If you're new to this stuff, here is a "primer" on how they work from Mike Ruppert. Yes, that Mike Ruppert. I've said before Mike was a bit insane. You have to be insane to even contemplate this sort of thing let alone investigate it! I wouldn't touch this with a ten-foot pole. The formatting will be funny here as I'm cutting and pasting.
Original source:
http://www.fromthewilderness.com/free/ww3/012005_ptech_pt1.shtmlIt combines datamining, artificial intelligence, and
"interoperability," the capacity for one program to read, operate,
and modify the source codes of other programs. Datamining is a
technique for detecting and extracting meaningful patterns hidden
within vast quantities of apparently meaningless data.
Programs based on datamining are powerful analytical tools;
finding meaningful patterns in an ocean of information is very
useful. But when such a tool is driven by a high-caliber artificial
intelligence core, its power gets spooky. The datamining capability
becomes a smart search tool of the AI program, and the system
begins to learn.
With neural networking, software has become much smarter than
it had been. Now it can perform multiple, related operations at the
same time through parallel processing; now it can learn from
setbacks, and use genetic algorithms to evolve its way out of
limitations. Now it can respond to more kinds of data from the
electronic environment, including "fuzzy" values that don't come in
discreet numerical packages. This kind of computational power
supports an inference engine that can digest the mined data into
results that are not only descriptive of the system's present state
but predictive for imminent and, to some degree, even middle
-term outcomes. That's why the same family of programs that
does enterprise architecture, which is descriptive (and
prescriptive if you take its descriptions as a mandate for cutting
costs by firing people - "process management"), comes to include
risk management software, which is predictive of the future. It
extrapolates from current trends in a more than quantitative way.
Conventional electronic surveillance finds patterns in the data of other instruments; [this software] can exploit the patterns it detects and extrapolate future probabilities.