HOME OF THE CREATIVITY MACHINE
The Big Bang of Machine Intelligence!
The IEI Blog is now open for discussion of new business ventures and opportunities!
for the Autonomous Generation of Useful Information*:
Completely Connectionist Approach to Cognition, Creativity, and Machine
S.L. Thaler, Imagination Engines, Inc.
Conference on Vision, Recognition, Action, Neural Models of Mind and Machine,
Boston University, May 28-31, 1997
By progressively increasing the magnitude of perturbation applied to the weights of a trained feedforward network, we observe a succession of output vectors that range between verbatim instances of training exemplars to utter nonsense. In the transition between these two extremes we observe a chaotic regime rich in unusual twists on exemplar concepts, each offering various degrees of utility and appeal. By cascading this chaotic network with a critic network that may observe and filter potentially valuable concepts, we form what has been coined a “Device for the Autonomous Generation of Useful Information” or “DAGUI.” This system has been shown to be not only a valuable research tool across all disciplines, but also a compelling model of mind. In this paper I will present the cumulative evidence that ties this connectionist architecture to human cognition, consciousness, and creativity in a three distinct stages:
(1) I demonstrate how such devices may be parametrically tuned to achieve maximum generate and search efficiency, and hence prescribe the most beneficial level of internal perturbation administered to such nets. Having found the ideal range for internal perturbation, we find that the most valuable network discoveries are in fact confabulatory states induced by instantaneous network degradation. This tempting analogy, combining static neurobiology with ubiquitous chaos, now suggests an intriguing cognitive model for which we seek additional corroborating evidence.
(2) Corroboration comes in the form of a rigorous statistical study comparing the temporal distribution of concept formation in both the tuned DAGUI and human volunteers. We discover an exact correspondence, with the artificial neural network cascade generating concepts with a prosody or rhythm identical to that of human thought generation. Accepting the equivalence between the artificial and biological paradigms, we see persuasive evidence that the modality of thought, either rote memory recall and novel idea formation may be simply described by the levels of perturbation involved.
(3) Finally, contact is made with the theories of consciousness focusing on the cortical-NRT (Nucleus Reticularis Thalami) interplay. Quantitative similarities are seen between experimentally observed “bubble” formation time scales in human cortex and the lifetime of synthetically introduced perturbations within the tuned DAGUI.
Related Work by S.L. Thaler:
Network “cavitation” in the modeling of consciousness, Toward a Science of Consciousness, Tuscon, AZ, March, 1996.
Neural Networks That Create and Discover, PCAI, May/June, 1996.
Creativity via Network Cavitation - An Architecture, Implementation, and Results, Adaptive Distributive Parallel Computing Symposium, Dayton, Ohio, 1996.
Is Neuronal Chaos the Source of Stream of Consciousness?, World Congress on Neural Networks, San Diego, 1996.
© 1997-2013, Imagination Engines, Inc. | Creativity Machine®, Imagination Engines®, and DataBots® are registered trademarks of Imagination Engines, Inc.
1550 Wall Street, Ste. 300, St. Charles, MO 63303 • (636) 724-9000