Saturday, May 20, 2017


AutoGnomic (AG) vs. Artificial Intelligence (AI)

         A rough line drawn through the history of AI starts with early attempts to mimic the exquisite complexity of logical reasoning in the early years, and passes through the most recent trends in industrial scale engineering applications afforded by modern processing power and access to digital data. Two of the biggest centers of attention in AI at the moment are Google and IBM. Google's recently announced public release of it's TensorFLow machine learning platform underlying the GoogleBrain project, and newsworthy successes such as learning about cats from unlabeled YouTube videos, demonstrate the power of large-scale computing applied to so-called "deep learning". Deep learning typically refers to artificial neural network (ANN) architectures that are stacked in increasing numbers of layers, where such deeper layers model more abstract features of the problem space than shallower ANNs which more directly relate their inputs and outputs.  Modern computing capacity has enabled ANNs to grow in depth as the inherent inefficiencies in the brute-force training of ANNs layers are overcome. Deep learning performance in narrow applications such as machine vision, speech recognition, and bioinformatics has been impressive and of substantial commercial value. 

On the other hand, the distance between contemporary deep learning systems and other presumed characteristics of mind and intelligence, such as performing logical reasoning, carrying on a conversation with another intelligent being, or explaining itself, is as great as ever in the history of AI. For AI that aspires to collaborate with and enhance human intelligence as a partner, deep learning offers little hope at present. On the other hand, IBM's Watson technology, made famous in its success in TV's Jeopardy! competition is specifically designed to work in the space of human intelligence. Branching out from competing with humans in game shows, Watson now provides advice to on-line shoppers and treatment recommendations to cancer doctors. Watson is specifically designed to respond to questions put to it by humans, in human language, with answers drawn from sources created by and for humans such as wikipedia and dictionaries. Rather than through any "deep" learning or reasoning, Watson's power comes from it's ability to index and search vast amounts of information based on statistical similarities grounded in keywords and phrases extracted from the question. While providing a valuable and impressive capability for the narrow activity it was designed for--answering questions--it is generally accepted that Watson offers little insight or capability in the wider range of intelligent activity. Like deep learning, it is a showcase for the practical potential of focusing vast computing power and data on a narrow problem.

"Real" intelligence--if humans are to be an initial guide--includes a broad range of capabilities such as creativity, resourcefulness, social cooperation, and so on, that are far from the reach of the high-profile, commercially compelling approaches coming from Google, IBM, and many others trying to emulate their technical success and revenues. But that does not mean there are not other opportunities as impressive and potentially profitable. One such application is in collaborative man-machine communities of inquiry, such as scientific and special interest communities. The type of intelligence required to participate in such a community includes full participation in the constant use and refinement of language to reflect the growing collective knowledge and experience of the community. It also requires the ability to learn and understand the differences between individuals and how to interact with them as individuals.

 Deep learning is too far removed from humanity, and Watson is too crude in it's brute statistics, to meet these needs. The AutoGnome, on the other hand, has charted it's own path through the history of AI. Although conceived near the beginning of modern AI history (circa 1960), its developers soon realized the limitations of the hand-crafted reasoning systems in those early years and instead adopted a large-scale learning approach that would only sound familiar nearly a half-century later in the "big data" and machine learning approaches of today. On the other hand, the AutoGnome contrasts with both deep learning and Watson-like approaches in modeling the underlying capability that enabled humans to evolve to reason, learn, and cooperate as they do—the capability to create, use, and refine a system of signs by and for itself. This capability makes the AutoGnome a superior approach to intelligent machines that will work closely with humans, first learning from them, then emulating them, and finally partnering with them in advancing knowledge. What the AutoGnome does share with GoogleBrain and Watson are the tremendous potential benefits of rapidly increasing computing power and information access. The window of opportunity is as open as it ever has been for commercialization of the AutoGnome, combining the demonstrated power of modern technology underlying the success of deep learning, addressing human needs in the complex modern world being partly but profitably met by Watson, but in applications not remotely addressed by either.
Over the past five decades, dating to the late 1950’s, an emerging trio of geographically and functionally independent researchers from the Linguistic Research Center, University of Texas at Austin (Pendergraft & Hirst) and from Johns Hopkins University and the Research Institute for Advanced Studies, Baltimore, MD (Hamann) coalesced as a distributed AG-founders team from 1979 to 1989 to focus on the completion of the formulation of the Specification for the AutoGnome (AutoGnomic Design Specification ) which led to the inception and formation of the AutoGnomics Group of Companies (AGC). The AGC, loosely formed beginning around 1990 by the Hirsts, Hamann, Pendergraft, and Reed, to develop and technologically implement a Standard Theory of Mind (the AutoGnome). The founding corporate parents of the AGC are The AutoGnomics Institiute, Inc.(TAI) and AutoGnomics Corporation (AC). AC has been inactive since 2010; hence, ThoughtSigns, Inc., (TSI) was subsequently formed as the prime managing director of AutoGnomics Theory and Technology Development, including aggregation and reformation of the now varied and global AutoGnomic and related development efforts into a concerted “Project AutoGnomics--Center of MIND” and AG Tech Apps., all Formulated on the Foundations of AutoGnomic MultiCoRelational Systems (MCRS)]. Key current AGC Partnering Companies include: First Peoples Capital Corporation (FPCC) and TrueFarmer, Inc. (TFI), focused on Indigenous and Rural Cultures, respectively, and StartSmartR Capital Corporation focused on the Global Metro cultures.

Since the initial founding of the AGC in 1994, through 2010, AG technology has realized investments (partly directly and partly indirectly through key affiliates) of approximately $15 million applied to its development efforts. This has led to 5 issued patents on semiotic-based AutoGnomic Processing Systems via a joint venture (UniGnomics) with Unisys Corporation (1997-2000). Also, through the continued support by Unisys from 2000 to 2008, a group led by one of the AutoGnomics/Unisys team secured over 40 additional issued patents on the AutoGnomic Data Storage System. All of this latter Data Storage System Intellectual Property was licensed back by Unisys to the AGC on August 28, 2010 and was subsequently traded to retire about $5 million of third party stockholdings. Approximately $10 million of the $15 million was derived from the UniGnomics effort with Unisys Corporation. From 2002 through 2010, the remaining $5 million of the $15 million was raised via four sequential private placements with the proceeds focused on the development and commercialization of a MyWebGnome product branded as “TrueThinker”.

Coincident with the wrapping up of UniGnomics at the end of 1999, Unisys exercised a warrant to purchase common stock in AutoGnomics; this transaction set an implied AG value at $41 million. Although the various AG software IP and related patents, copyrights and knowhow are a significant contribution to AG value, they are incidental in comparison to the 585 page proprietary Autognomic Design Specification. This latter has involved an uncountable number of hours of R&D effort under the AG-founders team from 1959 through 1989 with the concentrated work being from 1979 to 1999. This project has produced what arguably remains as the only publicly known globally existent experiential foundation for a standard theory of mind with a detailed conceptual specification for building a synthetic mind. Assuming a baseline value around $50 million in 1999 and conservatively assessing the value-added during the intervening decade and a half since, it is not unreasonable to set an estimated value in the order of $500 million to $1 billion based on present values of ongoing commercial development pro formas. 

No comments: