A self-watching model of analogy-making and perception

Abstract

This paper describes Metacat, an extension of the Copycat model of analogy-making. The development of Copycat focused on modelling context-sensitive concepts and the ways in which they interact with perception within an abstract microworld of analogy problems. This approach differs from most other models of analogy in its insistence that concepts acquire their semantics from within the system itself, through perception, rather than being imposed from the outside. The present work extends these ideas by incorporating self-perception, episodic memory, and reminding into the model. These mechanisms enable Metacat to explain the similarities and differences that it perceives between analogies, and to monitor and respond to patterns that occur in its own behaviour as it works on analogy problems. This introspective capacity overcomes several limitations inherent in the earlier model, and affords the program a powerful degree of self-control. Metacat's architecture includes aspects of both symbolic and connectionist systems. The paper outlines the principal components of the architecture, analyses several sample runs and examples of program-generated commentary about analogies, and discusses Metacat's relation to some other well-known models of analogy.

Self in NARS, an AGI System

“…For example, many works under the name of “metacognition” (Cox, 2005 ) have knowledge about various algorithms within the system itself and use this knowledge to select a proper one for the current problem. Although this process is self-reflective by nature, the systems typically does not have an explicitly represented “self.” Instead, the processes are separated into “object-level” and “meta-level,” where the latter monitor and control the former (Cox, 2005 ; Marshall, 2006 ).…”

“…In general, controlled processes are more flexible and adaptive, while automatic processes are more efficient and reliable. Such a model often uses meta-level processes to regulate object-level processes (Cox, 2005;Marshall, 2006;Shapiro and Bona, 2010;Rosenbloom et al, 2016), and such works are also covered in the study of machine consciousness (Chella et al, 2008;Baars and Franklin, 2009).…”

“…Although this process is self-reflective by nature, the systems typically does not have an explicitly represented "self. " Instead, the processes are separated into "objectlevel" and "meta-level, " where the latter monitor and control the former (Cox, 2005;Marshall, 2006).…”

40 years of cognitive architectures: core cognitive abilities and practical applications

“…It is universally agreed that symbols (labels, strings of characters, frames), production rules and non-probabilistic logical inference are symbolic and distributed representations such as neural networks are sub-symbolic. For example, probabilistic action selection is considered as symbolic in CARACaS [251], CHREST [476] and CogPrime [189], but is described as subsymbolic in ACT-R [321], CELTS [151], CoJACK [434], Copycat/Metacat [350] and iCub [459]. Likewise, numeric data is treated as symbolic in CAPS [266], AIS [212] and EPIC [279], but is regarded as sub-symbolic in SASE [586].…”

Interactive Effects of Explicit Emergent Structure: A Major Challenge for Cognitive Computational Modeling

“…There are, of course, a number of non‐connectionist computational models whose design is such that chunks are gradually formed from the input data and these chunks influence subsequent processing and representation of new input. For example, the “active‐symbol” computational models of analogy‐making initially developed in D. Hofstadter's research group in the late 1980s (French, ; Hofstadter, ; Hofstadter & Fluid Analogies Research Group, ; Marshall, ; Mitchell, ; Mitchell & Hofstadter, ) are examples of this type of model in which interactive emergence was a core design principle. In these models there is a constant interaction between representational and computational levels, between bottom‐up and top‐down pressures.…”