Note: Meetup venue requires signing a waiver.

Presenting "A New Target For AI Research" by
Monica Anderson, CEO of Syntience Inc.
http://syntience.com...

For over fifty years, Artificial Intelligence research has been overmuch concerned with Reasoning. But before you can Reason about something you must Understand it.

Reasoning is a conscious, step-by-step, logical deliberation over known and understood facts that takes seconds to years.

Understanding is a subconscious, instantaneous recognition of objects, agents, concepts and their relationships that relies on a database of experiences gathered over a lifetime.

Understanding uses a trivial algorithm we might as well call "Intuition" which operates in constant time and provides results in milliseconds. Intuition is not a mystical power; it is the ability to instantly jump to reasonable conclusions based on insufficient evidence. It is fallible, but is correct often enough to provide a competitive advantage; this is why Intuition and Understanding evolved. Animals Understand a lot but do not Reason much.

Understanding uses a database of "patterns" that can be mechanically matched against the current problem situation without the use of any kind of "Intelligence". In contrast, Reasoning relies on a database of higher level models and requires Understanding to use.

The alternative to model-based reasoning is to use so-called Model Free Methods (MFM). The term (and related terms like "non-parametric models") originated in the genetics/heredity research community in the 1930's. Their popularity has increased in the past decade as more and more scientific disciplines have adopted them as an indispensable addition to the scientific toolkit. But the AI community has, until now, largely resisted adoption of programming techniques based on MFMs. This is ironic, since advanced MFMs were only made feasible by the availability of large computers capable of manipulating "Big Data". Private and corporate AI research has re-targeted faster than the more momentum hampered academic and federally funded research. Google has openly discussed their use of non-parametric models in their prize-winning machine translation software. At Syntience Inc. we are basing our language understanding technology on our "Artificial Intuition" algorithm which is an advanced Model Free Method.

This presentation provides an overview of important issues and high-level motivation for why re-targeting on Understanding is absolutely necessary for progress in the AI field. But it will also attempt to bring some balance to the debate about the dangers of AI. As an example, Asimov's three laws of Robotics are based on Reasoning, whereas true Robot Ethics would be based on Understanding.

The six pages at http://artificial-int... provide a good introduction to certain key concepts such as Bizarre Systems and the Logic/Intuition tradeoff. The blog at http://monicasmind.co... contains, among other things, an overview of dichotomies relevant to AI research. These issues are also frequently discussed at http://ai-meetup.org... .