Webmind: A Complex, Intelligent Software System Applied to the Analysis of Financial
Information
Atuhor: Ben Goertzel
Webmind is a Java-based software system that evolves its own "digital
intuition," which it uses to pose and answer questions. It deals with textual and
numerical information on an equal footing, freely making generalizations that span
different types of data. It bears some resemblance to previous AI frameworks such as
neural nets and genetic algorithms and semantic networks, but is fundamentally different
from all of these.
Webmind integrates reasoning, language understanding and production, pattern
recognition, nonlinear rediction, concept learning and creative idea origination and
self-control into a single integrated software framework, consisting of a massively
parallel population of static and dynamic information agents that continually recompute
their relationships to other agents and act accordingly. Queries put to the system are
transformed into agents that can take advantage of Webmind's self-evolving structure to
discover context and concepts on behalf of users and their applications.
Webmind forms its own intuitions by dynamically "grounding" linguistic
concepts in terms of its own, non-linguistic, experience. So while Webmind does not have a
general intuitive understanding of our world, it can make human knowledge a part of its
own world, that of data and text feeds. It is the first AI system to create concepts that
apply across these different contexts. This is the qualitative difference between Webmind
and other AI systems, and why Webmind has the potential to bring AI beyond the role of a
specialized adjunct to unintelligent computer systems, and make it the heart of computing
itself.
In this talk I will describe the Webmind architecture in mathematical and conceptual
terms, and then summarize two of Webmind's applications to financial information
- market prediction, based on an integration of numerical data and information derived
from online news
- market prediction, based on an integration of numerical data and information derived
from online news
- news filtering, in which the system reads a variety of news from the Net and selects
only the news relevant to particular financial markets
I will also discuss the internal dynamics of Webmind, and the homeostatic mechanisms
that Webmind uses to control its own dynamical regimes.
WWW-site:
http://goertzel.org, http://intelligenesis.net
|