Intelligent Machines in the Near Future?

Scientists and programmers are beginning to look towards the human cortex as a starting point for creating intelligent machines. In particular, Jeff Hawkins is leading two ventures, the Redwood Neuroscience Institute and Numenta, that plan on creating a high level software representation of the cortex's layers in order to solve traditionally hard problems (voice and visual recognition, weather prediction etc...)
This video shows Jeff giving a talk about the neocortex, its structure, and the ways that computer models can be built on top of that. They even have an example of a model that allows for a computer to recognize shapes into 96 different categories with an amazingly simple underlying architecture.
This kind of research fascinates me because of the amazing possibilities of these techniques of 'solving' problems that are currently impossible. Imagine connecting one of these 'pseudo-cortexes' to live data about the stock market, and having the ability for a machine to predict stock prices for example.