
Hello, world!