Building synthetic brains capable of human level discovery and invention...

Summary - A radically new form of artificial intelligence technology has been conceived based upon false memory or "confabulation" generation in artificial neural networks. Based upon the notion that ideas are nothing more than "fibs" generated and then exploited by the brain's neural networks, a whole new theory of cognition, ideation, and consciousness was born, that inspired not only a vast suite of U.S. and international patents, but also products and services for international corporations, and a whole new generation of extremely clever, improvisational robots.

Confabulation is the neural process in S. L. Thaler's theory of cognition and consciousness in which all thoughts and ideas originate in both biological and synthetic neural networks as false memories nucleate upon various forms of neuronal and synaptic fluctuations (see Imagitron and Creativity Machine). Such novel patterns of neural activation are promoted to ideas as other neural nets perceive utility or value to them (i.e., the thalamo-cortical loop). Not only did Thaler theorize about this process in the early 90s, he reduced it to practice in 1994 in his landmark U.S. Patent 5,659,666, "Device for the Autonomous Generation of Useful Information." This patent stems from Thaler's discovery that when a trained neural network is subjected to increasing levels of internal synaptic perturbation, it shifts from rote memory production to confabulation generation (see figure to right). It is in this transition region that ideas are born. In 1995, Thaler first described a universal theory of animal cognition in which novel movement planning schemes arise from various forms of synaptic fluctuations within the central nervous system of a hypothetical creature. Later (1996c) he demonstrated that both inductive and deductive inferences could autonomously form within cascades of confabulating neural networks.

Much of the science behind the technology of "useful confabulation generation" is an outgrowth of Thaler's neural network simulations of near-death experience (Thaler, 1993, 1995a, 1996b, Yam, 1995b, Hesman, 2004), wherein neural networks were used to mine the confabulation stream of other internally stimulated neural networks for interesting or useful ideas. During that time period, such thalamo-cortical network pairs autonomously fine-tuned themselves (i.e., their internal perturbation level) to optimize the turnover rate of novel, useful, and coherent ideas, leading to the plot shown to right and its theoretical verification (1997a). The notion of multiple, confabulating neural network modules chaining their activation patterns into complex deductions and inductions was reduced to practice in 1995, and presented in San Diego in 1996 at the World Congress on Neural Networks (1996c). At that time, Thaler also presented, by invitation, this theory and its applications to both a well financed neural network company and a prestigious neuroscience institute also located in that area.

Thaler's latest patent US 7,454,388, "Device for the Autonomous Bootstrapping of Useful Information" enables a new form of improvisational AI that allows robots to devise not only novel movement planning schemes, but also Machiavellian battlefield tactics. This foundational AI patent relies upon confabulation generation by artificial neural nets and subsequent reinforcement learning of those false memories showing utility or value. In 2007, the very same neural network methodology used by these extremely clever robots, was used to autonomously generate the framework for a musical album entitled "Song of the Neurons." 

Over the last decade, Thaler and his company have generated a prodigious list of practical accomplishments using what many have called "the most important idea in history"...

• invention of new personal hygiene products for a major international corporation
• theoretical discovery of "mountains" of novel and technologically valuable chemical compounds and systems that include both ultra-hard, super-magnetic and high temperature superconducting materials
• metallurgical optimization for jet engines
• invention of new weapons for the U.S. Air Force
• pre-911 data mining using Creativity Machines to invent meaning (i.e., semantic parsing) to open source intelligence
• military resource allocation and optimization
• autonomous generation of new art forms that include a whole musical album
• development of the necessary AI system to optimize bandwidth on a constellation of military communication satellites
• fully improvisational robots for the U.S. Air Force
• autonomous rendezvous and docking systems for NASA
• totally self-determining virtual worlds in which characters and entities therein develop their own complex behaviors without the assistance of human programming.
• the invention of other neural network inventions that led to even more patents
• the development of SuperNets, essentially vast swarms of self-organizing, self-connecting artificial neural networks (not ancient Bayesian networks) that autonomously interconnect to form complex brain structures capable of not only pattern recognition, but human to trans-human level discovery in machines.

