Figure C.1
map of semantic states (percept categories) across region of human cerebral cortex
The Problem
Imagine owning a mobile phone which understands what you mean. Any phone manufacturer which could deliver this feature would soon dominate the market, and probably the world [1]. This level of functionality is beyond the current ability of science. A lot of very smart people have been working for a very long time in an attempt to solve this problem. In this discussion, we will try to discover why this problem is so fiendishly difficult, and perhaps suggest a way forward.
Imagine that there is a serial killer in town. The witness to the most recent murder gave police a partial description of the perpetrator - a white middle aged male, wearing a black shirt. So you meet up with a good friend for coffee, and say, "Joe has a black shirt"-s(1). Your friend asks whether you should provide the police with Joes name and address.
Now imagine that your friend is applying for a job as a barista. He knows about the various types of coffee, but he doesn't have the right clothes. "Joe has a black shirt" you suggest. Your friend phones Joe, who lends him the shirt for the interview.
Lets just think about what the term 'meaning' actually means in these two quite different contexts. The 'narrow' meaning of the statement s(1) is precisely the same in both situations. There is a black shirt folded up in a drawer somewhere in Joe's home.
The 'narrow' concept of meaning fails to include the goal-oriented nature of all spoken communication. We speak for a reason, which varies depending on the situation, defined roughly as the enclosing context. By using the concept of semantic state, we can place language within a cybernetic framework. Semantic state can then be treated just like any other homeostatically regulated variable, such as temperature or pH.
The goal of a speaker is to get their listener to understand something. They achieve this by using speech (which usually has some extrinsic semantic content) to change the intrinsic semantic state of the listener. The effective computational modelling of speech (or text or any other linguistic form) without also modelling the semantic states of both speaker's and listener's minds is impossible by its very definition. Without knowing the semantic state of the speaker's mind, it is impossible to ascertain their communicative goal (the desired semantic state of the listener). Without knowing the semantic state of the listener's mind, it is impossible for the speaker to compute the necessary semantic state differential.
Very few, if any, language researchers mention these critical variables. It is therefore not a mystery that their attempts to reproduce speech and learning in a machine have failed.
1. Selling this design to Nokia would give them the chance to regain their past glories. An asking price of half a billion dollars would not be unreasonable, contingent on results.
The Solution
The GOLEM/TDE research project was started in 2010 by Miro Dyer at Flinders University in South Australia. He left his lucrative career as a movie robotics (motion control) consultant for an uncertain future as an academic heavyweight. Having gained his (unremarkable) Honours in 2012, Miro threw himself at the exciting, but daunting task of solving the most difficult puzzle of the modern period, human cognition. For a solution to be worthwhile, it must include consciousness and emotionality as intrinsic components, yet should be realised in terms of existing computational techniques and technology. To be fully credible, It should not only fix all the known problems of AI, but identify and solve some as yet unknown problems as well.
In 2020, just before the global pandemic, this research task was satisfactorily completed. Consciousness and emotionality are integral, and understandable, parts of the solution, which describes a mechanism of mind whose component parts are analogous to those in real brains.
Figure C.1 depicts a real brain map, clearly showing that it is topographically organised according to semantic states (perceptual categories).
Figure C.2 depicts a neocybernetic neuron, the building block from which all animal and human minds can be built. Note that, although its connections are nominally feedforward in nature, it is about as far from a conventional neural network design as it is possible to be, utilising operating principles and mechanisms that are as subtle as they are novel. For greater understanding of the key network design differences, see the following websites. [3]
www.conscious-computation.webnode.com
Figure C.3 depicts the TDE (a recursive acronym meaning 'TDE Differential Engine'). The TDE clearly demonstrates how GOLEM (an acronym for Goal-Oriented Linguistic Emulation of Mind) theory produces reliable and accurate predictions for locations of well known neurocognitive features. For example, it correctly predicted that right cerebellar lesions would yield Broca's-like aphasias [2].
2. Silveri
3. Zero-cost websites such as Wix, Webnode and Simplesite were chosen as the best publication mechanism following established Open Science principles. They have the advantage that research findings are made available immediately, and transparently, without a journal editor or publication cycle to contend with. They have the obvious disadvantage of lack of peer review, unfortunately, which places the onus of academic rigor and intellectual honesty squarely on the shoulders of the author.
Figure C.2
Figure C.3