A framework for organizing the many disparate capabilities required of synthetic cognitive systems is proposed as a basis for assessing the status of existing and proposed cognitive architectures and systems, as well as a measure of progress towards human-level machine intelligence. This framework divides the “ingredients” of cognition into six dimensions. Capabilities within these dimension are organized roughly according to increasing levels of capability. The cognitive dimensions and their capabilities are described here as the basis for assessments of existing architectures provided in a companion paper.
As a cognitive architecture, ICARUS shares many aspects with other systems and the recent proposal for a standard model of human-like minds. But the architecture also commits to a unique combination of additional assumptions that are important. This paper discusses these aspects and proposes them as part of the standard model of human-like minds.
We are developing Companion Cognitive Systems, a new kind of software that can be effectively treated as a collaborator. Aside from their potential utility, we believe this effort is important because it focuses on three key problems that must be solved to achieve human-level AI: Robust reasoning and learning, performance and longevity, and interactivity. We describe the ideas we are using to develop the first architecture for Companions: Analogical processing, grounded in cognitive science for reasoning and learning, a distributed agent architecture hosted on a cluster to achieve performance and longevity, and sketching and concept maps to provide interactivity.
In this article we present DUAL-PECCS, an integrated Knowledge Representation system aimed at extending artificial capabilities in tasks such as conceptual categorization. It relies on two different sorts of cognitively inspired common-sense reasoning: prototypical reasoning and exemplars-based reasoning. Furthermore, it is grounded on the theoretical tenets coming from the dual process theory of the mind, and on the hypothesis of heterogeneous proxytypes, developed in the area of the biologically inspired cognitive architectures (BICA). The system has been integrated into the ACT-R cognitive architecture, and experimentally assessed in a conceptual categorization task, where a target concept illustrated by a simple common-sense linguistic description had to be identified by resorting to a mix of categorization strategies. Compared to human-level categorization, the obtained results suggest that our proposal can be helpful in extending the representational and reasoning conceptual capabilities of standard cognitive artificial systems.
A particularly stinging criticism of the entire artificial intelligence enterprise has been its' ability to produce systems that are far more capable than a typical human on domain-specific tasks but its' striking inability to produce systems that perform some of the simplest tasks that toddlers excel at. While various proposals have been made regarding the construction of a "mechanical child," going as far back as Turing (Turing 1950) and more recently in the realm of robotics, it has been rare for developers of architectures for general intelligence to inform their work with results from child development in the abstract. We describe a cognitive architecture in terms of a set of domaingeneral functional primitives capable of succeeding on a variety of well-known tasks from the child development literature spanning the so-called "core" cognitive domains of infant physics and folk psychology. Finally, we discuss the advantages of this approach in moving forward toward modeling other sorts of higher-order cognition, including the understanding and use of natural language.