ASSC4 Abstract: Evolvable architectures

ABSTRACT FOR ASSC4 Conference THE UNITY OF CONSCIOUSNESS:

BINDING, INTEGRATION, AND DISSOCIATION

Universite Libre de Bruxelles

JUNE 29th - JULY 2nd, 2000: BRUSSELS, BELGIUM

TITLE:

Evolvable Architectures For Human-like Minds

AUTHORS:

Aaron Sloman,
School of Computer Science, University of Birmingham
A.Sloman@cs.bham.ac.uk

Brian Logan,
School of Computer Science & IT University of Nottingham
bsl@cs.nott.ac.uk

Matthias Scheutz
School of Computer Science, University of Birmingham
M.Scheutz@cs.bham.ac.uk

Abstract:

Our multi-disciplinary project, combining empirical, philosophical and computational theories, aims to explain many human, animal and robot phenomena, e.g. varieties of affect, perceptual functions, reasoning, learning, and varieties of consciousness.

We regard intuitive concepts of mentality as "cluster" concepts referring to implicitly presupposed virtual machine (VM) architectures.

These pre-theoretical architectures are very crude approximations to diverse actual VM architectures, which vary during development, across species, and after damage or disease.

Many disagreements on topics like 'emotion', 'consciousness', 'intentionality' arise because researchers focus on different sub-clusters of capabilities and mechanisms within such architectures.

Researchers studying different aspects of mind are often unaware of what they are ignoring, like the proverbial blind men trying to describe an elephant.

Powerful theories require knowledge about evolution, ethology, neuroscience, psychology, phenomenology, and feasibility results from AI.

Our draft architecture-schema for human-like minds includes reactive, deliberative and meta-management layers, which evolved at different times. The layers operate concurrently and asynchronously, interacting in a variety of ways, including production of emotions and several kinds of learning, perception, motivation, etc.

Human-like self-consciousness requires the third layer, though more primitive sorts are implemented in older layers. Future work includes understanding many discontinuities and trade-offs in the evolution of mental architectures.

REFERENCES

Papers on our work can be found here: http://www.cs.bham.ac.uk/research/cogaff/