These synthetic, confabulating thalamo-cortical algorithms were producing useful results as early as 1975 and are now the best bet at creating human to trans-human intelligence in machines. Writing in the journal Neural Networks, as his key confabulation patent awaited patent prosecution, Thaler's words were prophetic, "...that network now acquires a ninth, additional memory, a confabulation, if you will, ..." (1999a), signaling a revolution in the field of contemplative, creative, and perhaps conscious artificial intelligence.

References:

Hesman, T. (2004). The Machine That Invents, St. Louis Post-Dispatch, Jan. 25, 2004.

Thaler, S. L. (1993), "4-2-4 Encoder Death," Proceedings of the World Congress on Neural Networks 2 (Mahwah, NJ: Lawrence Earlbaum & Associates, 1993), pp. 180-183.

Thaler, S. L. (1994a, 2005). US Patents 5659666, 6115701, 6356884, 7454388, Device for the Autonomous Generation of Useful Information.

Thaler, S. L. (1994b). UK Patents 2336227, 2308476, Device for the Autonomous Generation of Useful Information.

Thaler, S. L. (1994c), European Patent 786106, Device for the Autonomous Generation of Useful Information.

Thaler, S. L. (1994d). Canadian Patent 219996, Device for the Autonomous Generation of Useful Information.

Thaler, S. L. (1994e). Indian Patent 193381, Device for the Autonomous Generation of Useful Information.

Thaler, S. L. (1994f). Japanese Patent 4282760, Device for the Autonomous Generation of Useful Information.

Thaler, S. L. (1995a). Death of a gedanken creature, Journal of Near-Death Studies, 13(3), Spring 1995.

Thaler, S. L. (1996a) Creativity via network cavitation – an architecture, implementation, and results, Adaptive Distributive Parallel Computing Symposium, Dayton, Ohio, 8-9 August, 1996.

Thaler, S. L. (1996b). The death dream and near-death darwinism, Journal of Near-Death Studies, 15(1), Fall 1996.

Thaler, S. L. (1996c). A Proposed Symbolism for Network-Implemented Discovery Processes, In Proceedings of the World Congress on Neural Networks, (WCNN’96), Lawrence Erlbaum, Mawah, NJ.

Thaler, S. L. (1997a). A Quantitative Model of Seminal Cognition: The Creativity Machine Paradigm, Mind II Conference, Dublin, Ireland, 1997.

Thaler, S. L. (1997b). "The Fragmentation of the Universe and the Devolution of Consciousness," U.S. Library of Congress, Registration No. TXU00775586, 1997.

Thaler, S. L. (1997c). Device for the Autonomous Generation of Useful Information: A Completely Connectionist Approach to Cognition, Creativity, and Machine Consciousness. International Conference on Vision, Recognition, Action, Neural Models of Mind and Machine, Boston University, May 28-31, 1997.

Thaler, S. L. (1998). Predicting ultra-hard binary compounds via cascaded auto- and hetero-associative neural networks, Journal of Alloys and Compounds, 279(1998), 47-59.

Thaler, S. L. (1999a). No mystery intended. Neural Networks, Volume 12, Issue 1, January 1999, Pages 193-194.

Thaler, S. L. (1999b), AFRL-ML-WP-TR-1999-4033, Integrated Substrate and Thin Film Design Methods, Materials and Manufacturing Directorate, Air Force Research Laboratory, Air Force Materiel Command, Wright-Patterson Air Force Base, OH 45433-7750

Thaler, S. L., Patrick, M. C., Stevenson-Chavis, K. (2007). Demonstration of Self-Training Autonomous Neural Networks in Space Vehicle Docking Simulations, Aerospace Conference, 2007 IEEE, Volume , Issue , 3-10 March 2007 Page(s):1 - 6.

Yam, P. (1995b). As They Lay Dying, Scientific American, May 1995.

Interesting Sites:

Elsevier | The World 2 Come | PhysForum | Wikia | gamedev.net | Akademik | 3quarksdaily | PhysOrg | Psychology Wiki | Wikipedia | A Critical Review of the Creativity Machine (our favorite)

Interesting Google Excerpts: