<< JointMeta << Science << Machine Learning

painting of a clown>> (Experiment) Designing a Generic Unsupervised Learning Component: Knowledge dll


>> Objectives

To discover, through experimentation, the foundations of AI leading to software that appears to learn like a human being and as an eventual aim, to understand consciousness. Also considered here is the application of AI in video games, and what makes good videogame AI.

Artificial intelligence too often refers to computer programs that solve problems that are not mathematical in nature, or that were traditionally solved by human beings. Here, I would like the aim of AI to include skilled computer programs, but also to look beyond that structured approach at the idea of creating life, or with regard to video games, activity that feels intelligent (has apparent consciousness), rather than feeling like a series of optimized equations.

This experiment focuses on unsupervised learning - to form natural groupings or clustered patterns without explicit instruction. The ultimate aim is to create a system that can gain insight into the nature or structure of data independent of any former background knowledge. In particular to create a problem solving component that can then be applied to all sorts of situations.

Another aspect to be explored in this project is the idea of a linear memory, which does not categorize data in any way but stores a list of experiences and from that it derives the next step, by comparing the current sensory data with the data already collected.

More specifically, once the creation of the software is complete, it should be applied to a game in a 3d environment to judge its performance as both a problem solving system and a source of entertainment.

>> Next

Email Joint Meta 2006

Creative Commons License
This work is licensed under a Creative Commons License.