A.I. learned financial modeling

The holy grail for certain financial analysts is to be able to develop a comprehensive economic model that will take-in massive amounts of data and improve their odds at making forecasts. In order for this models to be meaningful we must enter the world of AI. Neural networks, genetic algorithms and extensive data mining are some of the tools needed to build such a daunting application. Typically it is very expensive to develop something meaningful, but there are some alternative options out there for the serious individual investor. Recently I found an application called RapidMiner which seems to be very powerful, modular and Open Source. I am currently trying to figure it out and brushing up my XML & JAVA in order to understand some of the concepts. It is very interesting, for example you can feed it data from TraderXL in order to construct your models.
Although it is not very encouraging that Goldman Sachs darling, the Global Alpha fund(9 billion) which trades are determined by computers and mathematical models, is down 16% for the year last time I checked, I have faith that computer modeling is a valuable tool for conducting meaningful forecasting.