Testing a Common Model for Human and Human Like Intelligence
Abstract
Cognitive architectures are general computational modeling tools that lie at the intersection of cognitive, neuroscientific, and Artificial Intelligence research. In the cognitive neurosciences, they have been recently surged in interests as a way to frame and understand large scale brain activity. In Artificial Intelligence, they have been proposed as the next step in machine intelligence. However, despite their importance, development of cognitive architectures has so far progressed in a cumulative but a systematic and haphazard way. Inspirations for architectures are often left to the preferences of individual researchers, while empirical tests have been few and limited, and do not take advantage of contemporary neuroscientific tools that integrate both regional activity and network dynamics and rely on modern Bayesian assessment methods. This proposal aims to put the architectural approach on a solid ground by developing the methodology and software tools to systematically test theoretical architectures against large scale human neurophysiological data, using measures of both regional fit and connectivity. This methodology will then be applied to the Human Connectome Project dataset, the largest repository of high quality fMRI data, including data from 1,200 healthy young adults. In this large scale test, alternative interpretations of the brain architecture that have previously suggested and published will be compared against the Common Model of Cognition, a consensus architecture specification that has been developed by integrating the lessons learned from multiple previous computational architectures, and covers both human and human like intelligent system. A strict comparison with human data would examine the feasibility of the CMC, and eventually suggest modification and further specification. The results of the proposal will be a thoroughly validated and standardized blueprint for cognitive architectures.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 14, 2022
- Source ID
- FA95501910299
Entities
People
- Andrea Stocco
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of Washington