CABot1: A videogame agent implemented in fLIF neurons

CABot1 is an agent in a simple videogame that assists a user in the game. Like the user, it views the game; it takes commands from the user, the commands are used to set goals, and the system interleaves all of these processes. Crucially, CABot1 is implemented entirely in simulated fatiguing Leaky Integrate and Fire neurons. The long term goal of this line of research is to develop a system that can solve the Turing test. The author believes that the best approach to building such a system is to mimic humans at a neural and psychological level. CABot1 makes use of the Cell Assembly hypothesis as the neural implementation of symbols, and as such, CABot1 processes symbols. Developing increasingly complex systems, like CABot1, that are grounded in an environment should lead to a system that is capable of grounding its symbols, learning new symbols, rules and relationships. If such a system is allowed to learn enough, it will be able to pass the Turing test. CABot1 does not pass the Turing test, but it does demonstrate how vision, language, and planning can be implemented in a neural system and integrated into a useful system.