Integral Review

A Transdisciplinary and Transcultural Journal For New Thought, Research, and Praxis

Posts Tagged ‘Michael Lamport Commons’

Hyper Smart Developmentally Based Stacked Neural Networks and Evidence that Allows for True Androids that Pass the Turing Test

Sofia R. S. Leite and Michael Lamport Commons

Abstract: This paper describes how a mathematically-based model of cognition and species evolution, the Model of Hierarchical Complexity (MHC) can be applied to create more effective and smarter artificial intelligence that is based upon how humans and animals solve problems. To more precisely emulate how a human acts upon the environment, a computer must learn from the environment in a way that is closer to the way that humans do. Moreover, the way humans learn from the environment is an evolutionary extension of how nonhuman animals learn. Hence, to more precisely emulate a human, nonhuman animal learning should also be taken into account. To do so, the MHC proposes an analytic, a priori measurement of the difficulty of task-actions called the Order of Hierarchical Complexity (Commons & Pekker, 2008). Task-actions mean actions directed toward problem-solving. According to the MHC, task-actions grow in complexity throughout development and evolution. The definitions for what makes an action more hierarchically complex will be presented. An application of the MHC to a general artificial intelligence architecture will be presented, and an application of it to a physics problem, called the balance beam problem, will be described. Following that, the possibility of creating truly intelligent androids based on the MHC architectural concept is discussed. An android is a computer based “organism” designed to act like an animal or human. It receives signals or input through sensors (simulating afferent nerves) and acts upon the environment through output agents (simulating muscles activated through motor nerves). The notions from the MHC that will be applied to the design of androids will also be discussed. Based on these developmental and evolutionary principles, such androids will be at least as smart as humans. They will not only pass the Turing test, as will be explained, but they will also be able to complete other tests specifically designed and appropriate for humans.

Tags: , , , , , ,

Measuring an Approximate g in Animals and People

Michael Lamport Commons

Abstract: A science of comparative cognition ultimately needs a measurement theory, allowing the comparison of performance in different species of animals, including humans. Current theories are often based on human performance only, and may not easily apply to other species. It is proposed that such a theory include a number of indexes: an index of the stage of development based on the order of hierarchical complexity of the tasks the species can perform; an index of horizontal complexity; and measures of g (for general intelligence) and related indexes. This article is an early-stage proposal of ways to conceive of g in animals and people. It responds to Geary’s argument that domain-general mechanisms are essential for evolutionary psychologists. Existing research is used to enumerate domains, such as problem solving behavior in pursuit of food, or behaviors in pursuit of mates and/or reproduction, and itemize identifiable human social domains. How to construct g, across domains and within domains, is described.

Tags: , , , , , , , ,






Current Issue

Recent Issues