Overview of the IBM Machine Intelligence Project

The first part of the talk (Wilcke) will give a general overview of the project. It will outline the  goal of building a system which uses fast associative reasoning to mimic certain aspects of human intelligence. The project is using the principles of Machine Intelligence (MI) - which are very different from those of Machine Learning. We will give a overview of the MI principles and the structure of our neural network. The project has four major components, which are the biological model, it's implementation as algorithms and software (called CAL), the building of a new 'neural supercomputer', ESCAPE 9000 and of several robots driven by Machine Intelligence. We will give a quick status overview of the status and current results from the four areas.

The second talk (Ozcan) describes in some detail the neural model used. We know that feedback is very important to deal with the real world and therefore include the thalamus in the model, in addition to that of the neo-cortex. The thalamus serves both as a router of signals and as a scratchpad for sharing information between regions.
Video of an early (2 region) version of CAL and of one of the robots in action will be shown during the poster sessions.