Banking secret, superbillionaires and IA

From: Fabrizio J. Bonsignore (fbonsignore_at_beethoven.com)
Date: 10/23/04 

Date: 23 Oct 2004 15:00:07 -0700

OK, so we have a bunch of superbillionaires playing games with ants, I
mean, people, because they are not under check. They can buy or
threaten everybody, leave a trace of millionaires and then send
Leprechauns to recover their treasure. They can keep themselves
anonymous and satisfy all their whims because they have so much money
that nobody counts, at least not poor people, like those who have less
than one billion dollars. But the banking secret is sacred, so there`s
not much to do...

To solve this problem intelligent agents are the perfect solution.
Banking accounts are kept in banking systems distributed around the
world. These databases can be read by specially designed tamperproof
interpreters hosting mobile intelligent agents. Each agent is composed
of executable code, knowledge bases and data, all of which travels
through the wire, in encrypted form, according to the agent`s needs
and decisions. The hosts itself provide enough intelligence to manage
transit and distribute knowledge to other hosts (availability, QOS,
agent successes, resources, multitasking, encryption, etc.). The
agents are composed of a hybrid AI system comprising NN, ES and FL
analyzing banking data to uncover patterns that denote suspicious
money, like sudden big amount transfers, name changes, steady
accumulation of resources, account depletion and closing, name
changes, unexpected deceases, persistent transfers to the same
account, etc. The agent establishes levels of suspiciousness and
decides when and where to transfer itself to follow resources or when
to spawn or call other agents to monitor suspicious accounts. An
initial learning period would determine (by using a learning mechanism
and training) patterns of usage that corresponds to legitimate, honest
accounts; this can be done with human intervention and counseling,
though crosschecked with other agents to avoid deceitful training.
Once an agent finds a suspicious account it locks on it and performs a
traversing of the network with its data until it either finds a point
of accumulation (not necessarily an account, but a human
individuality) or gives up the hunt when resources are found to follow
normal usage. Hosts can maintain sveral hypothesis and serve to guide
agents in their traversing. Once a match is confirmed, it is
published, in a distributed manner to avoid hiding the information.

Several architectures can compete according to successes or failures,
though once an agent is installed it should not be able to be modified
to avoid tampering with its data.unsucesful architectures (strains)
can be dropped relatively to more successful, new architectures. These
ideas can be further refined, of course.