For the past year or so I've been heavy into Cognitive Modelling and Adaptive Reasoning.

The small collage of screenshots to the right represent some of the early tool building stages of my work. There's tons of AI related libraries out there but I've always preferred building my own custom libs. The custom neural net libs I've written have evolved much from the early tools shown on the right. These days building a "blank" intelligent agent with general and domain specific reasoning is still a challenge but having a nice gui driven toolbox helps big time!

Everyone has their own style of thinking, but I tend to be very visual. It's pretty neat to watch a near real-time 3D render of your neural mesh evolve as it loads in the Core Knowledge Model but actually creating a simple 3D world for your AI to explore is beyond cool.

		Here on the left is an early version of a crude AI, created using the above toolset, wondering through a simple sandbox I created, a virtual world of sorts. The smaller view in the lower right represents what the agent actually sees as it explores. By the way, I chose Ogre as my rendering engine. Early on I was writing very low level OpenGL code but that was so tedious and taking just tons of time. Ogre is C++, multi-platformed, open source, has a very rich library, and integrates easily into Visual C++. I'm still a little ways off from showing some of the latest visuals of the project but things have definitely come a long ways from the early work shown above. If nothing else, hopefully you've been inspired a little...